The effects of HIV self-testing on the uptake of HIV testing, linkage to antiretroviral treatment and social harms among adults in Africa: A systematic review and meta-analysis

Introduction HIV infection is still a global public health problem. More than 75% of HIV-infected people are in Africa, and up to 54% are unaware of their HIV status, limiting access to antiretroviral treatment. Context and purpose of the study This review aimed to determine whether HIV self-testing (HIVST) increases the uptake of testing, the yield of new HIV-positive diagnoses, and the linkage to antiretroviral treatment, and social harms among adults in Africa. Methods PubMed, The Cochrane Central Register of Controlled Trials (CENTRAL), Pan African Clinical Trials Registry, The Cochrane Database of Systematic Reviews (CDSR), Databases of Abstracts of Reviews of Effectiveness (DARE), Social Sciences Citation Index, Web of Science and African Index Medicus databases were searched from 1998 to 2019 (updated in December 2019). Eligible trials employed randomized controlled trials (RCTs), before/after studies, and interrupted time series design comparing HIVST to standard HIV testing services or comparing different approaches to HIVST among adults living in Africa were systematically sought. Results After searching 2,617 citations eleven trials were identified including 59,119 participants from four (4) African countries. Meta-analysis of seven trials showed a significant increase in the uptake of HIVST compared to standard HIV testing services: Both fixed-effects (Rate Ratio (RR) = 2.64, 95% CI: 2.51 to 2.79), and random-effects (RR) = 3.10, 95% CI: 1.80 to 5.37, and a significant increase in the uptake of couples’ HIVST (RR = 2.50, 95% CI: 2.29 to 2.73 in fixed-effects models; and RR = 2.64, 95% CI: 2.01 to 3.49 in random-effects model). A decrease in linkage to care and ART was observed in HIVST compared to standard HIV testing services (RR = 0.88, 95% CI: 0.88 to 0.95 in fixed-effects models; and RR = 0.78, 95% CI: 0. 56 to 1.08 in random-effects models). Six RCTs measured social harms, with a total of ten reported cases related to HIVST. One RCT comparing two approaches to HIVST showed that offering home-based HIVST with optional home-initiation of antiretroviral treatment increased the reporting of a positive HIV test result (RR: 1.86; 95% CI: 1.16 to 2.98), and linkage to antiretroviral treatment (RR: 2.94; 95% CI: 2.10 to 4.12), compared with facility-based linkage to antiretroviral treatment. Conclusions HIVST has the potential to increase the uptake of HIV testing compared to standard HIV testing services. Offering HIVST with optional home initiation of HIV care compared to HIVST with facility-based HIV care increases HIV positivity and linkage to antiretroviral treatment. Reported incidences of intimate partner violence related to HIVST were rare. Future research should focus on the potential of HIVST to reach first-time testers, the effect of using different approaches to HIVST, and strategies for linkage to HIV services. Systematic review registration This systematic review was prospectively registered on the Prospero International Prospective Register of Systematic Review (CRD42015023935).


Introduction
Globally, an estimated 35.0 million people are living with HIV, with more than 2.1 million [1.9 million-2.4 million] new HIV infections and 1.5 million deaths in 2013 [1]. Africa remains the most affected region, contributing more than two-thirds of the global burden of HIV [2]. The UNAIDS/WHO has set a "95-95-95" global target to be reached by 2030. The global target stipulates that 95% of adults will know their HIV status, 95% of HIV positive will receive sustained antiretroviral treatment (ART), and 95% of those who are on ART will achieve viral load suppression by 2030 [3].
In Africa, HIV testing coverage remains low with up to 54% of adults are unaware of their HIV status, despite substantial efforts to increase HIV testing [4]. The low coverage observed undermines the ability to reach the first "90" target and is a critical barrier to scaling up HIV prevention, care and treatment programmes. Also, late presentation for ART and delay harms treatment outcomes among HIV-positive people in Africa, leading to high mortality, and high transmission of the virus [5,6]. Existing literature shows that HIV self -testing (HIVST) has the potential to provide an opportunity for individuals who have never tested for HIV to know their HIV status [7]. According to WHO and UNAIDS, HIVST broadly refers to "a process whereby an individual, who is willing to know his/her HIV status collects a specimen, performs the test and interprets the test results in private" [8].
• CD4 count: the proportion of diagnosed HIV-positive participants who had their CD4 count measured [26].
• Social harm: the number of participants for whom any episode of harm was observed or reported, during or after HIV testing (e.g. intimate partner violence, coercive testing by a partner, or suicide, etc.) [27].
Inclusion and exclusion criteria. All experimental studies which included RCTs, before/ after studies and interrupted time-series studies, and compared HIVST with the standard of HTS and/or which compared different approaches to HIVST were eligible for inclusion. Editorials, reviews, perspectives, and studies not evaluating self-testing strategies (e.g. home-based but non-self-test) were excluded. Studies, which did not clearly define the type of HIV testing strategies, were also excluded. Abstracts were included if full-texts were not available. Any disagreements in study inclusion/exclusion were resolved by consensus during a meeting between reviewers.
Setting. We included studies conducted in any country on the African continent.

Information sources
Electronic databases. A comprehensive search strategy was developed to identify both published and unpublished articles with no language restrictions from January 1998 to December 2019. The start-time period search restriction was used because since 1998 there has been an emergence of the advanced development of rapid HIV diagnostic tests including self-testing [28]. The review searched for related studies in PubMed, The Cochrane Central Register of Controlled Trials (CENTRAL), Pan African Clinical Trials Registry, The Cochrane Database of Systematic Reviews (CDSR), Databases of Abstracts of Reviews of Effectiveness (DARE), Social Sciences Citation Index, Web of Science and African Index Medicus. Also, we searched websites and databases for grey materials (generally refers to no publicly published literature) such as greynet.org, World Health Organization Library Information System (WHOLIS), WHO Global Index Medicus, The Joint United Nations Programme on HIV/ AIDS (UNAIDS resource library), Alliance of Health Policy & Systems Research, The World Bank. We also searched for conference abstracts of the following conference databases: International AIDS Conference 1st International symposium on self-testing for HIV, International AIDS Society Conference (IAS), and Conference on Retroviruses and Opportunistic Infections (CROI). Terms for self-testing only were used to search for HIV-related conference abstracts published between 2001-2019 to overcome the limitations of search functions. We also searched the HIV Self-Testing Research and Policy Hub (HIVST.org) for ongoing updates of ongoing studies. The search strategies for electronic databases incorporated medical subject headings (MeSH), free-text terms and comprehensive African search filter that was adapted to suit each database using an applicable controlled vocabulary [29,30]. We also checked reference lists of included studies for other eligible reports. We also checked reference lists of the included studies and other relevant systematic reviews for further eligible reports using Google Scholar, and Web of Science. Before the publication of this review, we conducted an updated search for HIVST studies in Africa that were published from 1 st January 1998 to 31 st December 2019.
Search strategy. This systematic review used a search strategy described in a previously published protocol and summarized in Additional file 1 [21].
The search was restricted to human subjects. The full search string for PubMed is included as supplementary material S1 File. We sought expert views on the search strategy to assess the search terms proposed for this review.

Study records
Data management. All search results were merged into reference management software EndNote and were all imported into Convidence (https://www.Convidence.org/), and BN and DD performed an automatic check to exclude duplicate entries.
Selection process. Two reviewers (BN and DD) independently screened the titles to exclude all irrelevant studies and select potentially eligible studies. In this process of abstract screening, the two reviewers included, excluded or classified abstracts as 'maybe'. The reviewers met to discuss the disagreements, with a consensus to include or exclude. BN then obtained the full text of potentially eligible studies and later BN and DD independently conducted the final study selection for inclusion in the review. The two reviewers provided the reasons for excluding studies at this stage. The final included studies for data collection and the screening process created by Convidence are presented as part of the PRISMA study flow chart shown in Fig 1. Data collection process. Two reviewers, independently extracted data using a predesigned data extraction form. A third reviewer (LA) was consulted to resolve any differences of opinion between the two reviewers that arose. We conducted a pilot trial of the data extraction form to check its adequacy and made necessary changes.
Data items. Extracted data included: study design, setting of the study (e.g. city/ country/ or rural/urban or facility-based /community-based), year of study, year of publication-1998 to PLOS ONE date, type of HIV self-testing (e.g. supervised or unsupervised), type of comparator (e.g. standard HIV testing services), study population (e.g. general population or key populations), and study outcomes of interest.
Assessment of risk of bias in included studies. Two authors (BN and DD) independently assessed the risk of bias (low risk of bias, high risk of bias, or unclear risk of bias) using the Cochrane risk of bias instrument, including an evaluation of random sequence generation and allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and unclear risk of bias (either due to lack of information or uncertainty over the potential bias). We resolved disagreements by consensus. Trials reporting > 20% of missing data were judged at a high risk of attrition bias. In HIV testing studies, it is difficult to bind participants, and/or researchers to the interventions resulting in a high risk of performance bias been made. However this should not infer that the methodological quality of the trials was poor, but the reviewer's acknowledging the inevitable bias related to lack of blinding [31]. The assessments of risk of bias are summarized in Tables 1, 3 and 4 and Fig 6. Investigation of heterogeneity. The I 2 statistics were used to evaluate the heterogeneity of results for each outcome, across trials. We interpreted values of 25, 50, and 75% as low, medium, and high heterogeneity respectively. The observed I 2 statistics relies on the magnitude and direction of effects, and the strength of evidence for heterogeneity. Although not a measure of absolute heterogeneity, the I 2 statistics describe the percentage of variability in the point estimates that is due to heterogeneity rather than sampling error [32].
Assessment of reporting biases. The review employed strategies to search and include relevant unpublished studies to reduce publication bias. The strategies included searching the gray literature, including conference proceedings (e.g., International AIDS Conference), and prospective trial registration database to over-come lime-lag bias. To investigate publication bias, if there were � 10 trials included in an analysis, we planned to use a funnel plot to explore the possibility of small study effects-a tendency for smaller studies to report larger beneficial effects. However, this review did not use a funnel plot test because of fewer trials (< 10 trials) were included in an analysis, as the power of tests is too low to distinguish chance from publication bias [33].
Data analysis. Where data from � 2 trials were available, pooled intervention effect estimates are expressed as risk ratios with corresponding 95% confidence intervals (CI) for each outcome included in the study. The review used the fixed effects models and random-effects models for analysis in the Cochrane Review Manager. Pooled estimates were calculated using randomeffects models were high heterogeneity was found among studies. The random-effects model approximates the extent of the heterogeneity and allocates a greater variability to the estimate of overall effect to account for the heterogeneity. In the result section we present forest plots of random-effects models, and forest plots of fixed effects models as Supporting file S2 Fig [33].
Sensitivity analysis. Sensitivity analyses were performed on outcomes where an effect of excluding outlier studies and those with high overall bias risk might affect the overall estimate. Visual inspection of forest plots was used to identify trials, which were outliers, and the I 2 was re-evaluated without the outliers. We also investigated the influence of a single study on the overall estimate by excluding 1 study in each turn [32].

Description of included studies
Search results. A total of 2,617 citations were identified through electronic search and other sources, which after screening resulted in eleven trials met the review eligibility criteria The study reported high retention rate for women participants (arm 1= 86%; arm 3 = 84%). Male partner follow-up was also high (arm 1=80%; arm 3=84%). Intent-to-treat analysis was done at 3-months.

Selective reporting:
The study reported findings for both primary and secondary outcomes.
Other potential source of bias: None noted (Continued ) The randomization list was computer generated in random blocks of size 3,6, and 9 and stratified by 3 study sites to ensure balance in study arms by site.

Allocation concealment:
Participants were sequentially allocated into one of the three study arms using sealed opaque, sequentially numbered envelopes.

Blinding of participants and personnel:
Due to the nature of the intervention, masking was not possible; however the peer educator's study arm assignment was concealed until all participants had been enrolled.

Blinding of outcome assessment:
Due to the nature of the intervention, masking was not done.

Incomplete outcome data:
The study reported high retention rates for FSWs. The randomization list was computer generated in random blocks of size 3,6, and 9 and stratified by 5 administrative divisions and evenly recruited peers to ensure balance in study arms by site.

Allocation concealment:
Participants were sequentially allocated into one of the three study arms using opaque, sequentially numbered envelopes.

Blinding of participants and personnel:
Due to the nature of the intervention, masking was not possible; however study personnel, peer educators, and participants were masked to study arm assignment prior to opening the sealed envelopes.

Blinding of outcome assessment:
Due to the nature of the intervention, masking was not done.

Incomplete outcome data:
The study reported high retention rate for women participants.

Sequence generation:
Participants were randomized in a 1:1 ratio stratified on clinic.

Allocation concealment:
Participants were assigned to either the SOC arm or the Choice arm by fieldworker from a sealed envelope with the randomization assignment.

Blinding of participants and personnel:
The study did not describe the blinding of participants. Field workers were not blinded after randomization.

Blinding of outcome assessment:
The study did not describe the blinding of outcome assessment.

Incomplete outcome data:
The study used intent-to-treat analysis. The study reported high retention rate for both SOC arm and the Choice arm. Attrition rate at 6-month follow-up was: 8.4% SOC arm, and 5.3% the Choice arm.

Selective reporting:
The study reported results of all study outcomes.
Other potential source of bias: None detected (Continued ) No adverse events were reported in the HIVST arm.

Sequence generation:
Participants were randomized in a 1:1:1 ratio using computergenerated random numbers and using a balanced block. 15 clusters were blocked by facility type, ownership, and region and allocated to either the SOC, OSOC, or HIVST arm. Allocation concealment: The study protocol did not describe the allocation concealment process.

Blinding of participants and personnel:
The study did not blind participants or personnel.

Blinding of outcome assessment:
This was a non-blinded study.

Incomplete outcome data:
The study used intent-to-treat analysis. Data on retention rate of participants on follow-up are not available.

Selective reporting:
Only preliminary results for HIV testing and HIV positivity are presented.
Other potential source of bias: None detected (Continued ) Adjusted for baseline (preintervention) ART initiation.

Sequence generation:
22 rural PHC were randomized 1:1 to receive the HIVST intervention or SOC.
In the HIVST intervention arm, CBDAs were randomized 1:1 to provide either referral to homebased ART or the usual referrals to the clinic. CBDAs were randomized in blocks of varying sizes from 2 to 12 CBDAs. Allocation concealment: The study protocol did not describe the allocation concealment process.

Blinding of participants and personnel:
Given the nature of the study the participants and personnel were not blinded to the allocation status.

Blinding of outcome assessment:
Trial statisticians were un-blinded after completion of data collection.

Incomplete outcome data:
Data on retention rate of participants on follow-up are not available.

Selective reporting:
The study presents preliminary results for HIV testing (i.e., in last 12 months and lifetime), and ART initiations.
Other potential source of bias: None detected (Continued )

Sequence generation:
In the HPTN 071 trialrandomization was done in public ceremonies simultaneously in Zambia and South Africa. The 21 clusters were initially divided into 7 matched triplets. Allocation concealment: 21 study clusters (7 clusters in each study arm) in the HPTN 071 trial were assigned to 3-study arms using restricted randomization process to ensure overall balance across study arms on cluster size, current ART uptake, and HIV prevalence.

Blinding of participants and personnel:
Unclear risk of bias (blinding of participants and personnel not described in the abstract).

Blinding of outcome assessment:
Unclear risk of bias (blinding of outcome assessors not described in the abstract).

Incomplete outcome data:
The study used 'time-to-event" analysis and follow-up information.
At 3 months after referral: 64% SOC vs. 65% HIVST were LTC Selective reporting: The study presented preliminary results on HIV positivity and linkage to care. Other potential source of bias: None detected (Continued )

Sequence generation:
Clinic day clusters were randomized in either SOC arm or HIVST arm using generated random numbers each day of recruitment. All women who were recruited from a particular clinic on a particular day were randomized to the same study arm Allocation concealment: Participants were sequentially allocated into one of the two study arms using separate lists of random numbers and sampling frames each day.

Blinding of participants and personnel:
Participants and personnel were blinded.

Blinding of outcome assessment:
The study did not describe on blinding of outcome assessors.

Incomplete outcome data:
All analyses were based on intention-to-treat. No severe adverse events were reported.

Sequence generation:
Clinic days where randomised in blocks of different sizes to any of the six trial arms in stage 1 of the trial. All three-clinic days available in one day where pre-identified before the randomization such that the arm recruiting at a particular clinic on particular day will be known within randomization sequence.

Allocation concealment:
The ANC clinic days were sequentially allocated using computer-generated random numbers by independent statistician.

Blinding of participants and personnel:
Participants and personnel were blinded.

Blinding of outcome assessment:
The study did not describe the blinding of outcome assessors.

Incomplete outcome data:
All analyses were based on intention-to-treat. The denominator was the number of women who were eligible and take into account the clustered design.
The study reported high retention rate for both SOC and "ST" arm. Attrition rate: At 4 weeks follow-up post women enrollment: 19.57% SOC arm, 17.6% "ST" arm.
Other potential source of bias: None detected (Continued )

Sequence generation:
Reported that 14 clusters were randomized in a 1:1 ratio( 7 clusters into intervention group; 7 clusters into control group).

Allocation concealment:
Used colored balls from an opaque bag held above eye level to select the distribution of clusters and allocation.

Blinding of participants and personnel:
Reported blinding of investigators but not the counselors, or study participants, or outcome assessors.

Blinding of outcome assessment:
Due to the nature of the study masking was not done.

Incomplete outcome data:
All participants completed the trial, attrition rate was: 28.7 % in IG, and 23.8 % in CG from ART, no treatment withdrawals, no trial group changes, and no major adverse events reported.

Selective reporting:
The study reported results of all outcomes (1 primary; 4 secondary).
Other potential source of bias: None detected.  Table 2. Characteristics of excluded studies (n = 41).

Study Reason for exclusion
Wawer et al. 1998 [68] Intervention ineligible and comparison. STD treatment was used as an intervention, and anthelminthic and iron-folate was used as control in this RCT.
Leon et al. 2010 [69] Ineligible intervention. Provider-initiated (opt-out) HIV counseling and testing was used in this controlled trial.
Luganda et al. 2010 [70] Ineligible intervention and comparison. Home-based and clinics-based HIV counseling and testing were used as an intervention and comparison in this RCT.
Choko et al. 2011 [46] A cross-sectional feasibility study to determine the uptake and accuracy of oral kits for HIV self-testing in high HIV prevalence setting in Blantyre, Malawi.
Sweat et al. 2011 [72] Ineligible intervention and comparison. Home-based and clinics-based HIV counseling and testing were used as an intervention and comparison in this RCT. A study conducted among HIV stakeholders to explore public opinion and perspectives on HIV self-testing in Nigeria.

Choko et al. 2015 [27]
A community-based prospective study to assess the uptake, accuracy, safety, and linkage into care over two years of promoting annual self-testing for HIV in Blantyre, Malawi.

Kurth et al. 2016 [52]
A cross-sectional study to assess the accuracy and acceptability of oral fluid HIV self-testing in a general adult population in Kenya.

Mavedzenge et al. 2016 [53]
A cross-sectional study to determine acceptability, feasibility, and preference of HIV self-testing in Zimbabwe.
Thirumurthy et al. 2016 [54] A cohort study to assess an approach of providing multiple self-tests to women at high risk of HIV acquisition to promote partner HIV testing and to facilitate safer sexual decision-making in Kenya.

Perez et al. 2016 [55]
A cross-sectional study on supervised oral HIV Self-testing in accurate in rural Kwazulu, Natal, South Africa.

Mugo et al. 2017 [56]
A feasibility study to determine uptake and acceptability of oral HIV self-testing among community pharmacy clients in Kenya.   5 Risk of bias: Downgraded the quality of evidence by 1 level for risk of bias. Blinding of outcome assessment was judged "unclear", and the risk of bias cannot be excluded. 6 Inconsistency: Downgrading for inconsistency was not done, despite observed heterogeneity. However, the estimates of effects were all in the same direction. 7 Risk of bias: Downgraded the quality of evidence by 1 level because the outcomes were self-reports and performance could not be excluded. 8 Imprecision: Downgraded the quality of evidence by 1 level because of the few numbers of events in both groups, and overlapping of confidence interval around the estimate of effect.
Ten of the trials compared HIVST to standard HTS [34][35][36][37][40][41][42][43][44][45], and one trial compared the optional home-based confirmatory testing and initiation of antiretroviral therapy (ART) to The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. 1 Single study.

HIVST compared with standard HIV testing services
Uptake of HIV testing. Three trials in Kenya, Malawi, and Uganda [35,40,42], recruited women (n = 5450), with no risk of intimate partner violence, and with their male partners of unknown HIV status or known HIV-negative status, at antenatal and postnatal care and invited them to either distribute HIV self-tests to their male partners (intervention arm) or an invitation card and referral vouchers for free HIV testing, here assumed to be standard HIV testing services (comparison arm). Two trials conducted in Zambia and Uganda [36,39] recruited adult female sex workers (n = 1,925), who reported the exchange of any vaginal, anal, or oral sex for money, goods, or other items of value and invited them to either direct delivery of HIV self-tests, or coupons for free collection of HIV self-tests (intervention arms) or referral for standard HIV testing services (comparison arm). One trial in Kenya [43] recruited male truck drivers (n = 305) and offered a self-administered HIVST under the supervision/ or unsupervised self-administered HIVST at home (intervention arms) or routine PITC HIV testing, here assumed to be standard HIV testing services (comparison arm). One trial in Malawi [37] recruited adult patients attending the out-patient department (n = 3918) and offered selfadministered HIVST (intervention arms), or routine PITC HIV testing, here assumed to be standard HIV testing services (comparison arm). A meta-analysis showed an increase in the uptake of HIV testing in the HIVST arm compared with the standard HTS arm (RR = 3.10, 95% CI: 1.80 to 5.37; Tau 2 = 0.52; Chi 2 = 808.64; df = 6 (p < 0.00001; I 2 = 99% in randomeffects models)). The I 2 of 99% is the proportion of the variation in observed effects due to variation of the true effects (Fig 2). A meta-analysis including three trials [34,35,40], among pregnant women (n = 8,397), showed an increase in the uptake of couples HIV testing in the HIVST arm compared to standard HTS (RR = 2.50, 95% CI: 2.17 to 2.89; Tau 2 = 0.01; Chi 2 = 5.08; df = 1 (P = 0.08; I 2 = 61% in random-effects models)). Couples HIV testing referred to the participant and her male partner testing for HIV together at the same time. The I 2 of 61% is the proportion of the variation in observed effects due to the variation of the true effects (Fig 3).
HIV positivity. Seven trials [35-37, 40-42, 45], assessed HIV positivity among male partners of pregnant women (n = 3), female sex workers following HIV testing (n = 2), adult patients attending health facilities (n = 1), and adults in the general population (n = 1). HIV positivity rate was determined as the proportion of participants diagnosed HIV-positive over those who accepted HIV testing. A meta-analysis showed no statistically significant difference in reporting HIV positivity among male partners of pregnant women, and female sex workers from HIVST compared to standard HTS (RR = 1.00, 95% CI: 0.87 to 1.15; Tau 2 = 0.00; Chi 2 = 5.95; df = 6 (p < 0.43; I 2 = 0% in random-effects models)). The I 2 of 0% is the proportion of the variation in observed effects due to variation of the true effects (Fig 4).
Linkage to care and antiretroviral treatment. Six trials, [35,36,[40][41][42]45], reported on linkage to care and ART among participants who were diagnosed HIV positive (n =). A metaanalysis showed that participants from HIVST are less likely to seek HIV-related care and initiating ART compared to participants from HTS (RR = 0.78, 95% CI: 0.56 to 1.08; Tau 2 = 0.10; Chi 2 = 24.54; df = 5 (p < 0.0002; I 2 = 80% in random-effects models)). The I 2 of 80% is the proportion of the variation in observed effects due to variation of the true effects (Fig 5).
Social harms. Six trials assessed social harms, following HIVST or standard HTS among adult participants. The trial in Kenya [35] reported a single incidence of harm in each group between two HIV-negative participants in the HIVST group, and the control group. In both groups, the reported harm was verbal and/ or physical intimate partner violence. The trial in Zambia [36] reported four incidences of harm, with three participants in the direct delivery of HIVST intervention arm, and one participant in the coupon intervention arm, reporting physical violence following primary partner learning of their HIVST use. None incidence of harm was reported in the standard of care arm. The trial in Uganda [39], reported four incidences of harm, with one participant in the direct delivery of HIVST intervention arm, and one participant in the coupon intervention arm, reporting verbal abuse following primary partner learning of their HIVST use/or mental distress following a positive HIV self-test result. Two participants in the standard of care arm reported verbal abuse after the disclosure of their HIV status. Two trials among male partners of pregnant women and patients attending health facilities in Malawi reported no social harms after HIVST [37,42]. Another trial in Malawi among the the general population of rural communities reported three incidences of social harm [44].
Sensitivity analysis. Exclusion of 5 studies from a meta-analysis reporting uptake of HIV testing in HIVST arm compared with the standard of HTS arm altered the overall estimate and heterogeneity (RR = 1.20, 95% CI: 1.12-1.27; P = 0.99, I 2 = 0%). Exclusion of studies from meta-analyses reporting uptake of couples HIV testing, HIV positivity, and linkage to care and ART did not alter the existing heterogeneity.

HIVST with optional home-initiation of HIV care compared to HIVST with facility-based HIV care
One RCT in Malawi (n = 16,660) compared different approaches to HIVST [25].
HIV positivity. The trial reported HIV positivity among participants following HIV testing. HIV positivity rate was determined as the proportion of participants diagnosed HIV-positive over those who accepted HIV testing. There was a significant difference in reporting a positive test result between the home and facility groups (RR: 1.86; 95% CI: 1.16 to 2.98; P = .010) to HIVST accompanied with referral to facility-based HIV care services.
Tested for CD4 counts. The study compared the receipt of CD4 count test results among newly diagnosed HIV-positive individuals. In the arm with the offer of optional home-based confirmatory testing and initiation of ART group, 72. 5% (n = 79/109) of newly diagnosed HIV positive individuals received CD4 count results compared to 51% (n = 23/63) in the facility group (RR = 0.70, 95% CI: 0.54 to 0.91; P = .007).
Linkage to care and antiretroviral treatment. The trial reported ART initiation and retention in care. In the arm with the offer of optional home-based confirmatory testing and initiation of ART group, initiation of ART was 2.2% (n = 181/8,194) compared to 0.7% (n = 63/8,466) in the control arm (RR, 2.94; 95% CI: 2.10 to 4.12; P < .0001). After adjusting for reported household mortality at baseline, the effect of availability of home ART care was (ARR, 2.44; 95% CI: 1.61 to 3.68; P < .001).
Assessment of risk of bias in included studies. The risk of bias in the included studies was assessed as described above. The assessments of risk of bias in included studies are presented in the risk of bias graph below (Fig 6), while additional details are included in Tables 1, 3

and 4.
Assessment of overall quality of the evidence. We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to assess the quality of evidence and generate absolute estimates of effect for the outcomes as described in the GRADE profiler (GRADEpro) software [31,83]. The GRADE methodology defines the quality of evidence for each outcome as ' the extent to which one can be confident that an estimate of effect or association is close to the quantity of specific interest" [20]. The quality rating across studies has four levels: high, moderate, low or very low. RCTs are initially categorized as providing high-quality evidence with the option of downgrading the quality. Quality of evidence can be downgraded based on limitations in design, indirectness of evidence, unexplained heterogeneity or inconsistency of results, imprecision of results or high probability of publication bias [20]. We did not downgrade based on lack of blinding alone due to difficulties of blinding participants, and or research personnel in HIV testing intervention trials. Downgrading was done based on the risk of bias when a lack of blinding was accompanied by additional high risks of bias (e.g., selection bias and incomplete outcome reporting). We summarized the quality of evidence for the studies separately for each outcome in the GRADE Summary of Findings Tables [31].
The overall quality of the evidence for the cluster RCT [25] was moderate. The quality of evidence was downgraded because of the risk of bias (Table 4).

Summary of main results
The objective of this systematic review was to synthesize the evidence on the effects of HIVST among adults in Africa on the uptake of testing, the yield of new HIV-positive diagnoses,

PLOS ONE
linkage to ARV treatment, and the incidence of social harms. This review included eleven trials conducted in Malawi, Kenya, Zambia, and Uganda.
Two models of HIV testing were assessed in the included studies of this review. The first model included ten trials comparing HIVST versus standard HTS, and the second model included one trial comparing different approaches to HIVST. In the first model, 10 RCTs in Kenya, Zambia, and Uganda found a positive effect on the uptake of HIVST compared to standard HTS. Similarly, in three trials in Kenya and Uganda reported a positive effect on the uptake of couples' HIVST compared to standard HTS.
Seven trials reported no statistically significant difference in reporting HIV positivity among participants from HIVST compared to standard HTS. Also, six trials reported no statistically significant difference in linkage to care and ART among HIV positive participants from standard HTS compared with those from HIVST. Finally, ten incidences of any kind of social harm were reported following HIVST and standard HTS from 6 RCTs in Kenya, Zambia, and Uganda.
In the second model, one trial in Malawi reported a statistically significant difference, showing participants who received HIVST and an optional home-based confirmatory testing and initiation of ART were more likely to report HIV positive test results than those in the control group. Further, the trial showed an increased proportion of newly diagnosed HIV positive participants in the homegroup received CD 4 count tests than those in the control group. Finally, the trial reported a statistically significant difference showing that participants who reported HIV-positive test results from home-based HIVST with optional home-based confirmatory testing and initiation of ART were more likely to initiate ART care compared with their counterparts from the home-based HIVST with facility-based HIV care.

Summary of evidence: HIVST versus standard of HIV testing services
Uptake of HIVST. Seven trials were included in the meta-analysis, and showed a moderate quality of evidence, for the uptake of HIV testing compared to standard HTS. Furthermore, two RCTs were included in the meta-analysis also showed the moderate quality of evidence for uptake of couple HIV testing. Although the results of the meta-analysis report that HIVST can double the uptake of HIV testing compared to standard HTS, there is some uncertainty because of heterogeneity in effect size between the studies. Besides, the evidence was limited to male partners of pregnant and post-partum women and adult female sex workers.
HIV positivity. Seven trials included in this review with moderate evidence showed no difference between HIVST compared to standard HTS in reporting HIV-positivity. The quality of evidence was downgraded for the risk of bias, detection bias, selection bias, and imprecision. None of the trials reported the yield of new HIV positive diagnoses. Yield and HIV positivity are important for designing appropriate and sustainable HIV testing interventions because it determines the number of tests needed to diagnose a new case of HIV, and has cost and resource implications [84]. However, all trials included in this review, none reported on the yield of new HIV positive diagnoses, underscoring the need to assess this important indicator in future studies, particularly in high HIV prevalence settings.
Linkage to HIV care. Six trials included in this review with moderate evidence showed poor linkage to HIV care following confirmatory testing between HIV positive participants in the HIVST group compared to those in the standard HTS group.
Social harms. One trial from Kenya compared the distribution of HIVST kits by pregnant women to their male partners with standard HTS, and 2 trials from Zambia and Uganda, compared the delivery of HIVST kits with standard HTS among adult female sex workers and assessed IPV resulting from HIVST. The trial reported a total of 10 incidences of any form of social harm related to HIV self-testing reported by different adult populations, such as pregnant women, and female sex workers. It is possible that self-testing does alter IPV experiences, but is currently not accurately measured by the extant literature. This finding of reported rare cases of IPV related to HIVST is consistent with several studies assessing harm across all forms of HIV testing services [27,85,86]. However, it is important to introduce HIVST with caution among vulnerable populations (e.g. female sex workers (FSW)) [8,54].

Summary of evidence: Comparison of different approaches to HIVST
HIV positivity. One trial from Malawi with moderate-quality evidence, assessed different approaches to HIVST and, observed that participants from home-based HIVST with optional home-based confirmatory testing and initiation of ART were significantly more likely to report a positive result than participants from home-based HIVST with facility-based HIV care [25].
Tested for CD 4 count. The same trial showed a high proportion of newly diagnosed HIV-positive participants, from the intervention arm compared to participants form the control arm tested for CD4 counts [25]. The finding supports the evidence that a high proportion of CD4 counts could be achieved outside the health facilities when offered together with HIV test results, and may facilitate immediate treatment for improved health and to decrease the probability of further HIV transmissions [18,27].
Started ART. Additionally, the same trial in Malawi reported an increased proportion of participants who reported HIV-positive results from the intervention arm compared to the control arm. However, this RCT reported the loss to follow-up 6 months after the home-initiation of ART compared with the facility-initiation of ART. ART initiators in the homegroup had higher rates of loss from ART compared with their counterparts in the facility group.
Overall completeness of the evidence and applicability of evidence. The observed results on increased uptake concur with results from other systematic reviews [16-18, 87, 88] and observational studies from Kenya [56], Lesotho [89], Malawi [27,46], and Zimbabwe [90], which reported similar increases in the uptake of HIV testing through HIVST. For instance, the systematic reviews reported an uptake of HIVST ranging from 50% to 96% in other parts of the world [16][17][18] and 20% to 74% in the key populations [18,88]. Also, a two-year prospective study in Malawi reported s consistently high uptake of community-based HIVST among adolescents aged 16-19 years of age; young people aged 16-29 years of age, women, and men [27].
Although HIVST primarily focuses on an individual, two trials in this review were conducted in Kenya where women distributed HIVST kits to their male partners. The trials also reported an increased uptake of the couple's testing from HIVST compared with the approach of giving male partners invitation letters or vouchers for free HTS. These results suggest that men may be more likely to prefer HIVST than standard HTS [91]. This observation underscores the importance of getting men to utilize HTS because it is well documented that men compared with women are known to underutilize HTS and present late for care ending up with worse outcomes on treatment [5,6].
Observational studies among key populations, including men who have sex with men (MSM) [88], people who inject drugs (PWIDs) [92], and female sex workers [93], revealed high uptakes of HIV testing following the offer of HIVST. It is well documented that to achieve universal knowledge of HIV status, it is imperative to have innovative interventions such as HIVST to increase access to HIV testing. This review confirms that HIVST is an important strategy to improve awareness of HIV status among adults in SSA, and provides additional HIV testing options to other approaches such as VCT, PITC, school-based, and work-place testing [8,92].
Earlier HIV diagnosis supports timely access to ART, with several beneficial consequences, such as improved life expectancy, reduced HIV transmission, decreased stigma related to HIV testing, and the provision of HIV prevention interventions [87]. These results are consistent with results from observational studies among the general populations in SSA, which reported HIV-positivity ranging from 3 to 14% [27,53,54,90], and from 1 to 30% among key populations [54,93]. As coverage of HIV testing increases, the proportion of HIV-positive tests and new HIV diagnoses will be likely to decrease, and hence calls for more focused HIV testing methods to continue to achieve the same or higher levels of HIV-positivity [8,92].
This review identified gaps in the evidence on linkage to care and ART. Linkages to further HIV testing and HIV prevention, treatment, and care services are a critical component in the HIV testing cascade. This is more important in HIVST because it requires linkage to confirmatory testing, particularly among those who report a positive HIV test result [87]. The most probable explanation to the observed low level of linkage could be due to a few diagnosed HIV-positives, under-reporting and the possibility that some men were aware of their HIVpositive status and already in care. These findings suggest that innovative follow-up strategies post HIVST such as the use of mobile phones or short message services, or facilitated HIV care assessments and ART initiation interventions could be put into place to encourage linkage [8,27,87,92].
HIV testing experts recognize the importance and complexity of monitoring, reporting, evaluating and mitigating social harms related to HIV testing. The rare cases of social harm related to HIVST reported in this review, help to alleviate a major concern of HTS experts, regarding IPV related to HIVST. The findings may suggest that HIVST does not directly influence the risk of IPV, but the risks are largely context-specific, including the settings (e.g. urban/rural) and the relationship dynamics of couples and partners. To emphasize this critical concern, the WHO recommends that HIV testing programmes consider context-specific strategies to implement HIV testing approaches, including HIVST [8,91]. Furthermore, fear of status disclosure or stigma, the possibility of false-positive diagnoses, lack of confirmatory HIV testing, and insufficient quality control procedures highlight the need to address legal and human rights issues related to HIVST [8,85,87].
Quality of the evidence. According to the GRADE system, well-conducted RCTs provide high-quality evidence, while observational studies provide low-quality evidence. This review included only trial data, and we found the quality of evidence reported from trials for the HIVST versus standard HTS was generally moderate. The model for comparison of different approaches for HIVST, including one high-quality trial provided moderate quality of evidence.
Potential biases in the review process. We used a broad search strategy to capture as many studies as possible. We limited our search to those studies conducted in Africa from 1998 to date to improve comparability. A random-effects model was used to pool data. Despite using a broad search strategy, we were able to capture eleven trials (8 full test studies, and 3 abstracts) conducted across four countries in SSA. Further, there are limited studies from other parts of Africa, which may limit the generalizability of the review findings.
We, therefore, cannot exclude the possibility that of the dispersion presented, what proportion is due to variance in true effects rather than sampling error [32]. Publication bias was not formally assessed because of the limited number of trials for each outcome, since analytical methods such as funnel plots and funnel plots asymmetry tests, were not appropriate [20].
Agreements and disagreements with other studies or reviews. To the authors' knowledge, this is the first systematic review evaluating the effects of HIVST on the uptake of HIV testing, linkage to care and ART, and social harms among adults in Africa. Previous reviews focused on HIVST strategies, the acceptability of HIVST, and different community-based HIV testing approaches globally. The findings of this review provide important information on the potential of HIVST as an option for HTS among adults in Africa. A key finding of our review is that the uptake of HIVST is promising among different adult populations in both rural and urban settings in Africa, particularly in SSA. Importantly, it is the observation of high participation by male partners in couples' HIV testing through HIVST. However, more data are needed on the yield of new HIV positive diagnoses, particularly from countries with high HIV prevalence in SSA in terms of the diverse range of context and settings.
Further research among key populations such as MSM, PWIDs, and vulnerable populations such as adolescents, and factors associated with their participation in HIVST, is important to inform policy and practice. In addition, further key areas for research includes: the effectiveness of HIVST in detecting previously undiagnosed HIV infection, or number of repeat nontesters, linkage to care following a reactive self HIV test result or a confirmatory positive test, retention of care among those identified HIV-positive, linkage to prevention services among participants with negative results (e.g. male circumcision), and social harms from HIVST, which are important to help guide policy on HIVST in Africa. Interventions to facilitate a timely linkage to care are ongoing in Zimbabwe (PACTR20160700171788), and Malawi (ISRCTN18421340), with preliminary findings reporting significant benefits on linkage to VMMC and ART using financial and non-financial incentives.
Implication for research. This review has reported moderate-quality evidence for the uptake of HIVST, suggesting that self-testing has the potential to provide an innovative strategy to increase uptake of HIV testing and increase awareness of HIV status among undiagnosed adults in Africa, particularly in SSA. However, we also reported no statistically significant difference in reporting HIV positivity between participants from HIVST and HTS, indicating a need for further research particularly in high HIV prevalence settings in SSA. This review recommends that high-quality evidence from trials would provide valuable insight into whether HIVST as an additional HIV testing option could facilitate early detection, early linkage to HIV care, treatment, and prevention. Additional trial findings will provide data on HIVST as an invaluable tool for the health authorities of African governments to increase access to HIV care, treatment, and prevention to achieve the "95-95-95" global target by the year 2030.