Bulletin of the World Health Organization

Viral suppression after 12 months of antiretroviral therapy in low- and middle-income countries: a systematic review

James H McMahon a, Julian H Elliott a, Silvia Bertagnolio b, Rachel Kubiak c & Michael R Jordan c

a. Infectious Diseases Unit, Alfred Hospital, Level 2 Burnet Tower, 85 Commercial Road, Melbourne, 3004, Australia.
b. HIV Department, World Health Organization, Geneva, Switzerland.
c. Department of Public Health and Community Medicine, Tufts University School of Medicine, Boston, United States of America.

Correspondence to James H McMahon (e-mail: ja.mcmahon@alfred.org.au).

(Submitted: 16 September 2012 – Revised version received: 01 January 2013 – Accepted: 21 January 2013 – Published online: 21 February 2013.)

Bulletin of the World Health Organization 2013;91:377-385E. doi: http://dx.doi.org/10.2471/BLT.12.112946

Introduction

At the end of 2011, over 8 million people in low- and middle-income countries (LMICs) were receiving antiretroviral therapy (ART).1 The guidelines of the World Health Organization (WHO) for the treatment of human immunodeficiency virus (HIV) infection recommend that, where possible, the viral loads of individuals receiving ART be measured every 6 months to detect viral replication and confirm treatment failure whenever it occurs.2 Although viral load tests are currently too costly for routine use in many LMICs, the potential for increased access to such tests exists as costs decrease and countries prioritize this method of patient monitoring.3

WHO’s guidelines for the treatment of HIV infection recommend that a viral load of > 5000 copies of viral ribonucleic acid (RNA) per ml be taken as indicative of virological failure.2 According to WHO’s strategy for the surveillance and monitoring of HIV drug resistance in LMICs, a viral load of < 1000 RNA copies per ml should be taken as evidence of viral suppression.4 Guidelines for the treatment of HIV infection in high-income countries stipulate that a viral load of < 50 RNA copies per ml – or a load below the limit of detection of the most sensitive assay available – be taken as evidence of viral suppression,57 and that a load of ≥ 50 RNA copies per ml,5,7 or one of ≥ 200 viral RNA copies per ml confirmed by repeat testing,6 be used as evidence of virological failure or rebound.

The proportion of a study cohort showing viral suppression is calculated as the number of patients with viral suppression divided either by the number of patients in the study cohort who began ART (i.e. the intention-to-treat population), or by the number of patients in the cohort who are alive and on treatment (i.e. the on-treatment population). On-treatment analyses reflect the effectiveness of ART for those receiving antiretroviral drugs. Intention-to-treat analyses, which use a denominator that includes individuals who die or are lost to follow-up during the study period, reflect factors at the individual or programme level that influence the risk of death and disengagement from care. Relatively high mortality in the first 6 to 12 months of ART and substantial loss to follow-up, both of which have been widely reported in LMICs, can therefore influence estimates of viral suppression based on intention to treat.8,9

Summary estimates of viral suppression (as measured, for example, 12 months after ART initiation) are needed to guide ART programme managers on the normative levels of population-level viral suppression and to define desirable levels of clinic and programme performance. Such estimates are also useful when creating and improving mathematical models of the different strategies that might be followed to provide ART in LMICs. Recent reviews of virological outcomes have focused on sub-Saharan Africa and levels of acquired resistance to antiretroviral drugs.10,11 Across LMICs, summary estimates of viral suppression based on different HIV-RNA thresholds are lacking. The objective of this systematic review was to establish estimates, based on different viral RNA thresholds, of the percentages of the intention-to-treat and on-treatment populations in LMICs that show viral suppression 12 months after ART initiation.

Methods

Study selection

We included publications (“papers”) or conference abstracts that reported the proportion of individuals in a study cohort from an LMIC for whom a virological outcome after 12 months of ART was reported, either as a primary or a secondary finding. If only the median duration of follow-up was reported for a study, that study was included in the review provided the median duration was between 9 and 15 months. Any definition of viral suppression (or failure) reporting a proportion of patients below (or above) a defined viral RNA threshold was accepted. If the threshold was not reported, the relevant authors were contacted and asked for details of the threshold that they had used. A study was excluded if (i) the threshold used could not be determined; (ii) only changes in viral RNA loads from the baseline values were reported; (iii) most of the patients were less than 13 years old; (iv) patients received only one or two antiretroviral drugs; or (v) the study was not from an LMIC. All the studies included in the review were published in English and were either clinical trials or cross-sectional or cohort in design.

Search strategy

We searched for relevant articles published between 1 January 2003 and 31 May 2011 through Ovid MEDLINE. Online databases containing the abstracts of presentations made at the International AIDS Society Conferences held in 2009–2010 and at the Conferences on Retroviruses and Opportunistic Infections held in 2009–2011 were also searched.12,13

Our search strategy combined the relevant medical subject headings (MeSH) with additional search terms to identify those studies that reported virological outcomes for HIV-infected participants receiving ART. We used other MeSH and search terms to identify studies from LMICs in which outcomes after 12 months of ART were investigated. When more than one article reported data from the same cohort of patients and used the same viral RNA threshold, only the article that contained the most detailed information was selected. Fig. 1 summarizes the search strategy and study selection process. The study protocol provides further details.14

Fig. 1. Search strategy and study selection used in the systematic review
Fig. 1. Search strategy and study selection used in the systematic review
Note: We excluded duplicate reports, studies on subjects not receiving antiretroviral therapy, studies with patients younger than 13 years or not infected with human immunodeficiency virus or not living in a low- or middle-income country, and reports that provided insufficient information.

The following data were abstracted from each study: first author, year of publication, study country or countries, health-care setting (i.e. public sector, private sector or nongovernmental organization), whether the study patients had to pay for their ART, dates the patients were observed, number of study sites, number of patients receiving ART, baseline demographics (i.e. age, gender, CD4+ T-lymphocyte count and clinical stage), ART regimen, whether patients were ART-naive at baseline, study definition of virological outcome, and the proportion of patients meeting that definition 12 months after initiating ART. When available, the percentages of subjects who died, transferred out, stopped ART or were lost to follow-up were also abstracted. Whether the proportion of the study cohort reported to have viral suppression was based on an on-treatment or intention-to treat analysis was noted. If intention-to-treat values had not been reported, they were calculated from the raw data (when available). In these calculations, the number of patients who had died or been lost to follow-up at the time of the estimation of virological outcome and, when available, the number that had stopped ART were included in the denominator. Patients who transferred out of the study cohort were excluded from the denominator, consistent with the indicators of retention in care recommended by WHO and the United States President’s Emergency Plan for AIDS Relief (PEPFAR).1518 If only the fraction of patients showing virological failure was reported, we calculated from this figure the proportion with viral suppression.

Data analysis

Proportions (%) of patients meeting the study definition of viral suppression were derived from text, tables or, if they could be accurately determined or estimated in this manner, from published graphs. Summary estimates were determined and categorized as on-treatment or intention-to-treat values. We assessed heterogeneity among proportions by calculating the I2 statistic. If heterogeneity was high (I2 > 75%), we pooled proportions using the Freeman–Tukey method with a random-effects model and DerSimonian–Laird weights.19,20 When possible, we determined summary estimates of viral suppression using different threshold ranges for viral RNA: ≥ 1000, 300–500 and/or ≤ 200 copies per ml. Additionally, we calculated summary estimates separately for studies in which one or two viral load tests were used per patient to define virological outcomes. We performed subgroup analyses to examine the influence of the year of ART initiation and of the time taken to publish the research findings (i.e. the “time-lag bias”) on the summary estimates for the on-treatment populations. Year of ART initiation was defined as the median year of ART initiation and reported as pre-2004, 2004–2005 or post-2005. The time from the median year of ART initiation until publication was categorized as ≤ 3, 4 or ≥ 5 years. Analyses were conducted using Excel (Microsoft, Redmond, United States of America) and version 2.7.9 of the StatsDirect software package (StatsDirect, Altrincham, United Kingdom of Great Britain and Northern Ireland).

Results

Overall, 49 studies (38 papers2158 and 11 conference abstracts5969), together comprising 48 cohorts and 30 016 individuals, were identified for inclusion (Fig. 1). Details of the cohorts from the papers and conference abstracts are presented in Table 1 and Table 2 (available at: http://www.who.int/bulletin/volumes/90/5/12-112946), respectively. Two reports29,52 described outcomes from the same cohort but made use of two different thresholds to define viral suppression. A further seven reports26,27,43,44,47,53,55 described outcomes using two or more thresholds of viral suppression. The data from another two papers24,25 were combined to obtain an estimate of viral suppression in a single cohort. The 48 cohorts in the systematic review comprised 43 single-country cohorts – 37 from sub-Saharan Africa, 3 from Asia and 3 from Latin America or the Caribbean – and 5 multi-country cohorts (of which 4 were from sub-Saharan Africa).

All but one study68 used a single viral load measurement per patient to define the virological outcome. The ART regimens used were reported for 39 of the studies. In each of 35 (90%) and 29 (74%) of these studies, at least 50% and 95% of the patients, respectively, had received a regimen based on a non-nucleoside reverse-transcriptase inhibitor (NNRTI). Most of the patients in another three studies – a clinical trial studying the efficacy of boosted protease-inhibitor regimens26 and two studies from regions with high prevalences of HIV-2 infection35,45 – had received regimens based on a protease inhibitor.

Summary estimates from on-treatment and intention-to-treat analyses are presented – for those studies in which outcomes were defined using a single viral load measurement– in Table 3 and Fig. 2. As high levels of heterogeneity were observed (I2 > 90%), all summary estimates and 95% confidence intervals (CI) were calculated using a random-effects model.

Fig. 2. Viral suppression after 12 months of antiretroviral therapy in low- and middle-income countries
Fig. 2. Viral suppression after 12 months of antiretroviral therapy in low- and middle-income countries
ITT, intention-to-treat; OT, on-treatment; RNA, ribonucleic acid. Note: The data shown summarizes OT or ITT analyses in which viral suppression was defined by a viral load – measured as copies of human immunodeficiency virus ribonucleic acid (HIV-RNA) per ml of plasma – that fell below a threshold value.

The summary estimate of the proportion of patients showing viral suppression, for all viral load thresholds in 43 on-treatment analyses, was 84.0% (95% CI: 81.3–86.6; n = 26 599). For the nine cohorts in which viral suppression was defined as < 1000 viral RNA copies per ml, 83.5% (95% CI: 77.8–88.4; n = 3192) of the combined on-treatment populations showed suppression. Thresholds ranging from 300 to 500 copies per ml were used for 32 cohorts (a threshold of 400 copies per ml was used for 27 cohorts) and 84.3% (95% CI: 80.4–87.9; n = 25 708) of the combined on-treatment populations of these 32 cohorts showed suppression. The reported outcomes for nine and six cohorts were based on thresholds set at or below 200 and at or below 50 viral RNA copies per ml, respectively. When thresholds set at or below 200 copies per ml were used, 76.1% (95% CI: 66.8–84.3; n = 2167) of the combined on-treatment populations showed suppression.

The summary estimate of viral suppression for all 27 cohorts (n = 13 134) in intention-to-treat analyses was 71.2% (95% CI: 66.5–75.7). The corresponding value for the four analyses in which a threshold of 1000 copies per ml was used was 77.5% (95% CI: 67.6–86.1; n = 1201). In intention-to-treat analyses of the data from 21 and 19 cohorts, thresholds of 300–500 and 400 copies per ml, respectively, were used. With thresholds that ranged from 300 to 500 copies per ml, 70.5% (95% CI: 65.2–75.6; n = 11 528) of the combined intention-to-treat populations showed viral suppression. Six studies reported the results of intention-to treat analyses based on thresholds set at or below 200 copies per ml – four studies used a threshold set at or below 50 copies per ml – and these produced a summary estimate for the frequency of viral suppression of 62.9% (95% CI: 51.2–73.8; n = 1654).

To investigate possible sources of heterogeneity, we performed subgroup analyses exploring the median time of ART initiation and the potential bias introduced by the time taken to publish results: the I2 values for the subgroups considered were all > 90%. When the median year of ART initiation was pre-2004, 2004–2005 and post-2005, the on-treatment estimates of suppression after 12 months of ART were 80.9% (95% CI: 73.0–87.7), 84.1% (95% CI: 78.3–89.2) and 84.3% (95% CI: 80.0–88.2), respectively. The corresponding values for delays of ≤ 3, 4 and ≥ 5 years between the median year of ART initiation and the year of the publication of results were similar: 83.3% (95% CI: 78.0–88.0), 84.0% (95% CI: 78.7–88.6) and 83.0% (95% CI: 76.9–88.3), respectively.

Discussion

This is the first systematic review to quantify population-level viral suppression 12 months after ART initiation in LMICs, and to stratify estimates by viral RNA thresholds in on-treatment and intention-to-treat analyses. Over 70% of the cohorts included in the review were investigated using thresholds ranging from 300 to 500 viral RNA copies per ml and, after 12 months of ART, viral suppression was noted in 84% (on-treatment) or 71% of patients (intention-to-treat). These summary estimates compare favourably with outcomes reported in high-income countries after 12 months of NNRTI-based ART, such as the 58–73% viral suppression seen in intention-to-treat analyses in early clinical trials and in a meta-analysis at thresholds set from 400 to 500 viral RNA copies per ml.7072 Observational data from Canada, the United Kingdom and the United States of America also indicate similar frequencies of viral suppression after 12 months of ART. For example, when a viral load threshold of 50 copies per ml was used, 82% of an on-treatment population investigated in the United Kingdom showed viral suppression.73 In Canada and the United States, in on-treatment analyses based on suppression thresholds between 500 and 1000 copies per ml, 60–63% of individuals initiating ART when they had CD4+ T-lymphocyte counts of < 200 cells per µl were found to have attained viral suppression 12 months later.74,75

As anticipated, summary estimates for viral suppression were found to be higher in the on-treatment analyses than in the intention-to-treat analyses and to increase as the viral RNA thresholds used to define suppression increased. Individuals who are lost to follow-up in studies on the efficacy of ART – and included in intention-to-treat analyses but excluded from on-treatment analyses – are defined as not having achieved viral suppression. In addition, as the viral load thresholds for suppression are increased, increasing numbers of patients with low-level viraemia are categorized as cases of viral suppression.

We found no evidence of studies in LMICs that had used viral load thresholds of 10 000 or 5000 copies per ml – as recommended in the ART guidelines published by WHO in 20062 and 2010,76 respectively – to define viral suppression. The reasons for this are unclear. Investigators in LMICs may simply have preferred to use the lower thresholds supported either by the national ART guidelines used in the country where the study was performed or by guidelines not specifically intended for use in a public health model of care.57 They may also have wished to use thresholds based on the lower limit of sensitivity of the viral load assay that they had available.

Only one study68 reported the frequency of viral suppression based on two viral load measurements per patient. In this case, use of a second test increased the percentage of patients showing viral suppression. A second test is particularly likely to increase the percentage of patients who are virologically suppressed if an intervention to improve adherence to ART occurs after an initial detectable viral load, in a strategy that is recommended by WHO2 and already followed in several LMICs.33,77,78 Given that the overwhelming majority of reported outcomes are based on a single viral load measurement per patient, efforts to understand and summarize virological outcomes in LMICs should also be based on a single test result for each patient.

As far as possible, this review used intention-to-treat estimates of viral suppression that, as recommended in the relevant international guidelines,1518 excluded individuals who transferred out of the included studies. Individuals were reported to have transferred out of only six cohorts included in the review and, in each case, such individuals were excluded from the intention-to-treat estimate of viral suppression.21,24,25,38,46,48,49 However, only two of the reports included in the review specifically stated that no patient had transferred out.44,53 For the other 19 cohorts with intention-to-treat estimates of viral suppression, no data on transfers out were presented. This lack of data left it unclear whether any patients had transferred out and, if so, whether such patients had been incorporated in the denominator used in the final analyses. If any patients who did transfer out were unreported and still included in the denominators used in the final analyses, the intention-to-treat summary estimates generated in this review may be too low.

The summary estimates presented in this systematic review are important for target-setting and benchmarking. They provide guidance to ART clinics and programmes on the mean rates of viral suppression achieved in LMICs (normative referencing). The managers of ART programmes may define adequate levels of programme performance as those that lead to levels of viral suppression that match or exceed the summary estimates (criterion referencing).79 A combination of normative and criterion referencing methods, incorporating the data summarized in this review, may be used to categorize poor, intermediate and optimal levels of performance. In its strategy for the surveillance and monitoring of resistance to antiretrovirals, WHO recommends that ART treatment sites achieve viral suppression (as indicated by a viral load of < 1000 copies per ml) in at least 70% of the intention-to-treat population after 12 months of ART.80 Although data from this review are limited, this target should probably be increased since, in the four reviewed studies that used the same viral load threshold (n = 1201), 78% of the intention-to-treat population achieved suppression. Although detectable viral RNA after ART does not prove the presence of resistance to antiretrovirals, individuals who are virologically suppressed on ART have no effective drug resistance. Detectable viral RNA in populations receiving ART is often associated with suboptimal adherence to ART, which is predictive of the emergence of drug-resistant strains of the virus.65

The estimates presented in this review may not be truly representative, despite being mean or normative levels of viral suppression. For example, ART programmes or clinics that evaluated and reported viral loads may have greater resources and better clinical outcomes than those where viral loads were not evaluated and/or where virological outcomes were not reported. Additionally, in settings where viral loads were evaluated but where poor virological outcomes were observed, researchers may have chosen not to disseminate the results, and this may have led to publication bias. The summary estimates may therefore overestimate the mean frequency of viral suppression in the broader population receiving ART in LMICs. In addition, high levels of statistical heterogeneity (I2 > 90%) were observed during the review. Analyses that focused on year of ART initiation and the delays between the recording and publishing of results revealed persistent heterogeneity but no major differences in the proportions achieving viral suppression between the subgroups considered. These findings suggest that other, unidentified factors are potentially contributing to the between-study variation seen in the proportions of viral suppression. Despite these limitations, the systematic method used to identify studies and the statistical methods used to generate summary estimates allow for a reliable estimate of viral suppression rates based on the data available from LMICs. Another potential limitation of the present review is that most of the data investigated came from studies that used thresholds between 300 and 500 copies per ml. Relatively few results from studies based on thresholds set at 1000 or at or below 200 viral RNA copies per ml were available. In the on-treatment analyses, the summary estimate of suppression seen with a threshold of 1000 copies per ml was similar to that seen with thresholds in the range of 300 to 500 copies per ml. However, only nine cohorts were included in the calculation of viral suppression at the high threshold (n = 3192), whereas 32 cohorts (n = 25 708) were included in the calculation at the lower thresholds. The summary estimates established at 1000 copies per ml may have been too underpowered to demonstrate a difference with the estimates established at thresholds between 300 and 500 copies per ml.

In conclusion, this is the first systematic review of viral-suppression rates from LMICs after 12 months of ART. It includes summary estimates, at multiple HIV RNA thresholds, based on both on-treatment and intention-to-treat analyses. At the most commonly reported viral RNA thresholds (i.e. 300–500 copies per ml), approximately 71% of the patients in the intention-to-treat analyses and 84% of those in the on-treatment analyses had attained viral suppression 12 months after ART initiation. These proportions compare favourably with outcomes observed in high-income countries and represent a substantial achievement for LMICs, where ART is generally provided under great resource constraints.

The data reported in this review have important public health implications. Researchers and managers of ART programmes in LMICs could use these results to support mathematical models of the effects of ART and set rates of viral suppression as performance targets. Use of these targets would help identify those ART clinics with suboptimal performance that would most benefit from focused interventions to improve service delivery and patient outcomes.


Acknowledgements

The authors alone are responsible for the views expressed in this publication, which do not necessarily represent the decisions or stated policies of the World Health Organization. JHM has a dual appointment with the Department of Public Health and Community Medicine, Tufts University School of Medicine, Boston, USA. JHE has dual appointments with the Department of Medicine, Monash University, Melbourne, Australia and the Burnet Institute, Melbourne, Australia. MRJ has a dual appointment with the Division of Geographic Medicine and Infectious Diseases, Tufts Medical Center, Boston, USA.

Funding:

The authors received support via a Postgraduate Scholarship from the Australian National Health and Medical Research Council (JHM) and from the United States National Institutes of Health 5K23AI074423-04 (MRJ.)

Competing interests:

None declared.

References

Share