INTERVENTIONS TO PROMOTE ADHERENCE TO ANTIRETROVIRAL THERAPY IN AFRICA: A NETWORK META-ANALYSIS

Sunday, 30th of November 2014

INTERVENTIONS TO PROMOTE ADHERENCE TO ANTIRETROVIRAL THERAPY IN AFRICA: A NETWORK META-ANALYSIS

The Lancet HIV, Volume 1, Issue 3, Pages e104 - e111, December 2014

Edward J Mills PhD a , Richard Lester MD b, Kristian Thorlund PhD a c, Maria Lorenzi MSc a, Katherine Muldoon PhD b, Steve Kanters PhD b, Sebastian Linnemayr PhD d, Robert Gross MD e, Prof Yvette Calderon MD f, K Rivet Amico PhD g, Harsha Thirumurthy h, Cynthia Pearson PhD i, Prof Robert H Remien j, Lawrence Mbuagbaw c, Prof Lehana Thabane PhD c, Michael H Chung MD k, Prof Ira B Wilson MD m, Albert Liu MD n, Olalekan A Uthman PhD o, Prof Jane Simoni PhD l, Prof David Bangsberg MD p, Sanni Yaya PhD r, Till Bärnighausen MD p q, Nathan Ford PhD s, Prof Jean B Nachega MD t

Best viewed at http://www.thelancet.com/journals/lanhiv/article/PIIS2352-3018%2814%2900003-4/fulltext?_eventId=login

Summary

Background

Adherence to antiretroviral therapy (ART) is necessary for the improvement of the health of patients and for public health. We sought to determine the comparative effectiveness of different interventions for improving ART adherence in HIV-infected people living in Africa.

Methods

We searched for randomised trials of interventions to promote antiretroviral adherence within adults in Africa. We searched AMED, CINAHL, Embase, Medline (via PubMed), and ClinicalTrials.gov from inception to Oct 31, 2014, with the terms “HIV”, “ART”, “adherence”, and “Africa”. We created a network of the interventions by pooling the published and individual patients data for comparable treatments and comparing them across the individual interventions with Bayesian network meta-analyses. The primary outcome was adherence defined as the proportion of patients meeting trial defined criteria; the secondary endpoint was viral suppression.

Findings

We obtained data for 14 randomised controlled trials, with 7110 patients. Interventions included daily and weekly short message service (SMS; text message) messaging, calendars, peer supporters, alarms, counselling, and basic and enhanced standard of care (SOC). Compared with SOC, we found distinguishable improvement in self-reported adherence with enhanced SOC (odds ratio [OR] 1·46, 95% credibility interval [CrI] 1·06—1·98), weekly SMS messages (1·65, 1·25—2·18), counselling and SMS combined (2·07, 1·22—3·53), and treatment supporters (1·83, 1·36—2·45). We found no compelling evidence for the remaining interventions. Results were similar when using viral suppression as an outcome, although the network contained less evidence than that for adherence. Treatment supporters with enhanced SOC (1·46, 1·09—1·97) and weekly SMS messages (1·55, 1·01—2·38) were significantly better than basic SOC.

Interpretation

Several recommendations for improving adherence are unsupported by the available evidence. These findings can inform future intervention choices for improving ART adherence in low-income settings.

Funding

None.

Introduction

Antiretroviral therapy (ART) has clinical and public health benefits by decreasing HIV morbidity and mortality and transmission to sex partners.1 Many patients experience difficulties in taking ART at some time and may take it only sporadically or take drug holidays.2 There are many possible reasons for not taking ART, including social, personal, and structural factors.3, 4 Promotion of adherence to ART is one of the chief public health concerns for populations living with HIV.5

Despite the importance of achieving and maintaining high rates of ART adherence, few interventions have proved successful among those experiencing difficulties.6, 7 In Africa, where most people with HIV infection live, specific social, structural, or health-system-related barriers exist including food insecurity, stigma, supply-chain interruptions, and a lack of human health resources.8 Previous systematic reviews have identified potentially effective interventions, but have not assessed their effectiveness with statistical measures.7, 9, 10

The past decade has seen important progress in evidence synthesis, particularly with the popularisation of network meta-analysis.11—14 In traditional meta-analysis, all included studies compare the same intervention with the same comparator. Network meta-analysis extends this concept by including multiple pairwise comparisons across a range of interventions and provides estimates of relative treatment effects on multiple comparisons for comparative effectiveness purposes on the basis of direct or indirect evidence. Direct evidence for the effect of treatment B compared with A would correspond to the evidence familiar to us in pairwise meta-analysis, combining all head-to-head comparisons. Indirect evidence corresponds to all comparisons of B and A through common comparators, such as standard of care. Thus, network meta-analysis allows for inference between two interventions even in the absence of head-to-head evidence. The conditions required for these analyses resemble those of traditional meta-analysis; although they require that direct and indirect evidence be in agreement, a condition called consistency. Therefore, we aimed to investigate what ART adherence interventions have been used in the African setting. We used a network meta-analysis approach to draw from both direct and indirect evidence from randomised trials.

Methods

Search strategy and selection criteria

This study has been designed and reported according to the pending Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) extension to network meta-analyses.15 The protocol for this study is available from the authors upon request.

To be included, RCTs must have had an intervention targeted to increase ART adherence, had at least 3 months follow-up, and reported adherence as an outcome. We restricted trials to African countries to avoid issues of dissimilarity that arise from variations in HIV risk groups (table 1).

Table 1Table image

Population, interventions, comparisons, outcomes and study design (PICOS) criteria for study inclusion

We searched medical literature for relevant trials that described interventions to improve ART adherence among patients with HIV, with the terms “HIV”, “ART”, “adherence”, and “Africa”. We searched using several electronic databases: AMED, CINAHL, Embase, Medline (via PubMed), and ClinicalTrials.gov from inception to Oct 31, 2014. The complete search strategy used to identify studies is available in the appendix (p 1). Two investigators (KM, ML) reviewed all abstracts and full-text articles. We contacted all study authors and requested the individual data on patients achieving adherence and viral suppression. We did not set any restriction based on publication date or language and included all studies available as of Oct 31, 2014.

Data extraction and variable definitions

Using a standard data sheet, we extracted the following data from articles that met the inclusion criteria: trial duration, trial location, year of publication, rate of loss to follow-up, ART experience, proportion of women, median age, sample size within each treatment arm, treatment within each arm, count of participants attaining adherence in each arm, the measures of adherence used, the number retained throughout the study. When data were unavailable or only partial, we requested data directly from authors. Data extraction from eligible studies was done independently and in duplicate.

We grouped treatment arms in the following categories: standard of care (SOC), enhanced standard of care (eSOC), alarm, eSOC plus alarm; eSOC plus calendar, daily short message service (SMS), weekly SMS, eSOC plus weekly SMS, eSOC plus treatment supporter, and SOC plus treatment supporter (table 2). SOC consisted of regular ART pick-ups including consultations with physician or pharmacist. In some cases adherence counselling was reported as part of SOC, and in others as a specific intervention, particularly when counsellors were involved. We did not differentiate such cases and deemed interventions that included adherence counselling in addition to SOC, either directly from the health practitioner or from adherence counsellors, to be eSOC. Finally, we did not differentiate treatment supporters that assisted in directly observed treatment (DOT) and those who provided other assistance.

Table 2Table image

Definitions used for categorisation of interventions in the network meta-analysis

The primary outcome was adherence as defined by the proportion of patients in each RCT arm meeting the trial-defined adherence criteria. Adherence was measured as the percentage of pills taken with various cutoff values, and when multiple measures were reported they were favoured in the following order: 95%, 90%, 80%, and 100%. We chose to place the 100% cutoff last in our order because it overestimates poor adherence.16 The proportion of patients achieving viral suppression was a secondary outcome. All outcomes were extracted at the end of study period.

Data synthesis and analysis

To inform comparative effectiveness between all interventions, we did a Bayesian network meta-analysis with all ten intervention types.17 This method provides better comparative evidence than pairwise meta-analysis because it combines direct (ie, head-to-head comparisons) and indirect evidence (comparisons across a common comparator) and in doing so increases the power of statistical comparisons while allowing for inferences of comparative effects between interventions that have not been compared head-to-head.13, 18 For the estimation of efficacy measures with Markov chain Monte Carlo methods, we used a burn-in of 20 000 iterations and 40 000 iterations for estimation. We assessed convergence with Gelman-Rubin diagnostics. Priors were normally distributed, centred at zero, with large variance for all variables except the probability of adherence and viral suppression, which both used a binomial prior distribution.

We performed edge-splitting to assess the consistency of direct and indirect evidence for interventions for which both types of information were available.19 We assessed the deviance information criterion (DIC) as a measure of model fit that penalises for model complexity.20 We modelled comparative log odds ratios (ORs) with the conventional logistic regression network meta-analysis set-up.17 All results for the network meta-analysis are reported as posterior medians with corresponding 95% credibility intervals (CrIs), the Bayesian analogue of CIs. Sensitivity analyses included period of trial follow-up and choices of adherence thresholds for measurement.

All analyses were done with WinBUGS version 1.4 (Medical Research Council Biostatistics Unit, Cambridge) and R version 3.0.1.

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

We identified 151 relevant abstracts (figure 1). Of these, 118 publications did not meet our inclusion criteria. Of the 33 further reviewed papers, we excluded 20 publications (appendix p 2): 12 were not RCTs,21—32 one did not report adherence interventions,33 one did not report adherence after 3 months,34 two had irrelevant interventions,35, 36 one did not report outcome,37 one had a cluster study design,38 one was in a paediatric population,39 and one was a substudy of another included trial.40 We included the remaining 13 publications, along with an additional poster41 provided after the search. These resources described 14 RCTs used in our analyses (table 3).38,39,41—54 Individual level data were available for nine of the RCTs.

Figure 1 Full-size image (11K) Download to PowerPoint

Study selection

RCT=randomised clinical trial.

Table 3Table image

Characteristics and outcomes of included trials reporting on adherence interventions for HIV-positive patients on ART

Our primary network includes data from 13 studies (5310 patients), comprising 30 treatment arms (figure 2). Follow-up time for adherence outcomes varied from 17 weeks to 192 weeks. Various measures were used to report adherence. The most common measure reported was the proportion of patients in each arm with at least 95% adherence by self-report; ten studies42—47,49,50,53,54 reported this measure. Four studies42, 47, 48, 50 reported the proportion of patients with no missed dose or 100% adherence, and two51, 52 reported the proportion with at least 90% adherence.

Figure 2 Full-size image (48K) Download to PowerPoint

Network for RCTs of interventions to improve ART adherence in HIV-positive patients

Nodes represent the individual or combined interventions. Lines between the nodes represent where direct (head-to-head) RCTs have been done. The numbers within those lines indicate the number of RCTs that have been conducted. RCT=randomised clinical trial. ART=antiretroviral therapy. SOC=standard of care.

To assess consistency across the network, we calculated direct and indirect evidence for each comparison for which both types of evidence were available (appendix p 3). Results were consistent between direct and indirect evidence, suggesting that conditions required for these analyses were met.

eSOC did better than basic SOC (table 4). Weekly SMS (with or without eSOC) was associated with better adherence than was SOC alone. The combination of eSOC with a treatment supporter was better than SOC, alarm alone, or daily SMS. Weekly SMS (without eSOC) was associated with higher adherence than was daily SMS; the difference between weekly SMS with eSOC from daily SMS was not statistically or operationally important. No other pairs of adherence interventions differed significantly. The product of the relative effect estimates for eSOC and weekly SMS was 2·41, which is close to the OR estimated for the combination of eSOC and weekly SMS. This implies a relation that is multiplicative with respect to relative effects as measured by ORs (additive within the logit model).

Table 4Table image

Odds of adherence to antiretroviral therapy in HIV-positive patients with different interventions

Neither follow-up time nor choice of adherence measurement affected the comparative efficacy measurements. As a sensitivity analysis for the adherence outcome, an additional network meta-analysis was done with the number remaining in the study (perprotocol) rather than intention to treat (appendix p 4). Comparisons of eSOC plus alarm with SOC, eSOC, or alarm alone were all statistically significant in the per-protocol analysis, suggesting differential loss to follow-up in these treatment arms.

Our secondary network meta-analysis included data from 13 treatment arms in six studies41, 42, 45, 49, 53, 54 (2738 patients; figure 3). Six interventions were included in the studies with available viral suppression data: SOC, eSOC, alarm, weekly SMS, eSOC plus treatment supporter, and SOC plus treatment supporter. For studies in which multiple timepoints were reported, the same timepoints were selected as in the adherence analysis where possible. Four studies reported the number of patients who had achieved plasma HIV RNA suppression (<400 copies per mL),41, 45, 53, 54 one study reported the number of patients on-study with viral failure defined as 400 copies per mL or more,41, 42 and one study reported the number of patients on-study with viral failure defined as 5000 copies per mL or more.43 We modelled viral suppression with an on-study analysis that treated measured lack of failure as equal to suppression irrespective of the cutoff.

Figure 3 Full-size image (30K) Download to PowerPoint

Network for RCTs assessing viral suppression with interventions to improve ART adherence in HIV-positive patients

Nodes represent the individual or combined interventions. Lines between the nodes represent where direct (head-to-head) RCTs have been done. The numbers within those lines indicate the number of RCTs. RCT=randomised clinical trial. SOC=standard of care.

As with adherence, we did edge-splitting to assess consistency between direct and indirect evidence across the network; results were reasonably consistent (appendix p 6), although there was a greater (but still non-significant) OR for eSOC than for SOC with direct evidence than with indirect evidence alone.

Both weekly SMS and eSOC plus treatment supporter were associated with higher suppression rates than were SOC, or SOC plus treatment supporter (table 5). No other pairs of adherence interventions differed with respect to viral suppression.

Table 5Table image

Viral suppression (<400 copies per mL) at last reported timepoint

Discussion

We found compelling evidence that enhanced standard of care improved adherence of patients to ART in Africa. Adherence was further improved when combined with weekly SMS messages and treatment supporters. The combination of enhanced standard of care, a cognitive intervention, and weekly SMS messaging, a behavioural intervention, seemed to be additive in nature, a new finding that could not be tested in the individual studies in the current evidence base. Our findings also suggest that evidence is insufficient to support alarms, daily SMS messages, and calendars. These findings are at odds with those in some previous reports and meta-analyses, and the difference could be partly explained by our analytical approach.10, 55 Our study found a large benefit for weekly but not for daily SMS messages. A dose-effect might exist whereby less is more because supportive SMS messages may become a reminder when too frequent, and reminders do not seem to support adherence.56

Our findings have operational and clinical implications. For example, we found a large, additive treatment benefit of the addition of weekly SMS messages to eSOC. Combinations of cognitive and behavioural interventions might therefore maximise the intervention efficacy. Although weekly SMS messaging is a low-cost intervention, it requires that patients have access to cell phones and can receive SMS messages confidentially.57 Given the high penetration of mobile technology in low-income settings such as sub-Saharan Africa and India. our findings may have global relevance and implications. Nonetheless, features of the weekly SMS messaging intervention need further research, such as whether patients will be able to respond to the messages and reach a care provider (so called, two way messages) or not (one way), and what content should be sent.58 The trials included in this study differed in this regard.

Similarly, we found a large effect of a treatment supporter in combination with eSOC. However, this intervention would be inappropriate where confiding ones HIV status to another person is not possible.49 Our finding that treatment supporters importantly increase adherence is at odds with some reviews examining treatment supporters and directly observed therapy.55, 59 Other reviews have included populations with competing mental health concerns and have used standard meta-analysis approaches. The use of a network meta-analysis allows for greater power and precision, which seems to explain why our findings are significant and others findings are not.60 Previous work has documented the feasibility, acceptability, and potential efficacy of treatment supporters as a community-based intervention (ie, widespread use of this method throughout the community).49, 61, 62

Across HIV programmes, treatment supporters can be defined in several ways and this has created a debate within implementation science as to what extent they should be promoted. Treatment supporters range from paid employees, such as accompagnateurs (community health workers) in Partners in Health projects, to unpaid family and friends in other programmes.55 Treatment supporters offer assistance that ranges from emotional support and reminding patients to adhere to therapy to more intensive services that include DOT and clinical monitoring. The evidence to support DOT is not convincing,55 but the evidence for social support that includes adherence discussions and reminders is much more broadly accepted. Our analysis is unlikely to settle the issue.

Our analysis has several strengths and limitations. Strengths include our extensive search, communication with trialists, and the statistical approach. We held meetings of those working in the field to identify any additional trials and received individual patient-level data where possible. Our statistical approach allows for greater power than standard meta-analysis as it incorporates data from both direct and indirect evidence (appendix p 5). Limitations of our review to generalisability include the lack of available data in specific populations such as HIV-infected children, adolescents, pregnant women, prisoners, and MSM. We found few studies for each individual intervention and so further confirmatory RCTs are warranted. We considered including studies from more developed settings, but given that the HIV epidemic in Africa is substantially different from that in other continents (in terms of a generalised epidemic) and that most RCTs in other settings have been directed at individuals with competing mental health concerns (eg, addictions) or marginalised people (eg, homeless, youth), we believe that restricting the analysis to Africa is necessary to meet the conditions required for the methods used for our analyses.

An important limitation to our study pertains to treatment definitions. As opposed to drugs, these behavioural and cognitive interventions varied across studies. Particularly eSOC, defined as SOC with an educational component, because the education component varied according to content and whether it was delivered in-group or one-on-one. Nonetheless, statistical heterogeneity was moderate, suggesting that differences in definitions were a minimum threat to the robustness of our analysis. Furthermore, we viewed various definitions of adherence and viral failure as equivalent. We considered self-reported adherence and more objective forms (such as medication event monitoring systems) as equivalent. However, self-report may overestimate adherence.63 An insufficient number of studies were available to assess this in sensitivity analysis. We included only RCTs, and other interventions used at the programme level outside of research might have important treatment benefits. Interventions to promote retention in programmes differ across and within countries and some programmes may use different adherence strategies also.64 Finally, we considered the RCT period as equivalent across studies and did a sensitivity analysis to investigate duration of follow-up. Although we did not identify time as an effect modifier, adherence will probably wane over the long term with any intervention.65, 66

Network meta-analysis should only be considered as valid as the individual comparisons within a network. Several of the nodes in our network are informed by just one or two trials and at most by five trials. In general, the more trials in a comparison, the greater the power to detect treatment effects.18, 67 Although we cannot add trials to our network, because no others exist, we can assess whether the comparisons are believable by assessing the transitivity of direct versus indirect evidence.68 When we assessed pairwise estimates versus network estimates we found no evidence of inconsistency between the direct and indirect evidence. This result increases our confidence that the network is sufficiently robust that the findings are unlikely to be spurious.68 As further evidence accumulates, this will further strengthen inferences from the network analysis.

Our study provides strong inferences that a standard of care that includes counselling of patients on adherence, SMS messaging, and treatment supporters can improve adherence for patients in Africa. As the provision of ART in Africa becomes more long term, sustainable efforts to promote adherence will be required. Future research should assess new adherence interventions individually or in combination, not only in adult populations but also in selected vulnerable populations such as children, adolescents, and pregnant women, for whom knowledge is scarce. Studies are also needed to assess the cost-effectiveness to inform policy makers, clinicians, and programme managers.

Contributors

EJM, RL, KT, ML, KM, SK, SL, RG, YC, KRA, HT, CP, RHR, LM, LT, MHC, IBW, AL, OAU, JS, DB, SY, TB, NF, and JBN conceived and designed the study; EJM, RL, ML, KM, SK, and JBN acquired the data; KT, ML, and SK did the statistical analyses; EJM, ML, KM, and SK drafted the manuscript; EJM, RL, ML, KM, SK, and JBN provided critical revisions of the manuscript for important intellectual content.

Declaration of interests

We declare no competing interests.

Acknowledgments

The authors thank Michael Stirrat (National Institutes of Mental Health, Bethesda, MD, USA) for critical insights and logistical support.

Supplementary Material

Supplementary appendix

PDF (224K)

References

1 Braitstein P, Brinkhof MW, Dabis F, et al. Mortality of HIV-1-infected patients in the first year of antiretroviral therapy: comparison between low-income and high-income countries. Lancet 2006; 367: 817-824. Summary | Full Text | PDF(122KB) | CrossRef | PubMed

2 Mills EJ, Lester R, Ford N. Promoting long term adherence to antiretroviral treatment. BMJ 2012; 344: e4173. PubMed

3 Langebeek N, Gisolf EH, Reiss P, et al. Predictors and correlates of adherence to combination antiretroviral therapy (ART) for chronic HIV infection: a meta-analysis. BMC Med 2014; 12: 142. CrossRef | PubMed

4 Mills EJ, Nachega JB, Bangsberg DR, et al. Adherence to HAART: a systematic review of developed and developing nation patient-reported barriers and facilitators. PLoS Med 2006; 3: e438. CrossRef | PubMed

5 Fauci AS, Marston HD. Achieving an AIDS-free world: science and implementation. Lancet 2013; 382: 1461-1462. Full Text | PDF(209KB) | CrossRef | PubMed

6 Thompson MA, Mugavero MJ, Amico KR, et al. Guidelines for improving entry into and retention in care and antiretroviral adherence for persons with HIV: evidence-based recommendations from an International Association of Physicians in AIDS Care panel. Ann Intern Med 2012; 156: 817-833. CrossRef | PubMed

7 Barnighausen T, Chaiyachati K, Chimbindi N, et al. Interventions to increase antiretroviral adherence in sub-Saharan Africa: a systematic review of evaluation studies. Lancet Infect Dis 2011; 11: 942-951. Summary | Full Text | PDF(144KB) | CrossRef | PubMed

8 Weiser SD, Palar K, Frongillo EA, et al. Longitudinal assessment of associations between food insecurity, antiretroviral adherence and HIV treatment outcomes in rural Uganda. AIDS 2014; 28: 115-120. PubMed

9 Chaiyachati KH, Ogbuoji O, Price M, et al. Interventions to improve adherence to antiretroviral therapy: a rapid systematic review. AIDS 2014; 28 (suppl 2): S187-S204. CrossRef | PubMed

10 Barnighausen T, Chaiyachati K, Chimbindi N, et al. Interventions to increase antiretroviral adherence in sub-Saharan Africa: a systematic review of evaluation studies. Lancet Infect Dis 2011; 11: 942-951. Summary | Full Text | PDF(144KB) | CrossRef | PubMed

11 Jansen JP, Fleurence R, Devine B, et al. Interpreting indirect treatment comparisons and network meta-analysis for health-care decision making: report of the ISPOR Task Force on Indirect Treatment Comparisons Good Research Practices: part 1. Value Health 2011; 14: 417-428. CrossRef | PubMed

12 Lu G, Ades AE. Combination of direct and indirect evidence in mixed treatment comparisons. Stat Med 2004; 23: 3105-3124. CrossRef | PubMed

13 Mills EJ, Ioannidis JP, Thorlund K, et al. How to use an article reporting a multiple treatment comparison meta-analysis. JAMA 2012; 308: 1246-1253. CrossRef | PubMed

14 Salanti G, Higgins JP, Ades AE, et al. Evaluation of networks of randomized trials. Stat Methods Med Res 2008; 17: 279-301. CrossRef | PubMed

15 Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009; 6: e1000097. CrossRef | PubMed

16 Bangsberg DR. Less than 95% adherence to nonnucleoside reverse-transcriptase inhibitor therapy can lead to viral suppression. Clin Infect Dis 2006; 43: 939-941. CrossRef | PubMed

17 Lu G, Ades AE. Combination of direct and indirect evidence in mixed treatment comparisons. Stat Med 2004; 23: 3105-3124. CrossRef | PubMed

18 Mills EJ, Thorlund K, Ioannidis JP. Demystifying trial networks and network meta-analysis. BMJ 2013; 346: f2914. PubMed

19 Dias S, Sutton AJ, Welton NJ, et al. Evidence synthesis for decision making 3: heterogeneity—subgroups, meta-regression, bias, and bias-adjustment. Med Decis Making 2013; 33: 618-640. CrossRef | PubMed

20 Spiegelhalter DJ, Best NG, Carlin BP, et al. Bayesian measures of model complexity and fit. J R Stat Soc B 2002; 64: 583-639. PubMed

21 Idoko JA, Agbaji O, Agaba P, et al. Direct observation therapy-highly active antiretroviral therapy in a resource-limited setting: the use of community treatment support can be effective. Int J STD AIDS 2007; 18: 760-763. CrossRef | PubMed

22 Kabore I, Bloem J, Etheredge G, et al. The effect of community-based support services on clinical efficacy and health-related quality of life in HIV/AIDS patients in resource-limited settings in sub-Saharan Africa. AIDS Patient Care STDS 2010; 24: 581-594. CrossRef | PubMed

23 Sherr KH, Micek MA, Gimbel SO, et al. Quality of HIV care provided by non-physician clinicians and physicians in Mozambique: a retrospective cohort study. AIDS 2010; 24 (suppl 1): S59-S66. CrossRef | PubMed

24 Stubbs BA, Micek MA, Pfeiffer JT, et al. Treatment partners and adherence to HAART in Central Mozambique. AIDS Care 2009; 21: 1412-1419. CrossRef | PubMed

25 Pienaar D, Myer L, Cleary S, et al. Models of care for antiretroviral service delivery. Cape Town: University of Capetown, 2006.

26 Thurman TR, Haas LJ, Dushimimana A, et al. Evaluation of a case management program for HIV clients in Rwanda. AIDS Care 2010; 22: 759-765. CrossRef | PubMed

27 Torpey KE, Kabaso ME, Mutale LN, et al. Adherence support workers: a way to address human resource constraints in antiretroviral treatment programs in the public health setting in Zambia. PLoS One 2008; 3: e2204. CrossRef | PubMed

28 Roux S. Diary cards: preliminary evaluation of an intervention tool for improving adherence to antiretroviral therapy and TB preventative therapy in people living with HIV/AIDS. Cape Town: University of the Western Cape, 2004.

29 Pirkle CM, Boileau C, Nguyen VK, et al. Impact of a modified directly administered antiretroviral treatment intervention on virological outcome in HIV-infected patients treated in Burkina Faso and Mali. HIV Med 2009; 10: 152-156. CrossRef | PubMed

30 Byron E, Gillespie S, Nangami M. Integrating nutrition security with treatment of people living with HIV: lessons from Kenya. Food Nutr Bull 2008; 29: 87-97. PubMed

31 Igumbor JO, Scheepers E, Ebrahim R, Jason A, Grimwood A. An evaluation of the impact of a community-based adherence support programme on ART outcomes in selected government HIV treatment sites in South Africa. AIDS Care 2011; 23: 231-236. CrossRef | PubMed

32 Rich ML, Miller AC, Niyigena P, et al. Excellent clinical outcomes and high retention in care among adults in a community-based HIV treatment program in rural Rwanda. J Acquir Immune Defic Syndr 2012; 59: e35-e42. CrossRef | PubMed

33 Cantrell RA, Sinkala M, Megazinni K, et al. A pilot study of food supplementation to improve adherence to antiretroviral therapy among food-insecure adults in Lusaka, Zambia. J Acquir Immune Defic Syndr 2008; 49: 190-195. CrossRef | PubMed

34 Mansoor LE, Dowse R. Medicines information and adherence in HIV/AIDS patients. J Clin Pharm Ther 2006; 31: 7-15. CrossRef | PubMed

35 Holstad MM, Essien JE, Ekong E, et al. Motivational groups support adherence to antiretroviral therapy and use of risk reduction behaviors in HIV positive Nigerian women: a pilot study. Afr J Reprod Health 2012; 16: 14-27. PubMed

36 Kiweewa FM, Wabwire D, Nakibuuka J, et al. Noninferiority of a task-shifting HIV care and treatment model using peer counselors and nurses among Ugandan women initiated on ART: evidence from a randomized trial. J Acquir Immune Defic Syndr 2013; 63: e125-e132. CrossRef | PubMed

37 Van Loggerenberg F, Grant A, Naidoo K, et al. Group counseling achieves high adherence to antiretroviral therapy: results of the CAPRISA 058 randomized controlled trial comparing group versus individualized adherence counseling strategies in Durban, South Africa. J Int Assoc Physicians AIDS Care 2010; 9: 260. PubMed

38 Jaffar S, Amuron B, Foster S, et al. Rates of virological failure in patients treated in a home-based versus a facility-based HIV-care model in Jinja, southeast Uganda: a cluster-randomised equivalence trial. Lancet 2009; 374: 2080-2089. Summary | Full Text | PDF(141KB) | CrossRef | PubMed

39 Wamalwa DC, Farquhar C, Obimbo EM, et al. Medication diaries do not improve outcomes with highly active antiretroviral therapy in Kenyan children: a randomized clinical trial. J Int AIDS Soc 2009; 12: 8. PubMed

40 Munyao P, Luchters S, Chersich MF, et al. Implementation of clinic-based modified-directly observed therapy (m-DOT) for ART; experiences in Mombasa, Kenya. AIDS Care 2010; 22: 187-194. CrossRef | PubMed

41 Gross R, Zheng L, La Rosa A et al. Partner-based intervention for adherence to second-line ART: a multinational trial (ACTG A5234). Conference on Retroviruses and Opportunistic Infections; Boston, MA, USA; March 3—6, 2014. Abstract 538.

42 Chang LW, Kagaayi J, Nakigozi G, et al. Effect of peer health workers on AIDS care in Rakai, Uganda: a cluster-randomized trial. PLoS One 2010; 5: e10923. CrossRef | PubMed

43 Chung MH, Richardson BA, Tapia K, et al. A randomized controlled trial comparing the effects of counseling and alarm device on HAART adherence and virologic outcomes. PLoS Med 2011; 8: e1000422. CrossRef | PubMed

44 Kunutsor S, Walley J, Katabira E, et al. Improving clinic attendance and adherence to antiretroviral therapy through a treatment supporter intervention in Uganda: a randomized controlled trial. AIDS Behav 2011; 15: 1795-1802. CrossRef | PubMed

45 Lester RT, Ritvo P, Mills EJ, et al. Effects of a mobile phone short message service on antiretroviral treatment adherence in Kenya (WelTel Kenya1): a randomised trial. Lancet 2010; 376: 1838-1845. Summary | Full Text | PDF(250KB) | CrossRef | PubMed

46 Maduka O, Tobin-West CI. Adherence counseling and reminder text messages improve uptake of antiretroviral therapy in a tertiary hospital in Nigeria. Niger J Clin Pract 2013; 16: 302-308. CrossRef | PubMed

47 Mbuagbaw L, Thabane L, Ongolo-Zogo P, et al. The Cameroon Mobile Phone SMS (CAMPS) trial: a randomized trial of text messaging versus usual care for adherence to antiretroviral therapy. PLoS One 2012; 7: e46909. CrossRef | PubMed

48 Mugusi F, Mugusi S, Bakari M, et al. Enhancing adherence to antiretroviral therapy at the HIV clinic in resource constrained countries; the Tanzanian experience. Trop Med Int Health 2009; 14: 1226-1232. CrossRef | PubMed

49 Nachega JB, Chaisson RE, Goliath R, et al. Randomized controlled trial of trained patient-nominated treatment supporters providing partial directly observed antiretroviral therapy. AIDS 2010; 24: 1273-1280. PubMed

50 Peltzer K, Ramlagan S, Jones D, et al. Efficacy of a lay health worker led group antiretroviral medication adherence training among non-adherent HIV-positive patients in KwaZulu-Natal, South Africa: results from a randomized trial. SAHARA J 2012; 9: 218-226. CrossRef | PubMed

51 Pearson CR, Micek MA, Simoni JM, et al. Randomized control trial of peer-delivered, modified directly observed therapy for HAART in Mozambique. J Acquir Immune Defic Syndr 2007; 46: 238-244. CrossRef | PubMed

52 Pop-Eleches C, Thirumurthy H, Habyarimana JP, et al. Mobile phone technologies improve adherence to antiretroviral treatment in a resource-limited setting: a randomized controlled trial of text message reminders. AIDS 2011; 25: 825-834. CrossRef | PubMed

53 Sarna A, Luchters S, Geibel S, et al. Short- and long-term efficacy of modified directly observed antiretroviral treatment in Mombasa, Kenya: a randomized trial. J Acquir Immune Defic Syndr 2008; 48: 611-619. CrossRef | PubMed

54 Taiwo BO, Idoko JA, Welty LJ, et al. Assessing the viorologic and adherence benefits of patient-selected HIV treatment partners in a resource-limited setting. J Acquir Immune Defic Syndr 2010; 54: 85-92. CrossRef | PubMed

55 Ford N, Nachega JB, Engel ME, et al. Directly observed antiretroviral therapy: a systematic review and meta-analysis of randomised clinical trials. Lancet 2009; 374: 2064-2071. Summary | Full Text | PDF(117KB) | CrossRef | PubMed

56 Chung MH, Richardson BA, Tapia K, et al. A randomized controlled trial comparing the effects of counseling and alarm device on HAART adherence and virologic outcomes. PLoS Med 2011; 8: e1000422. CrossRef | PubMed

57 Lester R, Karanja S. Mobile phones: exceptional tools for HIV/AIDS, health, and crisis management. Lancet Infect Dis 2008; 8: 738-739. Full Text | PDF(418KB) | CrossRef | PubMed

58 Thirumurthy H, Lester RT. M-health for health behaviour change in resource-limited settings: applications to HIV care and beyond. Bull World Health Organ 2012; 90: 390-392. CrossRef | PubMed

59 Hart JE, Jeon CY, Ivers LC, et al. Effect of directly observed therapy for highly active antiretroviral therapy on virologic, immunologic, and adherence outcomes: a meta-analysis and systematic review. J Acquir Immune Defic Syndr 2010; 54: 167-179. PubMed

60 Higgins JP, Whitehead A. Borrowing strength from external trials in a meta-analysis. Stat Med 1996; 15: 2733-2749. CrossRef | PubMed

61 Duwell MM, Knowlton AR, Nachega JB, et al. Patient-nominated, community-based HIV treatment supporters: patient perspectives, feasibility, challenges, and factors for success in HIV-infected South African adults. AIDS Patient Care STDS 2013; 27: 96-102. CrossRef | PubMed

62 Nachega JB, Knowlton AR, Deluca A, et al. Treatment supporter to improve adherence to antiretroviral therapy in HIV-infected South African adults: a qualitative study. J Acquir Immune Defic Syndr 2006; 43 (suppl 1): S127-S133. CrossRef | PubMed

63 Thirumurthy H, Siripong N, Vreeman RC, et al. Differences between self-reported and electronically monitored adherence among patients receiving antiretroviral therapy in a resource-limited setting. AIDS 2012; 26: 2399-2403. CrossRef | PubMed

64 Chi BH, Yiannoutsos CT, Westfall AO, et al. Universal definition of loss to follow-up in HIV treatment programs: a statistical analysis of 111 facilities in Africa, Asia, and Latin America. PLoS Med 2011; 8: e1001111. PubMed

65 Bangsberg DR, Mills EJ. Long-term adherence to antiretroviral therapy in resource-limited settings: a bitter pill to swallow. Antivir Ther 2013; 18: 25-28. CrossRef | PubMed

66 Wilson IB, Bangsberg DR, Shen JM, et al. Heterogeneity among studies in rates of decline of antiretroviral therapy adherence over time: results from the multisite adherence collaboration on HIV 14 study. J Acquir Immune Defic Syndr 2013; 64: 448-454. CrossRef | PubMed

67 Thorlund K, Mills EJ. Sample size and power considerations in network meta-analysis. Syst Rev 2012; 1: 41. PubMed

68 Salanti G, Del Giovane C, Chaimani A, et al. Evaluating the quality of evidence from a network meta-analysis. PLoS One 2014; 9: e99682. CrossRef | PubMed

a Stanford Prevention Research Center, Stanford University, Stanford, CA, USA

b School of Population and Public Health, University of British Columbia, Vancouver, BC, Canada

c Department of Clinical Epidemiology & Biostatistics, Faculty of Health Sciences, McMaster University, Hamilton, ON, Canada

d RAND Corp, Los Angeles, CA, USA

e University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA

f Department of Emergency Medicine, Albert Einstein University, New York, NY, USA

g Department of Health Behavior and Health Education, University of Michigan, Ann Arbor, MI, USA

h Gillings School of Global Public Health, University of North Carolina at Chapel Hill, NC, USA

i School of Social Work, University of Washington, Seattle, WA, USA

j HIV Center for Clinical and Behavioral Studies, NY State Psychiatric Institute and Columbia University, New York, NY, USA

k Department of Global Health, School of Medicine and Public Health, University of Washington, Seattle, WA, USA

l Department of Psychology, University of Washington, Seattle, WA, USA

m Department of Health Services, Policy & Practice, Brown University, Providence, RI, USA

n Center for AIDS Research, UCSF, San Francisco, CA, USA

o Centre for Applied Health Research & Delivery, Warwick University, Coventry, UK

p Harvard School of Public Health, Harvard, Boston, MA, USA

q Wellcome Trust Africa Centre for Health and Population Studies, University of KwaZulu-Natal, Mtubatuba, South Africa

r Faculty of Health Sciences, University of Ottawa, Ottawa, ON, Canada

s Department of HIV/AIDS, WHO, Geneva, Switzerland

t Center for Infectious Diseases, Stellenbosch University, Cape Town, Western Cape, South Africa

Correspondence to: Edward J Mills, Stanford Prevention Research Center, Stanford University School of Medicine, Medical School Office Building, 1265 Welch Road, Mail code 5411, Stanford, CA, USA 94305-5411