• Record: found
  • Abstract: found
  • Article: found
Is Open Access

Examining the effects of HIV self-testing compared to standard HIV testing services: a systematic review and meta-analysis

Read this article at

      There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.


      Introduction: HIV self-testing (HIVST) is a discreet and convenient way to reach people with HIV who do not know their status, including many who may not otherwise test. To inform World Health Organization (WHO) guidance, we assessed the effect of HIVST on uptake and frequency of testing, as well as identification of HIV-positive persons, linkage to care, social harm, and risk behaviour.Methods: We systematically searched for studies comparing HIVST to standard HIV testing until 1 June 2016. Meta-analyses of studies reporting comparable outcomes were conducted using a random-effects model for relative risks (RR) and 95% confidence intervals. The quality of evidence was evaluated using GRADE.Results: After screening 638 citations, we identified five randomized controlled trials (RCTs) comparing HIVST to standard HIV testing services among 4,145 total participants from four countries. All offered free oral-fluid rapid tests for HIVST and were among men. Meta-analysis of three RCTs showed HIVST doubled uptake of testing among men (RR = 2.12; 95% CI: 1.51, 2.98). Meta-analysis of two RCTs among men who have sex with men showed frequency of testing nearly doubled (Rate ratio = 1.88; 95% CI: 1.17; 3.01), resulting in two more tests in a 12–15-month period (Mean difference = 2.13; 95% CI: 1.59, 2.66). Meta-analysis of two RCTs showed HIVST also doubled the likelihood of an HIV-positive diagnosis (RR = 2.02; 95% CI: 0.37, 10.76, 5.32). Across all RCTs, there was no indication of harm attributable to HIVST and potential increases in risk-taking behaviour appeared to be minimal.Conclusions: HIVST is associated with increased uptake and frequency of testing in RCTs. Such increases, particularly among those at risk who may not otherwise test, will likely identify more HIV-positive individuals as compared to standard testing services alone. However, further research on how to support linkage to confirmatory testing, prevention, treatment and care services is needed. WHO now recommends HIVST as an additional HIV testing approach.

      Related collections

      Most cited references 27

      • Record: found
      • Abstract: found
      • Article: not found

      GRADE guidelines: 5. Rating the quality of evidence--publication bias.

      In the GRADE approach, randomized trials start as high-quality evidence and observational studies as low-quality evidence, but both can be rated down if a body of evidence is associated with a high risk of publication bias. Even when individual studies included in best-evidence summaries have a low risk of bias, publication bias can result in substantial overestimates of effect. Authors should suspect publication bias when available evidence comes from a number of small studies, most of which have been commercially funded. A number of approaches based on examination of the pattern of data are available to help assess publication bias. The most popular of these is the funnel plot; all, however, have substantial limitations. Publication bias is likely frequent, and caution in the face of early results, particularly with small sample size and number of events, is warranted. Copyright © 2011 Elsevier Inc. All rights reserved.
        • Record: found
        • Abstract: found
        • Article: found
        Is Open Access

        The Uptake and Accuracy of Oral Kits for HIV Self-Testing in High HIV Prevalence Setting: A Cross-Sectional Feasibility Study in Blantyre, Malawi

        Introduction The acceptability of HIV testing has increased dramatically in the last few years in countries with generalized HIV epidemics, as evidenced by increasingly low refusal rates when adults are directly offered convenient HIV testing either in health facilities [1],[2] or through home-based HIV testing [3]–. In Southern Africa, where adult HIV prevalence is far higher than in any other region [8], only a small minority of adults who do not already know their HIV status have no intent to test in future [8],[9]. However, use of facility-based and free-standing voluntary HIV testing and counseling services has remained well below levels of expressed interest and intent, particularly for men, rural populations, and the very poor [6],[8]–[10]. The inconvenience and cost involved in visiting fixed-site services, as well as a general aversion to visiting health facilities, appear to be major deterrents [4]–[6],[8]–[12]. Home-based HIV-testing services bypass these barriers, have much greater acceptability at population level [4],[5],[9],[12], and are being adopted as national policy in a number of countries [7],[8],[13]. Unfortunately, a common finding is that many people do not want to be counseled and tested by someone they know personally, or even access services being used by people they know [5],[7],[12]. This situation creates a strong tension between confidentiality and convenience, and adds considerably to the cost and logistical difficulty of providing home-based services [4],[5],[7],[12],[14]. Self-testing in private has considerable unexplored potential to contribute to first-time and repeat testing for HIV, and could potentially meet the requirement for anonymity at the time of home-testing [7],[11],[15],[16]. Self-testing, however, raises a number of issues, relating to its accuracy, the potential for adverse psychological reactions in the absence of face-to-face counseling [16], and the difficulty in organizing subsequent linkage into HIV/AIDS care [15]–[18]. Reassuring results have been obtained from the US, where home collection of specimens followed by telephoned results and counseling has been available for over a decade [16]: postmarketing surveillance has not identified any increased risk of suicide, and subsequent linkage into care appears similar to that obtained with other testing options [19]. Given the need to further scale up HIV testing and counseling in Africa [8] and encourage regular repeat testing [8], we carried out a mixed quantitative and qualitative study of self-testing for HIV using oral test kits in urban Blantyre, Malawi, with confirmatory blood-based HIV testing and post-test counseling. To the best of our knowledge, this constitutes the first study of the use of oral self-testing among the general population of a country affected by a generalized epidemic. The main aim was to test whether supervised oral self-testing could yield accurate results. We also explored reasons for accepting self-testing and respondents' preferences for HIV testing. Methods Community and Participant Selection Population-weighted random cluster sampling, using community-health worker catchment areas as the primary sampling unit, was used to select four community-health worker catchment areas from Ndirande, Likhubula, and Chilomoni high-density residential suburbs of Blantyre, Malawi. Boundaries were defined by capturing Global Positioning System (GPS) coordinates on a circumferential walk. These coordinates were superimposed onto high resolution satellite maps (March 2010 images: Geo Eye-1/Eurimage SpA) using Google Earth Pro software. Between March and July 2010, two groups of participants were randomly selected from within these catchment areas for interview and offer of HIV counseling and testing. Adults in Randomly Selected Households Identity numbers were assigned to all visible dwellings, followed by random selection of 15 dwellings from each selected catchment area. A random list of five further dwellings was available to replace any abandoned or nonresidential buildings. Dwellings were visited to introduce the study, identify all individual households (defined by sharing food), and allow a final random selection of a single household from each dwelling. All adults (≥16 y) were then invited to participate in interview and optional HIV testing and counseling (Figure 1) carried out in the participant's home. 10.1371/journal.pmed.1001102.g001 Figure 1 Schematic of study design. Purposive sampling was then used to select participants for in-depth interviews held within a few days of self-testing, aiming to include up to 36 participants to represent single men, single women, and couples, and to explore choice between the different HIV test options (see below). Interview guides aimed to situate individual experiences of self-testing within broader lifestyle narratives including decision-making, testing preferences and practices, health seeking experiences, and individual level barriers and facilitators. Community Peer-Group Members To facilitate informative focus group discussions, peer groups, such as sports teams, micro-finance, and church groups, that were active in the four selected catchment areas were identified through situational analysis and community mapping exercises. Eligibility was restricted to groups without a health or HIV theme. Six groups were randomly selected for inclusion, followed by random selection of 12 individuals from each group for a 2-h focus group discussion held at a local community venue. These discussions explored community views on barriers and facilitators to self-testing including use amongst couples, the role of supervision and counseling, preferences for different voluntary counseling and testing (VCT) options, and a demonstration and discussion of the self-testing process. All participants were then offered HIV testing and counseling on an individual basis (see Figure 1). Optional HIV Testing and Counseling Consenting adults selected as above were asked to consent to a questionnaire capturing demographic and socioeconomic data, previous HIV test history, and test preferences (with a stigma module also administered to adults from randomly selected households), before being offered the choice between: self-test for HIV followed by standard VCT; standard VCT only; and no HIV testing or counseling. Participants were asked to consider HIV testing even if they knew themselves to be HIV positive, to allow general population HIV prevalence to be estimated and to maximize power to investigate specificity. Participants opting to self-test had pretest counseling and a brief demonstration of an oral test kit. They then collected their own specimen without intervention, developed and read the results, guided by illustrated instructions. A brief anonymous self-administered questionnaire captured participant recruitment group, 10-y age band, results, ease, credibility, self-identified errors, and self-test results. This questionnaire was read to illiterate participants. Participants could request additional help if needed. After self-testing, a counselor reread the self-test kit, completed a checklist of potential errors, and confirmed the result using two rapid HIV test kits run in parallel from a finger-prick blood specimen. After post-test counseling and (where indicated) written referral into HIV care services, an exit interview was conducted by a research assistant unaware of the participant's HIV result. The exit questionnaire included a section filled by the counselor on self-test accuracy, and errors made, but not the test result. Ethical Considerations All participants provided written (or witnessed if illiterate) informed consent. Pre- and post-test counseling was provided to all participants. All participants testing positive were referred to the nearest primary health center for HIV care. Approval was obtained from the ethics committees of London School of Hygiene and Tropical Medicine (UK) and College of Medicine Research Ethics Committee, Blantyre, Malawi. Laboratory Methods Self-testing used OraQuick ADVANCE HIV I/II (OraSure Technologies) run from an oral mucosal transudate specimen. Confirmatory testing used Determine (Abbott Laboratories) and Unigold (Trinity Biotech plc), with a third test (SD Bioline HIV I/II Standard Diagnostics, Inc.), together with repeat of all three previously used kits, used in the event of any discordance. Test kits were stored in a temperature-controlled laboratory in compliance with recommended storage between 2 to 27°C, before use at ambient temperatures that were within the manufacturers recommendations (15–37°C) at the time of the study. Sample Size Sample sizes were based around sufficient precision to exclude accuracy of self-testing worse than 95% if our measured estimate was 98%. Our sampling strategy aimed to provide 200 to 250 participants accepting self-testing, on the basis of three adults per household and an 80% or higher uptake of self-testing in community participants. Statistical Methods Quantitative data were entered and managed using Epi info version 3.4.3 (Centers for Disease Control and Prevention, Atlanta, Georgia, US). Analysis used Stata version 10.0 (Stata Corporation). Questionnaires used mainly close-ended questions with predefined categorical coding, on the basis of previously published and author-supplied questionnaires as far as possible [20]–[26]. Demographic variables were recoded into categorical variables on the basis of frequency of responses before substantive data analysis. Where possible, variables with similar themes were combined. Occupation was classified into six categories, using the nature of the job and security of contract. Twelve questions related to stigma were combined into an ordered categorical scale, having first shown acceptable internal consistency using Cronbach's alpha [27]. 95% confidence intervals for the proportion of self-tests accurately read, and for sensitivity and specificity of the tests were calculated using the Agresti-Coull approach [28]. Participants with no confirmatory results were excluded; inconclusive results were treated as inaccurate for sensitivity and specificity analyses. The clustered sampling design was taken into account using the complex survey commands in Stata, which use linearization to estimate standard errors. Categorical baseline characteristics were compared between the two selection groups, and between males and females, using design-based F-tests calculated from applying the second order Rao and Scott correction [29],[30] to the usual Pearson chi-squared test statistic for two-way tables. Continuous baseline characteristics were compared between selection groups using logistic regression, adjusted for clustering. Risk factors were explored for the main outcomes of interest (uptake, accuracy, acceptability, and reported ease of self-testing) using the F-test described above for categorical variables and logistic regression, adjusted for clustering, for continuous or ordered categorical variables. Logistic regression, adjusted for clustering was used for multivariate analysis, including variables identified as having a priori association with outcomes of interest, as well as other variables with univariate association at p≤0.1. Thereafter a stepwise approach was taken, removing variables with p>0.1 from the multivariate model. Qualitative data were recorded, transcribed, translated, quality checked, imported to NVIVO software for qualitative data analysis version 8.0 (QSR), and coded thematically using a framework developed deductively from the study objectives and reordered inductively from a grounded approach to theme identification. Analysis used a constant comparative approach, identifying patterns in the data, situating these and testing developing hypotheses on all cases, reexamining data contradictions within broader social contexts. Results Participants 283 (95.6%) of 298 selected individuals consented to interview, including 216 of 226 (95.6%) adults from the 60 randomly selected households, and 67 of 72 (93.1%) community peer-group members (see flow diagram, Figure 1). Baseline characteristics by selection group are shown in Table 1. 147 (51.9%) of participants were women, with median age of 27 y. Only 30 (10.6%) participants were in regular employment, and regular or recent food shortage in the household was reported by 39 (58.2%) peer-group members and 55 (25.5%) adults from randomly selected households (design-based F(1,3) = 15.65, p = 0.029). Peer-group members were significantly more likely to report worrying a lot about HIV (27; 41.5%) than adults from randomly selected households (42; 19.4%, design-based F(1,3) = 85.69, p = 0.003). Otherwise baseline characteristics were similar between selection groups. 10.1371/journal.pmed.1001102.t001 Table 1 Participant characteristics by selection group. Characteristic Peer-Group Member (n = 67) Household Member(n = 216) p-Value Gender Male 30 (44.8%) 106 (50.9%) 0.532* Female 37 (55.2%) 110 (49.1%) — Age Median (IQR) 27 (23–32) 26.5 (22–32) 0.302** Occupation Regular employment 8 (11.9%) 22 (10.2%) 0.726* Marital status Single 20 (29.9%) 66 (30.6%) 0.977* Married 40 (59.7%) 128 (59.2%) — Divorced/widowed 7 (10.4%) 22 (10.2%) — Stigma score Low ND 94 (44.1%) ND Low-medium ND 52 (24.4%) — Medium-high ND 46 (21.6%) — High ND 21 (9.9%) — Literacy Unable to read 4 (6.0%) 19 (8.8%) 0.611* Education Part primary/None 5 (7.5%) 17 (7.9%) 0.176** Primary or lower 20 (29.9%) 72 (33.3%) — Higher 42 (62.7%) 127 (58.8%) — Goes hungry Sometimes 27 (40.9%) 52 (24.1%) 0.029** Often/always 12 (18.2%) 3 (1.4%) — Previous HIV test Ever tested before 38 (56.7%) 137 (63.4%) 0.372* HIV test within last year Yes 16 (23.9%) 48 (22.2%) 0.652* HIV Risk: own perception now or future Medium 16 (24.6%) 35 (16.2%) 0.570** High 19 (29.2%) 79 (36.6%) — Worries a lot about HIV Yes 27 (41.5%) 42 (19.4%) 0.003* Knows AIDS death personally Yes 50 (74.6%) 171 (79.2%) 0.657* Testing choice Self-test + VCT 62 (92.5%) 198 (91.7%) 0.842* No test 5 (7.5%) 18 (8.3%) — *p-Value from F-test (categorical variables). **p-Value from logistic regression (ordered categorical/continuous variables) adjusted for clustered design. ND, not determined. Approximately two-thirds of the women (118, 67.4%), but only one-third (55, 32.6%) of the men had previously tested for HIV (design-based F(1,3) = 56.20, p = 0.005). Recent HIV testing (within the last year) was reported by 47 (32.0%) women and 17 (12.5%) men (design-based F(1,3) = 18.01, p = 0.024). Uptake of Offer of Self-Testing and HIV Prevalence Choices made between the three testing options (no test, VCT only, self-test plus VCT) are shown in Table 1. All 260 (91.9%) participants who consented to VCT also opted to self-test, with the remaining 23 (8.1%) choosing not to self-test. Differences between those accepting and declining self-testing were minor, with significant differences being limited to a higher proportion of those who reported to be worried a lot about having HIV (design-based F(1,3) = 14.44, p = 0.016). Two participants with missing confirmatory results (one opt out, one incomplete form) were not included in accuracy analyses. HIV prevalence was 18.5% (48 of 260), with positive results in 37 (18.8%) of 197 randomly selected community adults, and 11 (17.5%) of 63 community peer-group members. HIV prevalence among participants who had previously tested HIV-negative or not tested at all was 12.0% (29 of 241 participants). Accuracy of Self-read Self-testing Compared to Confirmatory Blood Testing Self-testing was highly accurate, with clear and concordant results for 256 (99.2%; 95% CI 97.0–100.0%) of 258 participants with both self-test and blood results (Table 2). One HIV-positive participant was unable to interpret a faint positive test (recording the result as “uncertain”), and one HIV positive participant had no discernible line within the recommended test period. In both cases a repeat test performed by the counselor gave the same findings. 10.1371/journal.pmed.1001102.t002 Table 2 Cross tabulation of OraQuick and confirmatory blood test results in all participants and participants not already known to be HIV positive. OraQuick Self-test Positive Negative Inconclusive Unconfirmed Total All participants Positive 46 0 0 0 46 Negative 1 210 0 2 213 Inconclusive 1 0 0 0 1 Total 48 210 0 2 260 Excluding HIV-positive participants who have tested positive before Positive 27 0 0 0 27 Negative 1 210 0 2 213 Inconclusive 1 0 0 0 1 Total 29 210 0 2 241 Participants with no confirmatory results are not included in sensitivity and specificity analyses. Inconclusive results are considered to be inaccurate (false) results in sensitivity and specificity analyses. Overall sensitivity for self-test self-read was 97.9% (95% CI 87.9%–100.0%) with specificity of 100% (95% CI 97.8%–100%), respectively (Table 2). Results according to previous test results (positive, negative, or no previous test) are summarized in Figure 2. Table 2 shows the cross-tabulation of self-test and confirmatory results in 241 participants not previously known to be HIV positive. In this subgroup sensitivity was 96.4% (81.7%–99.9%) with specificity 100% (98.3%–100.0%). 10.1371/journal.pmed.1001102.g002 Figure 2 Accuracy of self-test self-read results against gold standard. Parallel testing of finger-prick blood by a counselor using Determine and Unigold, Bioline if discordant. Reported Ease of Self-testing, Need for Help and Errors At exit interview, 256 (98.5%) of participants rated self-testing as “very easy” to do. The visual and verbal demonstration of kit use was considered as useful as the illustrated instructions provided by the counselor when carrying out self-testing (250/260, 96.2% and 252/260, 96.9% respectively). Additional help beyond a brief demonstration and illustrated instructions was requested by 26 (10.0%) of self-test participants (Table 3), most commonly to clarify taking the mouth swab. Illiterate individuals (6/22, 27.3% compared to 15/236 literate, 6.4%; design-adjusted F(1,3) = 7.51, p = 0.071), those with lower levels of schooling (design-based F(1,3) = 7.93, p = 0.067) and women (15/134, 11.2% compared to 6/124, 4.8% men, design-adjusted F(1,3) = 12.14, p = 0.040) requested significantly more assistance. The nature of help requested is summarized in Table 3. 10.1371/journal.pmed.1001102.t003 Table 3 Help required and errors made by participants when self-testing. Category n Percent Needed help Asked counselor to watch and confirm correct approach to taking mouth swab 8 3.1 Asked for clarification after making a minor error 6 2.3 Asked for help opening developer fluid vial 4 1.5 Unable to use timer 2 0.8 Unable to take own mouth swab (asked counselor to do for them) 2 0.8 Unable to read result (asked counselor to read with them) 2 0.8 Wanted counselor present during collection 2 0.8 Total 26 10.0 Errors made Incorrect or incomplete swab of upper and lower gums 4 1.5 Touched flat pad 15 5.8 Spilt developer fluid 2 0.8 Fumbled vial or cap when opening developer fluid 1 0.4 Removed kit from developer too early 3 1.2 Read incorrectly (faint positive read as negative) 1 0.4 Total 26 10.0 Required second kit because of inconclusive result 2 0.8 Procedural errors were identified for 26 participants (10.0%, Table 3). The most serious errors, with highest potential to affect the results were: removing kit from developer too early (three participants, 1.2%) and spilling the developer fluid (two participants, 0.8%). Two participants (0.8%) required a second kit because of inability to read a very faint test line on a self-test result. Acceptability of Self-testing in the General Community Acceptability was assessed both before and after self-testing, with results summarized for adults from randomly selected households in Table 4. As detailed in the Methods, self-testing options were ranked second to door-to-door standard VCT by an external provider as being most likely to successfully increase HIV testing in the community (Table 4). Local provision of standard VCT by a neighbor was unacceptable to most women 71 (64.6%) and a substantial minority of men 41 (38.7%, design-based F(1,3) = 13.13, p = 0.036), whereas local distribution of self-test kits by a neighbor without having to disclose results was acceptable to 205 (94.5%) of participants with no differences between males and females (design-based F(1,3) = 0.03, p = 0.878). 10.1371/journal.pmed.1001102.t004 Table 4 Preferences and acceptability of HIV testing and counseling options. Adults from Randomly Selected Households in Blantyre n = 110 Women (%) n = 106 Men (%) p-Value* (Gender) Before self-testing Willingness to test if HIV services were provided by a neighbor? — — 0.036 Willing to self-test (do not have to reveal result) 66 (60.0) 38 (35.9) — Willing to accept both self-testing and standard VCT 38 (34.6) 63 (59.4) — Not willing to accept either self-testing or standard VCT 6 (5.5) 5 (4.7) — Strategy most likely to be successful in increasing testing in the community Door-to-door testing by counselor from outside community 59 (53.6) 50 (47.2) 0.737 Door-to-door testing with local person as counselor 11 (10.0) 15 (14.2) — Mobile clinics providing VCT 12 (11.0) 10 (9.4) — Free self-test kits available from local vendors and stores 14 (12.7) 6 (5.6) — Door-to-door self-testing provided by local person 14 (12.7) 25 (23.6) — Either of above self-testing option as most/next-most successful 68 (61.8) 76 (71.7) — Agree strongly/somewhat “these days if people have tested before” No need for counseling 12 (10.9) 10 (9.4) 0.378 Telephone hotline sufficient 32 (29.1) 28 (26.4) 0.282 Information leaflet 42 (38.2) 33 (31.1) 0.323 Community worker available 54 (49.1) 53 (50.0) 0.67 At exit interview Would recommend this kit for self-testing to friends and familya 100 (100) 97 (100) ND If testing for HIV in future would want next HIV test to be — — 0.078 In a VCT centre 3 (2.7) 12 (11.4) — At a hospital or clinic 25 (22.7) 13 (12.4) — Provided by a counselor at home (door-to-door service) 16 (14.6) 11 (10.5) — Self-testing at home 64 (58.2) 68 (64.8) — Testing campaign in community (mobile clinic/stand) 2 (1.8) 1 (1.0) — a Question only asked to those who accepted self-testing (100 women and 97 men). *F-test adjusted for the clustered design. ND, not determined. After self-testing, all participants would recommend self-testing to friends and family. Self-testing at home was the preferred option for future HIV tests among both men and women (Table 4) and did not significantly vary by age, marital status, schooling, or socioeconomic status (unpublished data). Counseling Despite the popularity of the concept of self-testing, all but a small minority of individuals (22/216, 10.2%) disagreed that people “these days know enough about HIV to do without counseling for a repeat HIV test.” Alternatives to face-to-face counseling for repeat testers, such as telephone helpline, information leaflets, or availability of a local community health worker were also not considered acceptable substitutes by most participants (Table 4). Discussion The main findings of this study of self-testing for HIV are that high levels of accuracy (99.2%) were obtained from oral self-testing after a brief demonstration and illustrated instructions, and that there were strong indicators of community readiness to adopt self-testing alongside other HIV counseling and testing strategies in an African urban high HIV prevalence setting. Self-testing was the preferred option for future HIV tests for 56.4% of participants, being the most common choice for both men and women. There was no obvious subgroup for whom self-testing was not acceptable. These findings are strengthened by our having used random selection of participants from the main high-density residential areas in urban Blantyre, and by the 95.6% participation rate. High-density areas are particularly interesting for self-testing because of the difficulty in securing confidential space for HTC due to lack of soundproofing and intense crowding. The high accuracy contrasts with much lower levels obtained from blood-based self-testing with a less simplified test kit in Singapore [31], but is in the same range as reported for oral self-testing in Americans attending public hospital services [24],[32]. Also remarkable is that 91.9% of individuals who agreed to be interviewed opted to test for HIV, all choosing self-test plus VCT. This finding concurs with results from other recently conducted community-based studies [4],[6] and indicates that knowledge of HIV status has now become sufficiently desirable in many parts of rural and urban Africa to make universal knowledge of recent HIV status by all adults a realistic goal. Given the accuracy achieved and strong preferences around future testing, we consider our results to be sufficiently promising to warrant further exploration of self-testing options as a potential way to make progress towards meeting Universal Access goals [8]. The need for innovation around community-based testing is underscored by the identification of previously undiagnosed HIV in 29 (12.0%) of our participants in this current study. Despite the highly effective public health approach to HIV testing and care taken in Malawi, which is among the most successful programmers in Africa and considered a model for the continent [33],[34], less than half of HIV-infected participants were previously diagnosed, and just over half of undiagnosed HIV infections were in individuals who had previously tested HIV negative. Self-testing may prove especially valuable for encouraging regular repeat-testing, couple testing, and first-time testing in otherwise hard-to-reach groups such as men and older individuals [5],[8],[14],[15]. Other investigators have shown high acceptability, uptake, and high accuracy of oral self-testing in American populations [24],[32], and home-collection of specimens for laboratory HIV testing and telephone results and counseling is well established in America and also available in Europe [15],[16],[18]. Kenya was the first country in Africa to develop policy guidelines for self-testing, although no kit has yet been fully registered [13]. The manufacturers of the kit used in this study have applied to the US Food and Drug Administration (FDA) for registration as an over-the-counter test, and it is likely that self-testing options will become available in many different countries over the next few years. The high uptake and high proportion of participants selecting self-testing at home as the preferred “next test” option in this current study suggest that carefully considered promotion of self-testing has potential to contribute greatly to scale-up of HIV testing in Africa. For example HIV prevention through “universal test-and-treat” [23]–[26],[35],[36] strategies are being considered with the goal of reducing HIV transmission by prompt identification and initiation of antiretroviral treatment in HIV-infected individuals irrespective of extent of immunosuppression. To be effective such strategies require a high proportion of HIV-infected individuals in communities to be diagnosed and started on treatment within 12 mo of infection, calling for highly acceptable and convenient initial and repeat HIV testing strategies. The high uptake of home-based testing for HIV and much higher rates of couple testing than alternative strategies, make it the ideal community arm to complement facility-based routine provider-initiated testing in countries with generalized HIV epidemic [3]–[7]. However, sustainability is a key issue, which is compounded by the requirement for counselors who come from outside the community [4],. Home-based self-testing provides privacy from neighbors and may allow development of novel local community strategies: although most of our participants were not willing to accept VCT from a neighbor, 94.9% indicated willingness to accept a self-test kit from a neighbor if they did not have to confide the results. Support is still needed at the time of a positive test, with most participants in the current study considering telephone hot line or information leaflets to be inadequate replacements for face-to-face counseling. Operational research will be needed to investigate whether or not linkage into HIV care is any worse after self-testing than it is for current facility-based HIV testing and counseling services in Africa (estimated linkage into care of less than 50% within 1 y), and ways of providing psychological support following a positive self-test. Examples of potential models include 24-h telephone hotlines, or community workers trained in pretest counseling, distribution of kits, and available for face-to-face post-test counseling. The private nature of self-testing, low potential for inadvertent breach of confidentiality, and ease of incorporating self-testing into busy daily lives, particularly for couple testing, should ideally be retained. Limitations of the study include the lack of data on linkage into care following a positive HIV test. This area needs to be investigated in any program implementing self-testing, as risk of drop-out between testing HIV positive and entering HIV care is known to be high even under standard VCT models [37]. Linkage into care was not investigated in this study because self-testing was offered in a supervised context, as a first step, so that HIV counselors were able to immediately refer participants into care. Our data on acceptability and accuracy include 19 (7.3% of 260) participants who already know themselves to be HIV-positive and so may have been more willing to test and able to read positive test results accurately. Regulatory strengthening may be required to prevent sales of counterfeit and substandard test kits (as for example has been described in TB diagnostics) [38]. The need to avoid storage over 27°C may be a practical barrier to scale up in hotter climates, although this same recommendation applies to most rapid HIV test kits currently in programmatic use. In conclusion, we have shown high willingness to self-test for HIV at home when self-testing was linked to confirmatory blood tests, with self-testing the preferred option for future tests. Highly accurate results were obtained by randomly selected African adults in a poor urban setting following a brief demonstration and illustrated instructions. Africans have been living with the spectra of HIV for two decades, bringing personal loss, hardship, and a more uncertain future to almost everyone in the high HIV prevalence Southern region [8],[9],[12]. With the scale up of HIV care, home-based testing and counseling has become a welcome intervention with very high uptake rates [4]–[7],[9]. In settings where regular annual visits to every home by external VCT providers are not feasible, options based on self-testing may offer a more readily sustainable approach that can contribute towards Universal Access goals, provided that mechanisms can be identified to ensure that safety, accuracy, and post-test support are not unduly compromised.
          • Record: found
          • Abstract: found
          • Article: found
          Is Open Access

          Supervised and Unsupervised Self-Testing for HIV in High- and Low-Risk Populations: A Systematic Review

          Introduction On July 3, 2012 the US Food and Drug Administration (FDA) approved an oral point-of-care (POC) HIV self test, OraQuick, for over-the-counter (OTC) sale on the basis of a positive recommendation from their Blood Product Advisory Committee (FDA-BPAC) [1]. This decision, the first for a self test for an infectious disease, is an important step towards normalizing the process of receiving an HIV diagnosis, currently beset with stigma and discrimination. The FDA approved the oral HIV test even though it is less sensitive than a blood test because of its potential to allow more people to know their sero-status and thus potentially avert thousands of cases of HIV transmission. Although this approval has paved the way for a self-testing paradigm complementary to facility-based testing, evidence is needed to understand whether self-testing will lead to more people knowing their HIV status and whether self-testing can be implemented, and operationalized as a strategy in global settings. If HIV self-testing is to realize its promise of increasing the number of at-risk individuals knowing their sero-status evidence is needed to demonstrate that self-testing strategies can be an acceptable or preferred mode of HIV testing. Crucially, it is important to demonstrate that individuals who are self-testing can be given reasonable assurance of test accuracy, especially for populations with varying backgrounds and literacy levels. Furthermore, evidence on whether self-testing will offer a private, confidential alternative to facility-based HIV testing with a safe conduit to care and treatment is needed to improve outcomes for both individuals and at-risk populations. If self-testing is proven to help increase knowledge of sero-status in those individuals that do not seek facility-based testing and improves linkages to care and treatment rates in the community, then it will stand to impact control HIV at the population level. However, as of 2013, that vision remains unrealized. Facility-based HIV testing strategies (voluntary testing and counselling, provider-initiated and conventional client-initiated testing and counselling) have been in place for decades [2]. However, stigma and discrimination faced in these settings remain key barriers to testing [3]. Additional barriers include fear of visibility, fear of lack of confidentiality of a positive test result, a lack of privacy, and increased waiting time to obtain a test result [4]. It is no surprise therefore that about six in ten individuals living with HIV infection remain untested globally and as a consequence, are unaware of their HIV sero-status [5]. In this context, self-testing offers some promise in alleviating or eliminating these barriers. It provides individuals with the option of knowing their HIV status in the privacy of their homes and therefore has the potential to negate the effects of stigma, and reduce perceived discrimination. Although self-testing has the potential to help increase the numbers of individuals aware of their HIV sero-status, this strategy has been beset with concerns and challenges about its success [6]–[8]. This is also partly attributed to a lack of evidence on self-testing behaviors, and effective linkage initiation to treatment and counselling after self-testing. To our knowledge, only one published narrative review has evaluated the potential role of self-testing in health care worker populations [2]. A related meta-analysis compared the diagnostic accuracy of oral versus finger-stick antibody POC tests to demonstrate that oral POC tests could play a role in future self-testing initiatives [9]. Another systematic review on home-based rapid testing initiatives where people are encouraged to get tested and counselled by health care workers in home visits (thus different from self-testing) alluded to the potential of offering self-testing to individuals in resource constrained settings [10]. With a view to fill a knowledge gap in understanding supervised and unsupervised self-testing strategies that are being evaluated globally, and to guide their effective implementation in high-income and resource constrained settings, we performed a systematic review. Methods We systematically reviewed the literature to evaluate the current evidence on two common strategies: (a) supervised and (b) unsupervised self-testing strategies in high- and low-risk populations worldwide. A supervised strategy (self-testing and counselling processes) was always aided by a health care professional (HCP). An unsupervised strategy was performed by a self-tester without any help, but with counselling and linkage to care offered off-site (e.g., over the phone) by a HCP. Our specific objectives included documentation of all outcomes from implementation research associated with self-testing and counselling strategies (acceptability, accuracy, feasibility, cost, and counselling preferences). Further, data on challenges, concerns, and barriers documented from qualitative research or mixed methods studies were also synthesized. This review was reported following PRISMA guidelines. Search Strategy and Identification of Studies For the period of January 1, 2000–October 30, 2012, we searched seven electronic databases (Medline [via PubMed], Biosis, PsycINFO, Cinahl, African Medicus, LILACS, and EMBASE) and abstracts from six major HIV/sexually transmitted infections (STIs) conferences (Canadian Association of HIV Research [CAHR], International Society for Sexually Transmitted Diseases Research [ISSTDR], International AIDS Society [IAS], Conference on Retroviruses and Opportunistic Infections [CROI], Infectious Diseases Society of America [IDSA], and the Inter Science Conference on Antimicrobial Agents and Chemotherapy [ICAAC]). Additionally, we reviewed bibliographies and contacted the authors for original data. We included abstracts if full-texts were not available. Our search string (limited to humans) was: (1) “HIV”[MeSH] OR “HIV Seropositivity”[MeSH] OR “HIV Infections”[MeSH] AND (2) (“Self Care”[MeSH]) OR “Self Administration”[MeSH]) OR “Point-of-Care Systems”[MeSH] OR “self*test*” OR “rapid*test*.” Study Selection Two reviewers (JS and SS) independently screened all citations. Please refer to the flow chart for details (Figure 1). 10.1371/journal.pmed.1001414.g001 Figure 1 Flow chart of study search and selection. Eligibility Criteria Full-text articles, brief reports, or abstracts that evaluated HIV self-testing strategies in any part of the world were included. Reviews, perspectives, editorials, and studies that did not evaluate self-testing strategies (home-based non self test) were excluded (Figure 1). Data Abstraction Two reviewers independently abstracted data from quantitative (JS and SS) and qualitative (SP and JS) studies. Concordance between reviewers was high at 90%. Disagreements were resolved by consultation with a third reviewer (NPP). A pre-piloted data abstraction form was used. Variables such as study characteristics, populations, study design, type of strategy, and outcomes were tabulated (Tables 2–4). 10.1371/journal.pmed.1001414.t002 Table 2 Characteristics of included studies. Test Strategy Author Year Study Setting Sample Size Survey Response Rate HIV Self-Test Type Study Design Population Summary Score for Quality Critiquea Test strategy: supervised Skolnik 2001 [30] USA 354 365/380 NA Quantitative survey HIV clinic attendees 56% (18/32) Spielberg 2003 (a) [32] b USA 240 NA Oral and finger-stick Quantitative cross-sectional HIV positive patients 55% (6/11) Spielberg 2003 (b) [31] USA 460 460/865 NA Quantitative survey MSM. Persons from needle exchange site STI clinic attendees 63% (20/32) Lee 2007 [27] Singapore 350 NA Finger-stick Quantitative cross-sectional STI clinic attendees 69% (22/32) Spielberg 2007 [19] b India 27 27 NA Qualitative survey General population, brought to community internet center NA Chavula 2011 [17] b Malawi 92 92 Oral (post-survey) Qualitative survey General urban population NA Choko 2011 [33] Malawi 283 NA Oral Quantitative cross-sectional General urban population 72% (23/32) Gaydos 2011 [24] USA 478 NA Oral and finger-stick Quantitative cross-sectional Emergency department 59% (19/32) MacPherson 2011 [18] Malawi 216 216/226 NA Quantitative survey General urban population 59% (19/32) Carballo-Dieguez 2012 (a) [16] USA 57 NA Oral Qualitative and quantitative cross-sectional Urban MSM 59% (19/32) Pant Pai 2012 [29] b Canada 100 NA Oral Quantitative and qualitative cross-sectional University students 82% (9/11) OraSure 2012 (phase IIb) [13] USA 1,031 NA Oral Quantitative cohort study Known HIV positives, general population 29% (10/34) Belza 2012 [21] Spain 208 NA Finger-stick Quantitative cross-sectional Attendees at a rapid HIV testing site 41% (13/32) Ng 2012 [28] Singapore 994 NA Oral Quantitative cross-sectional HIV positive, at-risk patients from family practice clinics 66% (21/32) Test strategy: unsupervised Kalibala 2011 [12] Kenya 765 NA Oral Qualitative and quantitative cross-sectional Health care professionals 28% (9/32) Katz 2012 [25] b USA 108 NA Oral RCT Urban MSM 59% (10/17) OraSure 2012 (phase III) [13] USA 5,798 NA Oral Quantitative cohort study General population high- (n = 4,689) and low- (n = 1,109) prevalence sites 29% (10/34) Fuente 2012 [23] Spain 313c NA Finger-stick Quantitative cross-sectional Attendees at a rapid HIV testing site 56% (18/32) Lee 2012 [26] b USA 500 NA Oral Quantitative cross-sectional General population at unknown risk of HIV 64% (7/11) Carballo-Dieguez 2012 (b) [22] b USA 28 NA Oral Qualitative and quantitative cross-sectional HIV uninfected, urban non-monogamous MSM 55% (6/11) Helm 2012 [20] b Netherlands NA NA Oral NA NA NA a The summary score for quality critique represents the number of criteria reported, over the total number of criteria. b Sample size for “cost preference and willingness to pay (WTP) (USD)” and “feasibility linkages errors” (Table 3) outcomes was 519, as data were reported in combination with participants from another testing program. c Abstract. NA, not available/not applicable; STI, sexually transmitted infection. 10.1371/journal.pmed.1001414.t003 Table 3 Study outcomes: acceptability, accuracy, agreement and cost preference. Test Strategy Author Year Study Setting Acceptability Accuracy Agreement or Concordance Cost Preference and WTP (US$) Test strategy: supervised Skolnik 2001 [30] USA 24% NA NA NA Spielberg 2003 (a) [32] USA NA NA Oral fluid test: 95%, blood-based test: 89% 70% WTP≤US$15; 40% WTP US$20 Spielberg 2003 (b) [31] USA NA NA NA WTP US$10–US$15 Lee 2007 [27] Singapore NA NA κ value = 0.28 (p US$15 Chavula 2011 [17] Malawi NA NA NA NA Choko 2011 [33] Malawi 92% Sensitivity: 97.9% (95% CI 87.9–100) Specificity: 100% (95% CI 97.8–100) NA NA MacPherson 2011 [18] Malawi 92% NA NA NA Test strategy: unsupervised Kalibala 2011 [12] Kenya 78% NA NA NA Katz 2012 [25] USA NA NA NA 45% WTP≤US$20; 25% WTP US$20–US$40, 17% WTP≥US$40, 13% WTP free OraSure 2012 (phase III) [13] USA NA High-prevalence setting: Sensitivity: 92.9% (95% CI 86.5–96.89); Specificity: 99.98% (95% CI 99.87–100); PPV: 99.1% (95% CI 94.86–99.98); NPV: 99.81% (95% CI 99.63–99.92) Low-prevalence setting: Sensitivity: 100% (95% CI NA; 0 FN); Specificity: 100% (95% CI 99.66–100); PPV: 100% (95% CI 2.5–100); NPV: 100% (95% CI 99.66–100) NA NA Fuente 2012 [23] Spain NA 8% (95% CI 4.8–11.2) invalid tests NA 18% WTP>US$38, 22% WTP US$25–US$38, 5.2% WTP free Lee 2012 [26] USA NA Specificity: 99.8% (95% CI 98.1–100), 1.8% testing error NA NA Carballo-Dieguez 2012 (b) [22] USA 84% NA NA NA Helm 2012 [20] Netherlands NA NA NA NA FN, false negative; K, kappa statistic; NA, not available; NPV, negative predictive value; PPV, positive predictive value; WTP, willingness to pay. 10.1371/journal.pmed.1001414.t004 Table 4 Study outcomes: counselling preference, feasibility, linkages, errors, motivation, label comprehension, and test preference. Test Strategy Author Year Study Setting Counselling Preference Feasibility, Linkages, Errors Motivation, Label Comprehension Test Preference Supervised self-testing strategies Skolnik 2001 [30] USA 100% for in-person pre and post test NA Convenience, speed, privacy, and anonymity NA Spielberg 2003 (a) [32] USA NA Errors noted in placing test device in developer solution NA 61% preferred testing at home Spielberg 2003 (b) [31] USA NA NA NA 20% prefer home self-testing versus conventional test Lee 2007 [27] Singapore 79% for post-test counselling NA Convenience, speed, privacy, and anonymity NA Spielberg 2007 [19] India Computer-based pre and post test counselling NA Convenience, speed, privacy, and anonymity NA Gaydos 2011 [24] USA NA 5%–10% difficulties in test performance and test interpretation NA 91% preferred oral fluid versus blood-based tests Carballo-Dieguez 2012 (a) [16] USA NA Errors in conduct: (1) touch test pad; (2) swab multiple times; (3) eating/drinking just before taking the test; (4) almost drinking the solution in the vial 87% for would likely self-test if available OTC and 80% would likely use it to test partners at home NA Pant Pai 2012 [29] Canada 78% for post-test at community clinics, 53% post-test on the phone, 31% at pharmacies, 29% online NA 98% convenience, 96% time efficient, 84% pain free NA OraSure 2012 (phase IIb) [13] USA NA 1.82% error rate in population of unknown status; 4.76% error in HIV positive population NA NA Belza 2012 [21] Spain NA 1% invalid tests NA NA Ng 2012 [28] Singapore 72.5% for pre-test counselling; 73.9 for post-test counselling Errors due to conduct: (1) use of collection pad to swap external lips; (2) touching the swab during removal from packaging; (3) spilling the test solutions; (4) misinterpret negative or invalid test results Convenience, speed, privacy, and anonymity; kit instructions easy to understand 87.4% would but an OTC rapid test kit and 89% wanted to conduct HIV testing in private Chavula 2011 [17] Malawi Post-test counselling considered essential NA NA NA Choko 2011 [33] Malawi 90% preferred pre-test; 70% prefer in-person counselling over telephone counselling or information leaflets Errors in conduct: (1) early removal of kit from the developer; (2) spilling the developer fluid NA NA MacPherson 2011 [18] Malawi NA NA NA NA Test strategy: unsupervised Kalibala 2011 [12] Kenya Telephone-based counselling NA NA Katz 2012 [25] USA NA Convenience, speed, privacy, and anonymity NA OraSure 2012 (phase III) [13] USA 88% sought post-test counselling Test system failures: interpretational and operational errors: high-prevalence settings: 1.25% (95% CI 0.95%–1.63%); low-prevalence settings: 0.37% (95% CI 0.10%–0.93%); 96% newly diagnosed HIV positive subjects wanted to seek linkages and follow up; 97% would recommend oral self tests to others, 79% would use for self test NA Fuente 2012 [23] Spain NA 5.4% (95% CI 3.4–7.4) misinterpretation of self-test result picture; 6.6% with valid results find instructions “somewhat” or “quite difficult” versus 20% with invalid results 83.9% felt more motivated after taking the test to self-test in future Lee 2012 [26] USA NA NA High-label comprehension among intended user populations 98.8% (95% CI 97.4–99.6) NA Carballo-Dieguez 2012 (b) [22] USA NA NA Availability of OTC would increase testing frequency “High acceptability” among ethnic minority participants and ethnic minority sex partners Helm 2012 [20] Netherlands NA NA NA NA NA, not available; WTP, willingness to pay. Quality Assessment and Data Synthesis A quality critique of quantitative data from cross-sectional (Tables S1 and S2) and cohort studies (Table S3) was performed using the STROBE reporting checklist [11]. Two articles, although not peer reviewed, were critiqued using the STROBE checklist as they were reporting outcomes of cohort and cross-sectional studies [12],[13]. Similarly, a conference abstract reporting a RCT (Table S4) was appraised using the CONSORT guidelines [14]. A guide [15] for critically appraising qualitative research was used to appraise qualitative studies [12],[16]–[19]. The only study that could not be quality critiqued was an announcement of an implementation strategy through a conference abstract [20]. Due to lack of standardized reporting of primary and secondary outcomes, a meta-analysis was not conducted. Results A total of 1,207 studies were identified from databases and bibliography searches and 14 abstracts retrieved from conferences for a total of 1,221 citations (please refer to Figure 1). After removing duplicates, 1,108 citations were reviewed in the first screen (Figure 1). Of 40 articles that were reviewed in the second screen, 20 were included. Reasons for exclusion of the remaining 20 studies were: repeat of reporting in abstract and full-text (n = 4), narrative review (n = 1), and topic irrelevant to this review (n = 15). One article reported on two separate studies, thus, a total of 21 studies were synthesized. Table 1 provides a list of definitions for the key outcomes or characteristics documented in this review. Table 2 presents the description of the 21 included studies, and Tables S1 to S4 present their detailed STROBE and CONSORT quality reporting assessment, where applicable. 10.1371/journal.pmed.1001414.t001 Table 1 Definition of outcomes. Outcome Definition Acceptability Numerator defined as those individuals who chose to self-test. Denominator defined as all those who were offered and consented to test. Uptake = numerator/denominator (computed as a percentage) Accuracy Accuracy was defined by sensitivity and specificity parameters. Index test was a self-test result. Reference standard tests were combination of conventional lab tests for HIV (rapid tests or ELISA with p24 and/or Western blot depending on high- versus low-resource setting) Concordance Concordance for self-testing was reported as a measure of agreement between the test result between the individual and health care worker quantified either as percentage agreement or with the Cohen's Kappa (κ) inter-rater agreement Feasibility Documented completion of self-testing and counselling process. Includes ease of performance and interpretation of self-testing results, and documentation of errors, initiation of linkages. Motivators Factors contributing to the acceptability of HIV self tests Description of Included Studies Of the 21 studies included, 16 (n = 16/21, 76%) were conducted in high-income [13],[16],[20]–[32] versus five (n = 5/21, 24%) in resource constrained country settings [12],[17]–[19],[33]. Total sample size varied from 27 to 5,798. Two main strategies for HIV self-testing were identified, supervised and unsupervised, and the studies classification is illustrated in Figure 2. The total sample size for the supervised testing strategy was 4,890 individuals and 7,512 for the unsupervised testing strategy. Our review provides data from 1,383 participants in resource constrained settings, compared to 11,019 in high-income settings, thus the bulk of data (89%) was from high-income settings. Study populations varied from high at-risk for HIV to low-risk general populations. 10.1371/journal.pmed.1001414.g002 Figure 2 Self-testing strategies: a classification. A vast majority of studies evaluated oral self tests (n = 14/21, 67%), while others used finger-stick-based tests. The two self-testing strategies, unsupervised and supervised, were evaluated with two oral POC tests, OraQuick and Calypte [12],[13],[16],[17],[20],[22],[24],[26]–[28],[29],[32],[33]. With the exception of one completed randomized controlled trial (RCT) [25], all other studies (n = 20/21, 95%) were observational, of which 14 (n = 14/20, 70%) were cross-sectional or cohort studies [12],[13],[16],[21]–[24],[26]–[29],[32],[33], while five (n = 5/20, 25%) were surveys [17]–[19],[30],[31], and one reported a study in progress that evaluated an unsupervised testing strategy coupled with Internet counselling for high-risk populations in the Netherlands [20]. A supervised self-testing strategy was evaluated in 14 (n = 14/21, 67%) studies, in both high- and low-risk populations with varying level of education, and access to resources. Although study participants performed the self test themselves, a HCP assisted the self-tester in any aspect of self-testing and counselling, if called upon. This assistance varied across studies, e.g., understanding the conduct of self-testing, helping with result interpretation, counselling, and initiating linkages for confirmatory testing. In the unsupervised self-testing strategy (n = 7/21 studies, 33%), no assistance was offered by HCPs in the conduct and interpretation of self tests, but counselling was available off-site (on the phone or over the Internet). Unsupervised self-testing was evaluated primarily among urban and educated populations in high-income and resource constrained settings, such as a proposed evaluation in the Netherlands [20], HCPs in Kenya [12], urban men who have sex with men (MSM) in the US [25], general literate populations in the US [13], general population at unknown risk of HIV [26], urban non-monogamous MSM in the US [22], and attendees of a rapid HIV testing site in Spain [23]. Quantitative Data Table 1 provides the definitions used for reporting outcomes, facilitating their interpretation, documentation, and synthesis stratified by strategies. Acceptability For both supervised and unsupervised self-testing strategies, acceptability (Table 3) was high (range: 74%–96%). For supervised strategies, acceptability was documented in seven studies. Acceptability varied from 74% (n = 42/57) in MSM in the US [16], to 78% (n = 208/267) in clinic attendees from Spain [21], to 85% (n = 478/564) in clinic attendees from the US [24], to 92% (n = 198/216 and n = 206/283) from household participants in Malawi [18],[33], to 95% (n = 100/105) in educated student populations in Canada [29], but was much lower (n = 85/354, 24%) in an early 2001 study in the US [30]. It should be noted that acceptability was measured in a research context, where participants voluntarily show up for self-testing. For unsupervised strategies, acceptability was documented in two studies, and ranged from 78% (n = 230/295) in HCPs in Kenya [12] to 84% (n = 27/32) in non-monogamous urban MSM in the US [22]. Accuracy Accuracy of self tests was reported by sensitivity and specificity parameters with the index self test being compared to the reference standard prevalent in each research setting. In high-resource country settings, the reference standard algorithm used was dual ELISA and Western blot performed on blood samples. In resource constrained settings, the three rapid test algorithm was employed. Across both strategies, irrespective of the reference standard used, the specificity of the HIV self tests was consistently documented to be high, but sensitivity estimates varied greatly. In the supervised testing strategy, high overall accuracy was generally observed. For example, a high sensitivity (97.9%; 95% CI 87.9–100) and a high specificity (100%; 95% CI 97.8–100) was reported in Malawi [33], as well as in a US-based study (sensitivity 97.9%; 95% CI 95.0–99.4—specificity 99.8%; 95% CI 98.1–100) [13], and in a Singapore-based study (sensitivity 97.4%; 95% CI 95.1–99.7—specificity 99.9%; 95% CI 99.6–100) [28]. However, in the unsupervised strategy, one study from the US reported a slightly lower sensitivity of 92.9% (95% CI 86.5–96.9) in a high-prevalence setting (HIV prevalence>1%), compared to a sensitivity of 100% (95% CI 99.7–100) in a low-prevalence setting (HIV prevalence≤1%) [13]. Agreement and Concordance Agreement and concordance between the self-tester and HCP could only be reported and computed for supervised testing strategies. Six studies reported data on concordance of self-testing results compared to test performed by HCPs [21],[24],[27]–[29],[32]. The five studies using oral self tests noted a higher agreement for oral self tests (95%–100% or κ = 0.97) [21],[24],[28],[29],[32] in comparison to finger-stick self tests (κ value = 0.28) [27]. Furthermore, a time trend was observed across studies, with a lower agreement reported in early studies (from 2001). Feasibility Ten studies evaluated the feasibility of self-testing strategies, with various feasibility outcomes documented across studies. Feasibility outcomes included test conduct, test interpretation, post-test counselling, and treatment linkages. In the unsupervised strategy, about 95% (n = 37/39) of urban MSM in the US found the kits to be “very easy to use” and no test errors were reported [25]. However, an implementation study evaluating the same strategy in the US in general populations reported on operator errors (interpretational and observational) [13]. These failures varied from 0.37% (n = 4/1,093) in low-prevalence settings to 1.25% (n = 56/4,465) in high-prevalence settings. A Spanish study evaluating an unsupervised strategy reported an overall interpretation error rate of 5.4% (n = 28/519) [23], which included a 1.1% (n = 11/1,038) error in interpretation of a positive result as negative. Seven studies that evaluated supervised strategies documented errors in test performance [13],[16],[21],[24],[28],[32],[33]. Errors documented in general population in urban Malawi requested significantly more assistance from HCPs during self-testing with oral tests; errors were noted as a result of early removal of kit from the developer and spilling of the developer fluid [33]. Two studies from Singapore and Spain with finger-stick tests reported a high degree of errors, where 0.5% (n = 1/200) of participants obtained invalid self-test results [28] and about 1% (n = 2/208) of the finger-stick self-test results were found to be invalid [21]. In comparison, test conduct and performance errors noted with oral tests were: (i) collecting the oral mucosal sample, (ii) reading test instruction, and, (iii) interpreting test results [24]. In the study that evaluated both strategies, a higher error rate (4.76%, n = 24/504) was observed in test performance and interpretation by populations who tested positive for HIV in a supervised setting compared to unscreened or low-risk populations in an unsupervised strategy (1.25%, n = 56/4,465) [13]. This difference was attributed to the fact that, being already aware of their sero-status, individuals testing positive for HIV that self-tested did not care to read the instructions carefully and subsequently made more errors than unscreened populations. Two studies from the US have demonstrated that an improvement in self-test instructions for test conduct and interpretation reduce the incidence of errors [27],[32]. Data on post-test linkages were sparse. Linkage data were reported in only one study from the US that has evaluated an unsupervised strategy. It demonstrated that 96% (n = 102/106) of participants testing positive for HIV stated they would seek post-test counselling [13]. Overall in both strategies, a pattern of errors in self-test conduct was noted: (1) failing to place the oral test device in the developer solution after swabbing [32], (2) removing the test kit from the developer solution too early [33], (3) spilling the developer fluid [13],[28],[33], and lastly (4) dipping the test device in the developer solution before swabbing the gums [13],[16]. Likewise, errors in self-test interpretation were also reported: (1) difficulty in interpreting the test result on the device [24],[28] and (2) inability to read or interpret faint or weakly positive test lines on the test device [13]. Motivators to Self-Test Across all settings, both for the supervised and unsupervised self-testing strategies, motivators to self-testing were: (a) convenience, (b) speed and time to test result, (c) privacy, (d) a sense of empowerment, and, lastly, (e) a control of one's health choices [13],[16],[19],[22],[25]–[30]. In a supervised strategy evaluated in an educated population in Canada, convenience (99%, n = 99/100), time efficiency (97%, n = 97/100), and the pain-free procedure (84%, n = 84/100) were identified as motivators to self-testing [29]. In supervised strategies, in Malawi, participants preferred the distribution of test kits by one person (neighbourhood counsellor or trained person) and counselling by another person to maintain anonymity of self-testing [17]. Similarly, in a study with HIV clinic attendees in the US, 92% (n = 67/73) of participants cited provision of instant results, and 45% (n = 33/73) mentioned anonymity, confidentiality, and privacy as factors favoring self tests over conventional tests [30]. Likewise, in another supervised strategy evaluated in non-monogamous urban MSM in the US, 87% (n = 50/57) were likely to buy self tests, and 80% (n = 46/57) wanted to use it to test partners at home [16]. In unsupervised strategies, 7% (n = 19/288) of participants with valid results and 20% (n = 5/25) with invalid results found the test instructions “somewhat” or “quite” difficult to read, while most participants did not document difficulties following the test kit instructions [23]. High comprehension of instructions for use was also observed among intended user population in a study based in the US [26], and a study among urban non-monogamous MSM in the US documented that OTC availability of self tests would increase the frequency of testing in this population [22]. In another study evaluating an unsupervised testing strategy, a high percentage (79%, absolute number not reported) of HIV negative subjects expressed a desire to use a self test at home if available OTC, and 97% (absolute number not reported) of subjects newly diagnosed with HIV expressed a desire to recommend this OTC test to a friend [13]. Preference Three preference outcomes were documented: (a) preference for test and strategies, (b) cost preference and willingness to pay, and (c) preference for counselling. Preference for tests and strategies This outcome was variously documented. It included: (a) preference for oral versus finger-stick self test, (b) preference for self-testing strategy over conventional testing strategy, and (c) preference for self-testing of partners. A majority (n = 14/21, 67%) of studies evaluated oral fluid tests for both strategies, and preference was attributed to the tests' non-invasiveness, convenience, and ease of specimen collection. Furthermore, in studies from the US where participants were offered a choice between oral and finger-stick, 91% (n = 433/478) chose oral fluid tests over blood-based tests (9%, n = 45/478) in one study [24], while in the other, 61% (n = 146/240) of participants preferred the oral test over the finger-stick test [32]. In a Canadian study with students, 81% (n = 81/100) preferred self tests over conventional lab tests, and 74% (n = 74/100) expressed a desire to buy them OTC [29]. For unsupervised strategies, in another US study in general populations and individuals of known positive HIV status, about 79% (absolute number not reported) preferred to self-test using oral over conventional tests, and 97% (absolute number not reported) wished to use oral tests to test their partners [13]. In Spain, 84% (n = 436/519) of the participants undergoing unsupervised testing felt more motivated after taking the test to self-test in the future [23]. Finally, a study among urban non-monogamous MSM in the US documented “high acceptability” of the unsupervised self-testing strategy among ethnic minority participants and their sex partners [22]. Cost preference and willingness to pay Cost preference for self tests and willingness to pay if sold OTC were documented for both supervised and unsupervised strategies, and varied across populations, settings and strategies [23],[25],[27],[29]–[31],[32]. In supervised strategies, in Canada, 32% (n = 32/100) of university students were willing to pay up to US$10 and 41% (n = 41/100) up to US$20 [29], while in Singapore, 88% (n = 370/420) were willing to pay between US$7 and US$13, and 28% (n = 118/420) more than US$15 [27]. In a study in the US, 70% (n = 168/240) were willing to pay up to US$15, and 40% (n = 96/240) would be willing to pay US$20 for it [32]. In another study in the US, at-risk participants from homeless shelters wanted free self tests [30]. In unsupervised strategies evaluated in an urban MSM population, 45% (n = 49/108) were willing to pay less than US$20, 25% (n = 27/108) between 20 and US$40, 17% (n = 18/108) more than US$40 and 13% (n = 14/108) wanted it free [25]. Additionally, in Kenya, HCPs were unwilling to pay and wanted the government to provide it for free, as HIV was perceived to be an occupational risk [12]. A study in Spain evaluating an unsupervised strategy in attendees at a rapid HIV testing site documented 18% (n = 56/313) willingness to pay more than US$38, 22% (n = 69/313) between US$25 and US$28 and 5% (n = 16/313) wanting it for free [23]. Preference for counselling Preferred mode and medium of counselling varied by strategies, as well as by educational and socio-economic status of the self-testers, which dictated their access to resources such as mobile phones, Internet, and pharmacies. In both supervised and unsupervised strategies, participants agreed that both pre- and post-counselling were essential components that needed to be integrated in the testing process [12],[13],[17],[19],[27]–[30],[33]. Regarding preference for counselling in supervised strategy, in two studies from Singapore, 79% and 74% (n = 275/350and n = 735/994) of participants felt that confidential post-test counselling was essential [27],[28]. In India, participants from the general population brought to community internet centers preferred computer-based counselling [19], while in San Francisco, homeless men preferred face-to-face counselling [30]. In a study from Malawi, participants indicated post-test counselling was essential, but agreed that it did not have to be immediately available following self-testing for test negative participants [17]. Regarding mode of counselling, in a study conducted in student populations in Montreal, students' preferred counselling options that were through community clinics (78%, n = 78/100), by phone (53%, n = 53/100), at pharmacies (31%, n = 31/100), and by Internet (29%, n = 29/100) [29]. In Malawi, most participants (including repeat testers) preferred face-to-face counselling over phone-based counselling [33]. Data for preference for and mode of counselling for the unsupervised strategy was limited. In the US, a high preference (88%, n = 93/106) was noted for post-test counselling, and 96% (n = 102/106) of subjects newly diagnosed with HIV indicated they would follow up for positive self-test results with a doctor or a clinic [13]. Among the subjects who tested positive for HIV who received post-test counselling, 69% (absolute number not reported) were calm upon learning their status, while 31% (absolute number not reported) indicated some level of anxiety; however, no intervention (suicide helpline) was required for any of these subjects [13]. In a study from Kenya, telephone-based counselling from a call-center was deemed useful by HCPs [12]. Qualitative Data Qualitative data were sparse and reported in five studies [12],[16]–[19]. A mixed methods study in Kenyan HCPs incorporated qualitative data to assess feasibility, acceptability, and barriers and the influence of self-testing on couples testing choices [12]. Another US-based study questioned MSM on the use of self tests to screen potential sexual partners [16]. A study in urban India surveyed general population at community internet centers [19], while two other US-based studies elucidated social and structural barriers (challenges, concerns, and barriers) [17],[18]; these issues are discussed in detail below. Challenges Challenges in self-testing with a partner were discussed for both strategies: unsupervised strategy among HCPs in Kenya and supervised strategy in high-risk MSM in the US [12],[16]. Kenyan HCPs noted that the key challenge was avoiding the potential misuse of self tests, including non consensual testing [12]. In another study that evaluated a supervised strategy with non-monogamous MSM, refusal of new partners to self-test was flagged as a possible positive HIV status that would deter them from having sex or encourage them to use condoms [16]. Participants also indicated that the self test would “kill the mood” if used right before sex and being under the influence of drugs and/or alcohol could also dissuade testing altogether [16]. Concerns Concerns regarding both strategies included accuracy, stigma, misuse, and potential abuse of self tests. A lack of trust in the accuracy of self-test results was reported in two studies [29],[30], while in one US-based study, 94% (n = 407/433) of participants trusted their oral fluid results more than their finger-stick results (87%, n = 39/45) [24]. In supervised self-testing strategies, where part of the self-test distribution or assistance was offered by counsellors or provided at visible testing centers, perceived stigma as a barrier was reported by participants in two studies [18],[19]. HCPs in Kenya described potential misuse of self tests including non-consensual testing of partners and children/infants, maliciously infecting others if found positive, and unauthorized selling of the test kit [12]. Barriers Study participants for both strategies in both settings perceived barriers. In the supervised testing strategy, stigma and discrimination, and sequelae surrounding a diagnosis of HIV was high in Malawi, with 22% (n = 47/216) of participants fearing verbal abuse, 14% (n = 29/216) thinking they would be excluded by friends, and 10% (n = 11/110) of women and 11.3% (n = 12/106) of men fearing that their partner would leave them in case they tested positive for HIV [18]. In an unsupervised strategy evaluated in Kenya, a fear of visibility discouraged HCPs because they did not want to be seen carrying the self-test kits [12]. Quality of Studies Quality of studies varied. A lack of standardization in reporting and documentation of outcomes was observed. Qualitative data were sparse and evaluation of quality limited by incomplete reporting of data in abstracts. Combined reporting of quantitative and qualitative findings along with insufficient reporting of the qualitative methods and data collection tools (e.g., content of interview guide) may have masked potentially useful qualitative evidence. In a US study, the rationale for employment of different methodologies was clearly explained, and care was taken to reduce bias [16]. However, lack of clear presentation of themes in the results limited our understanding of collected data. Discussion We identified 21 studies that assessed supervised or unsupervised HIV self-testing strategies. The majority of the evidence from included studies was for supervised strategies from high-resource settings, although seven recent studies reported on an unsupervised strategy [12],[13],[20],[22],[23],[25],[26]. Included studies recorded a high score for acceptability (range: 74%–96%), preference over facility-based testing (range: 61%–91%), and partner self-testing (range: 80%–97%). Sensitivity and specificity for both unsupervised and supervised self-test strategies was high, but a lower sensitivity was reported for unsupervised (range: 92.9%–100%; one study) versus supervised (range: 97.4%–97.9%; three studies) strategies. Only one of the included studies reported on linkage to counselling and care where 96% (n = 102/106) of individuals who tested positive for HIV stated they would seek post-test counselling. While the evidence of high acceptability for supervised strategies is clear, it is not so for unsupervised strategies, especially in resource constrained settings. Only one study evaluated it, and the study focused on acceptability in a health literate population (HCPs) [12]. In terms of interpretation, self-testing has to be viewed as a process that requires a higher level of motivation and pro-activity compared to facility-based conventional testing. It involves a certain level of independence and assumes a basic level of literacy (typically grade 6 or high school) to ensure confidence in the self-test conduct and interpretation and enough personal involvement to seek post-test counselling and linkage to care and treatment. In addition to motivation, the conduct of a self test and its interpretation are key to its successful implementation. Additionally, demystifying the process of self-testing on television and social media will improve passive absorption, help de-stigmatize the HIV testing process, and promote the visibility of a self-testing strategy while providing information on HIV. Therefore, more research on exploring the best strategy (i.e., supervised versus unsupervised strategies) for different high- and low-risk populations in resource-constrained settings is clearly needed. Evidence for preference for self tests was evident for both supervised and unsupervised strategies, while mode and medium of preferences for counselling varied across settings, populations, and strategies. In terms of interpretation, qualitative research will help understand and tailor the preferred counselling option to high- and low-risk populations in diverse settings. Counselling strategies must be tailored to the literacy levels, lifestyle needs, and preferences, especially for populations living in resource constrained settings. For cost preferences, low-income populations in both high- and low-resource settings tended to prefer free self tests whereas higher income groups in both settings were willing to pay typically up to US$20. This finding implies that the price of a self-test kit will be an important factor in determining the uptake of self tests. Qualitative research will help guide the ranges of cost preference. This is crucial to policy initiatives for both strategies in both settings. Regarding findings on accuracy, while a consistently high specificity was noted for both strategies across all settings, a wide variability in sensitivity estimates was observed. Lower sensitivity was more prominent in studies that used the unsupervised self-testing strategy. Interestingly, lower estimates of sensitivity were also reported from high-prevalence settings, conducted by at-risk testers that were experienced with testing (including individuals who knew their sero-status) but generated more errors with the self test by not complying with instructions compared to individuals who did not know their sero-status. In contrast, naïve self-testers in low-risk settings seemed to follow instructions diligently resulting in high estimates of sensitivity. Although this process of passive absorption and priming to self-test instructions was present in supervised strategies in high-income settings where it resulted in fewer errors and highly accurate results, it was notably absent or weakly implemented in the unsupervised strategy. This finding highlights the importance of improving clarity of instructions to self-test. In terms of generalizability, adapting these instructions to contexts and populations with use of various media (pictorial and textual instructions, web, social media, or smartphone-based content in a user friendly format) and tailoring to languages and levels of literacy especially for low-resource settings will be essential. In terms of feasibility, none of the existing literature addressed issues related to seeking linkage to care (obtaining a CD4 count or having visited a clinic for ART initiation) and evidence from only one study reported that 96% (n = 102/106) of individuals who tested positive for HIV stated they would seek post-test counselling. Linkages could be better documented in the next phase of controlled studies. Furthermore, in planning roll out of self-testing strategies, clarification of some concepts and limitations of the self tests to potential self-testers is necessary. These include presence of faint positive lines and their interpretation as a positive test result, and the limitation of antibody-based self tests to detect an HIV infection only after 90 d. With either strategy, occurrence of false negatives or indeterminate test outcomes cannot be eliminated with antibody-based oral or finger-stick self tests so repeat testing at 3-mo intervals should always be encouraged. At any time, calling or meeting a counsellor for clarification of test results must be emphasized. For poor and less literate populations who cannot afford self tests or cannot comprehend the process of testing, the supervised self-testing strategy may remain the best option. But this option requires a careful controlled evaluation in many sub-Saharan and Asian settings. Engagement of providers and stakeholders is a necessary first step before implementation of a self-testing strategy because treatment and staging must be available for newly diagnosed individuals with HIV. Efforts must be made to avoid loss to follow up. In under regulated and under funded health systems, it will be important to regulate the quality of self-test kits along with the process of self-testing. In contrast, in well-managed health systems in high-income countries (e.g., US, Canada, UK, Australia, Singapore, Japan) and populations with high socio-economic status in emerging economies and middle-income countries (China, South Africa, Russia, India, Brazil, Colombia), affordable OTC self tests sold at pharmacies or online should work well with linkages operationalized by counsellors via different mediums, such as over the phone, the Internet, or face-to-face at a clinic. Evidence of extreme adverse effects, for example suicide, and the potential for abuse of self tests was consistently absent. However, as such evidence is collected in well-controlled studies, such data from a real world context will be hard to document. Anecdotal evidence suggests that unregulated rapid tests are available in markets of African and Asian countries and that such abuse will happen regardless of the introduction of self tests. Nevertheless, safeguards introduced by public health systems could help minimize the problem. To reach preliminary self-testers, a 24/7 toll free number to reach counsellors in time, and the offer of an option of face-to-face counselling could help obviate the occurrence. In terms of new research, controlled clinical trials in different populations and settings with exploration of tailored and novel linkage modalities are urgently needed. This will generate the best evidence on offering an optimal and tailored strategy suited to the needs of different populations and settings. The impact of self-testing on detecting new cases of HIV in the community and changes in sexual behaviour of self-testers post-testing are yet to be assessed. Because data on prior HIV testing behaviour for unsupervised strategies were undocumented, new knowledge on whether self-testing reached new testers is unclear and therefore needed. Data on the impact of self-testing in increasing community engagement in screening, and on long term HIV outcomes remain limited. These data can be documented through community-based cluster RCTs or implementation research. However, it is hard to conduct such trials in communities when self tests are not approved for use in many countries and especially in health systems where linkage to counselling is not in place. Lastly, formative research that further explores individual motivations in using self tests, their fears, perceptions, and concerns, is needed. It will assist in tailoring self-testing to the needs and preferences of individuals and communities. In sum, self-testing offers an alternative for individuals who desire privacy and confidentiality to find out their sero-status, as well as for under-resourced health care systems and settings where stigma and discrimination may prevail. It offers the potential to bring more people to self-screen and proactively seek linkage to care and prevention, but its potential in optimizing linkages remains unproven. However, if optimized, self-testing could facilitate and motivate individuals who preliminarily tested positive for HIV to seek care and counselling. Furthermore, care seeking should not be left to the motivated self-tester, but facilitated through various means, including technological innovations. But in any such system, if any component of self-testing, linked counselling and initiation of care are poorly managed, then the self-testing strategy could cause more harm than good. At all times therefore, linking positive self-test results within an 8–24-h window period to post-test counselling followed up by CD4 count and ART initiation should continue to be emphasized. Strengths and Limitations Approximately 70% of the included studies were cross-sectional in design. The majority of the data were primarily derived from implementation studies and from high-income settings (89%), only one of which was an RCT. Few studies were conducted in resource constrained countries, none of which were RCTs. Lack of data from resource constrained settings limited our comparisons of supervised versus unsupervised test strategies (one unsupervised study of HCPs in Kenya [12] and one supervised study of an urban population in Malawi [33]). Data on linkage to care was sparse with only one unsupervised self-testing study reporting linkage to care as an outcome. Selection and volunteer bias (self-selection of participants) within included studies cannot be ruled out. Publication and reporting bias are also likely. Errors in test interpretation were recorded in three studies. With these errors, a potential misclassification of false negative test results and misinterpretation of faint positive lines is probable; hence information bias (misclassification of test results) is also likely. The quality of reporting in included studies was generally poor. STARD and QUADAS checklists could not be used to assess quality because they are tailored for diagnostic accuracy, which was not the primary outcome in most studies. The use of observational study designs, data from pilot studies, with incomplete reporting of data items and lack of compliance with STROBE reporting criteria, were typical problems encountered in our quality assessment of the studies. A general lack of standardized reporting of outcomes beyond accuracy (patient centered outcome or implementation research outcomes) in the field of POC diagnostics has been noted [34]. It is therefore not surprising that these outcomes were heterogeneously and incompletely reported across many studies. The wide range of sources searched, the lack of language restrictions, and the evaluation of study quality are strengths of this review. Conclusion Privacy, anonymity, time-savings, and convenience facilitated a high acceptability and preference for both supervised and unsupervised self-testing strategies across diverse global settings. Included studies demonstrated that it was feasible to implement both supervised and unsupervised oral fluid-based self-testing strategies, despite variable accuracy estimates obtained by self-testers. However data on linkages to care were only reported by a single study of unsupervised self-testing. Self-testing can be an alternative option to facility-based testing for individuals who desire more privacy. However, controlled trials wherever possible, and implementation studies that document linkage to care, are warranted to confirm findings from observational studies. Although self-testing offers the potential to increase the number of individuals to self-screen for HIV and therefore deliver more people to care, systems that can maintain confidentiality and operationalize linkage to care within a reasonable time frame are pertinent to its success. More data from diverse settings are needed to inform global scale-up and policy recommendations for HIV self tests. Supporting Information Table S1 STROBE reporting criteria for cross-sectional studies (full-text). (DOCX) Click here for additional data file. Table S2 STROBE reporting criteria for cross-sectional studies (conference abstracts). (DOCX) Click here for additional data file. Table S3 STROBE reporting criteria for cohort study (full-text). (DOCX) Click here for additional data file. Table S4 CONSORT reporting criteria for RCT (conference abstract). (DOCX) Click here for additional data file. Text S1 PRISMA checklist. (DOC) Click here for additional data file.

            Author and article information

            [ a ] Department of HIV, World Health Organization , Geneva, Switzerland
            [ b ] Social and Behavioral Interventions Program, Department of International Health, Johns Hopkins University Bloomberg School of Public Health , Baltimore, MD, USA
            [ c ] Division of Global and Community Health, Department of Psychiatry and Behavioral Sciences, Medical University of South Carolina , Charleston, SC, USA
            [ d ] Independent Clinical Epidemiologist , Cape Town, South Africa
            [ e ] Department of Essential Medicines and Health Products, World Health Organization , Geneva, Switzerland
            Author notes
            [ § ]Corresponding author: Cheryl C Johnson, Department of HIV, World Health Organization , 20 Avenue Appia, Geneva 1201, Switzerland. Tel: +41 22 791 4335. ( Johnsonc@ )
            J Int AIDS Soc
            J Int AIDS Soc
            Journal of the International AIDS Society
            Taylor & Francis
            15 May 2017
            : 20
            : 1
            © 2017 Johnson CC et al; licensee International AIDS Society

            This is an Open Access article distributed under the terms of the Creative Commons Attribution 3.0 Unported (CC BY 3.0) License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

            Figures: 5, Tables: 3, References: 46, Pages: 11
            Funded by: World Health Organization 10.13039/100004423
            This systematic review was supported by UNITAID, the Bill and Melinda Gates Foundation [Grant Number: OPP1134510] and the United States Agency for International Development [Grant Number: GHA-G-00-09-00003] to inform the WHO guidelines on HIV self-testing.
            Review Article
            Review Article

            Infectious disease & Microbiology

            hiv test, hiv/aids, hiv self-test, public health


            Comment on this article