Introduction

In 2021, more than 38.4 million people were living with HIV globally, and sub-Saharan Africa (SSA) has the highest burden of HIV [1]. People living with HIV (PLWH) are linked to care to ensure viral suppression [2]. Usually, the HIV continuum includes HIV diagnosis, linkage to antiretroviral therapy (ART), and retention in care [3]. An important step in the HIV continuum is HIV testing. Provider-initiated testing in health facilities or clinics has been advocated for over the years, but 15% of PLWH globally are still unaware of their seropositivity status as of 2021 [4].

HIV testing remains suboptimal even in countries where HIV is endemic, such as South Africa. For instance, up to 75.1% (range: 51.4–86.1%) of individuals aged 15 years and above interviewed across 52 districts in South Africa had ever tested for HIV [5]. Only 20.3% of young people surveyed in the United States had ever tested for HIV [6]; 20.4% of 770 men who have sex with men (MSM) in the Netherlands had ever tested for HIV [7]. A recent meta-analysis reported a 55.30% (95% CI 54.97–55.69%) pooled prevalence of HIV testing among young women in Eastern Africa [8].

Individual barriers, such as low level of education, low income, poor HIV knowledge, perceived low risks of HIV, and privacy concern, are some reasons for not testing for HIV [7, 9]. Also, structural barriers, such as distance to hospital, long waiting time, stigma and discrimination from health workers, are major barriers to HIV testing [10, 11].

In addition to conventional HIV testing and to break the barriers to HIV testing, the World Health Organization (WHO) recommended the HIV self-test (HIVST) in 2016 [12]. The HIVST differs from the conventional rapid diagnostic test; a layperson in private conducts the former, while the latter is conducted by health personnel, usually in a clinical setting [13]. HIVST involves the collection of blood or saliva samples, performing a rapid test, and interpretation of results in private [12]. The HIVST provides users with more privacy, confidentiality, and autonomy compared to providers-initiated testing, including facility-based barriers [14,15,16]. Regarding the effectiveness of HIVST, the HIVST kit has been proven to have excellent accuracy with a sensitivity of 94% (95% CI 90.2–96.7%) and a specificity of 99.7% (95% CI 99.3–99.9%) [17]. A mathematical model-based study reported that HIVST could lead to 96.5% diagnosis of HIV among priority populations and pregnant women [18]. The WHO further encourages individuals on pre-exposure prophylaxis (PrEP) to perform HIVST every 3 months to self-reassess HIV, suggesting its effectiveness [19].

Several systematic reviews and meta-analyses have emerged since the advent of oral and blood-based HIVST kits, and they have mostly focused on linkage to care. Of note, linkage to care is not synonymous with ART initiation because linkage to care can occur and ART initiation is delayed or declined, as evident in previous trials [20, 21]. Previous reviews have focused on the general population, sub-group population, or priority populations (e.g., MSM and female sex workers [FSW]) from a global or local perspective, with interest in comparing HIVST with the standard of care (SoC) and/or linkage to care. For instance, some studies have investigated the influence of HIVST on linkage to care or ART among young adults and men only in Africa or SSA [3, 15, 22, 23]. Some reviews have assessed the effectiveness of HIVST (linkage to care) compared to SoC among priority populations, such as MSM, FSW, and transgender [2, 24]. Also, some reviews have focused on HIVST among the general population [13, 25, 26].

Earlier reviews have focused on linkage to care after seropositive confirmation [13, 25, 26]. Ideally, HIVST should be accompanied by linkage to care in the form of ART or prevention such as PrEP. However, earlier reviews have not documented linkage to preventive practices, such as PrEP use, condom use, and condomless anal intercourse (CAI) following HIVST intervention. From the prevention side, our review is the first to systematically document the effectiveness of HIVST on PrEP use and sexual behavior change. This study aimed to assess the impact of HIVST interventions on (1) linkage to care or cART, (2) linkage to PrEP, and (3) sexual behaviors, such as consistent condom use and reduction in the number of sexual partners.

Methods

This systematic review and meta-analysis follow the Updated Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) guideline [27]. The study protocol has been published elsewhere [28].

Eligibility Criteria

The eligibility criteria followed the population/participants, interventions, comparison, outcome, and study design framework [29]. Population/Participants: we included studies that focused on the general population or priority populations, such as MSM and FSW, who were at least 15 years old. Interventions: studies that leverage HIVST as part of an intervention were considered. Comparison: only studies with a comparator group were included. Outcome: we evaluated the effectiveness of HIVST in terms of linkage to care after seropositive confirmation and adopted preventive practices, such as PrEP uptake, reduction in CAI, and number of sexual partners, after seronegative confirmation. Study design: we included only quasi-experimental or randomized control trials published in peer-review journals.

Database and Search Strategy

We searched six databases from their inception to November 11, 2022, without date or language limitation: PubMed, Web of Science, SCOPUS, Cochrane Library, CINAHL Plus (EBSCO), and MEDLINE (Ovid). We hand-searched Google Scholar and ResearchGate for grey papers on November 14, 2022.

All the articles generated from each database were exported into the Zotero reference manager to detect and delete duplicates. All duplicates were checked manually by one author (OAB). The remaining articles were exported to Rayyan for the title and abstract screening [30]. Two authors (OAB and CNN) independently screened the title and abstract of the included articles with an agreement of 92.9%. Conflicts were resolved through discussion between OAB and CNN; there was no basis to invite a third reviewer for resolution. The full-text screening was conducted independently by two authors (OAB and CNN), and discrepancies were resolved through discussion.

Data Extraction

One reviewer (OAB) extracted data from the included studies, verified independently by LEB and OAA. The following details were extracted from the articles: author and year of publication, country of study, characteristics of the participants (such as study population, proportion of gender, mean or median or mode age), study design (randomized or non-randomized trial), description of the intervention, HIVST type (oral or blood), duration of intervention, sample size, time of analysis (midline or endline), analysis type (intention-to-treat or otherwise), the proportion of linkage to care across intervention arms, and proportion of preventive behavior after HIVST (such as PrEP use, condom use, decrease in the number of clients, etc.). Corresponding authors were contacted when data were not clear or available. The extraction process was computed in Excel (version 2016).

Quality Assessment

Cochrane’s risk of bias assessment tool for randomized controlled trial (RoB 2.0) was used to appraise the quality of the included studies [31]. The risk of bias tool has five domains, including “bias arising from the randomization process”, “bias due to deviations from intended interventions”, “bias due to missing outcome data”, “bias in measurement of the outcome”, and “bias in selection of the reported results”. Each domain is measured against five possible options: Yes, Probably yes, No, Probably No, and No information. Yes and Probably Yes were merged as Yes, while No and Probably No were merged as No [31]. Regarding the scoring, studies with “low risk” across the five domains were classified as low-risk studies. Studies with at least one domain rated as “some concerns” were classified as having moderate risk studies. Studies with “high risk” in at least one domain or “some concerns” in two or more domains were classified as high-risk studies. The risk of bias was completed by OAB and verified by LEB and OAA.

Data Analysis

Review Manager (version 5.4) was used to analyze data. Linkage to care and ART were summarised with an odds ratio at a 95% confidence interval. Cochrane’s Q test was used to assess heterogeneity between the studies, while Higgins and Thompson’s I2 was used to quantify the level of heterogeneity. I2 value of 25%, 50%, and 75% was considered low, moderate, and high heterogeneity, respectively [32]. A random-effect model was fitted due to moderate heterogeneity. The subgroup analysis was computed for study population (priority population and general population), country of study according to WHO regional classification, duration of intervention, analysis type (intention-to-treat and otherwise), and mean and median age. Publication bias was assessed visually by inspecting funnel plots [13, 33]. Apart from the number of clients and non-clients per night that were included in the meta-analysis, other sexual behavioral changes, such as number of CAI and number of sexual activities, were narratively synthesized because the available data did not permit meta-analysis.

Results

A total of 2669 articles were retrieved from the six databases and two other sources (see Fig. 1). Of these, 1499 duplicates were deleted; 1170 were subjected to title and abstract screening. A total of 1111 articles that did not meet the set eligibility criteria were excluded, leaving 59 articles for full-text screening. Forty-four articles were excluded after full-text screening; the reason for exclusion include being a conference paper (n: 16), not an intervention study (n: 17), main outcome not HIVST (n: 8), and no comparison group (n: 3). Only 15 articles were included in this review. On appraisal with RoB 2.0 tool, nine studies were judged as “low risk”, while three studies each were considered to have “some concerns”, and “high risk” (Table 1). Also, we found no evidence of publication bias from the studies included in the meta-analysis (Supplementary Fig. 7).

Fig. 1
figure 1

Search strategy flowchart

Full size image
Table 1 Characteristics of included studies
Full size table

Description of Studies

Of the 15 studies, 14 were randomized control trials, while one study was a quasi-experimental study [34]. The majority were conducted in Uganda (n: 3), while two each were conducted in Malawi, South Africa, Zambia, China, and the United States. One study each was conducted in the Democratic Republic of Congo and Zimbabwe. Classifying the countries according to the WHO regional classification (Table 1), 11 of the 15 studies were conducted in the African region, two in the region of the Americas [35, 36], and two in the Western Pacific region [37, 38].

All interventions were randomized control trials except that of Pai et al. [34], a quasi-randomized trial. Similarly, all the studies used oral HIVST (OraQuick), except two studies that used blood-based self-test kits [37, 39]. The pattern of intervention varies across the studies; 12 of the 15 (80%) studies provided HIVST kits to participants in the intervention arm and information about HIVST and where it can be done in the SoC arm; three studies included a third arm, usually a coupon or incentive arm [20, 21, 40].

Regarding the duration of intervention, two studies collected endline data at 1 month [21, 34]; three studies obtained endline data at 3 months [14, 41, 42]. Four studies ended after 4 months [20, 40, 43, 44]; four studies’ endline was 6 months [35, 38, 39, 45]. One study collected endline data at 12 months [37], and another was a 15-month intervention trial [36]. Of the 15 trials, only six did not report intention-to-treat analysis [34,35,36, 39].

The population of interest for all trials differs; some studies focused on the general population, while others focused on priority groups, such as MSM and FSWs. Specifically, five trials recruited from the general population [34, 35, 39, 42, 45]. Two studies recruited male partners of women attending antenatal care [21, 41], and one study recruited male partners of women living with HIV [14]. For the studies that focused on the priority group, four trials focused on FSW [20, 40, 43, 44]; and three studies focused on MSM [36,37,38].

The mean, median, and mode age were reported across the studies. Ten of the 15 trials had participants with less than 30 years on average in both the intervention and SoC arm, while five had an average age of more than 30 years [14, 35, 36, 42, 45].

Effectiveness of HIVST Intervention on Linkage to Care

Nine of the 15 studies reported sufficient data for a linkage to care meta-analysis (Fig. 2). However, the study by Katz et al. [36] was excluded from the meta-analysis because 100% of MSM recruited were linked to care in both the HIVST and SoC arms leading to an inestimable analysis. Overall, our meta-analysis showed that linkage to care was similar between the HIVST arm and SoC (effect size: 0.92 [0.45–1.86]; I2: 51%; p: 0.04). In the population subgroup analysis, FSWs in the HIVST arm were significantly linked to care compared to the SoC arm (effect size: 0.53 [0.30–0.94]; I2: 0%; p: 0.41). Linkage to care does not differ between the partners of antenatal care women in the HIVST arm attending antenatal care, and those in the control arm (effect size: 2.41 [0.90–6.50]; I2: 0%; p: 0.85); linkage to care in the general populations was similar between HIVST and SoC arm (effect size: 1.18 [0.21–6.58]; I2: 54%; p: 0.85).

Fig. 2
figure 2

Linkage to care by population

Full size image

Subgroup analysis by region revealed that linkage to care was similar between HIVST and SoC arms in the African region (effect size: 0.81 [0.42–1.55]; I2: 46%; p: 0.09; Fig. 3). Also, subgroup analysis by the duration of intervention, analysis type, and age showed a similar link to care between HIVST and SoC arms (see supplementary figures).

Fig. 3
figure 3

Linkage to care by region

Full size image

Effectiveness of HIVST Intervention on ART Initiation

Eight studies provided relevant ART initiation data. HIVST intervention did not significantly improve ART initiation in the HIVST arm compared to the SoC arm (effect size: 0.90 [0.45–1.79]; I2: 74%; p: < 0.001). Subgroup by population shows that linkage to care does not differ between HIVST and SoC among FSW, male partners, or other populations (Fig. 4). Similarly, other subgroup analyses (region, duration of intervention, analysis type, and age) did not show differences in ART initiation between HIVST and SoC arm (Fig. 5 and Supplementary figures).

Fig. 4
figure 4

ART initiation by population

Full size image
Fig. 5
figure 5

ART initiation by region

Full size image

Effectiveness of HIVST on Linkage to PrEP and Sexual Behavioral Change

Only one study reported the impact of HIVST intervention on PrEP uptake [14]. The study found that more male partners of women living with HIV in the SoC arm initiated PrEP compared to partners in the HIVST arm (16% versus 7%).

Five studies provided relevant sexual behavioral data, and all reported these behaviors among two priority groups—FSW and MSM. The identified behavioral changes include the number of clients per night, number of non-client partners in the past month among FSWs, CAI, number of male and female partners and consistent condom use with regular and casual partners among MSM.

Only two studies with a similar endline of 4 months reported on the number of clients per night and non-client partners past month among FSW [40, 44]. The meta-analysis shows no difference between the HIVST and SoC arm regarding the number of clients FSW seen per night (effect size: − 0.66 [1.35–0.02]; I2: 64%; p: 0.09; Fig. 6); a similar observation was noticed with non-clients past month (effect size: − 1.45 [− 1.45 to 1.38]; I2: 93%; p: < 0.001; Fig. 7).

Fig. 6
figure 6

Number of clients per night

Full size image
Fig. 7
figure 7

Number of non-clients per night

Full size image

Regarding MSM, the mean number of CAI with a male partner was higher in the HIVST group than in the SoC group (2.34 versus 2.14), as reported by one study [36]. However, Zhang et al. [37] found that MSM in the HIVST group were insignificantly less likely to have a higher number of male partners.

One study reported non-concordant CAI and found that compared to the SoC group, MSM in the HIVST group had more non-concordant CAI (29% versus 24%) [36]. Similarly, another study found that HIVST intervention among MSM did not significantly reduce the number of female and male sexual partners compared to the SoC arm [37].

Two studies reported the impact of HIVST intervention on condom use among MSM, but the data presented in these studies did not meet the criteria for meta-analysis. One of the studies found that the HIVST group regularly and significantly used condoms with the main partner more than the SoC group [37]; the other study found a similar insignificant finding [38]. The influence of HIVST intervention on condom use with casual partners differs between both studies, although they found an insignificant association. Zhang et al. [37] found that the HIVST group were less likely to use a condom with a casual partner at 12 months, while Zhu et al. [38] found that a fewer percentage of those in SoC regularly use a condom with a casual partner than the intervention group at 15 months. Only one study reported HIVST intervention on condom use during CAI; the use of condoms during insertive or receptive CAI was similar between both study arms [38].

Discussion

This review investigated the linkage to care, ART, or PrEP following HIVST intervention and the influence of HIVST intervention on sexual behaviors. We found that linkage to care does not improve with HIVST compared to SoC. Two earlier reviews and one observational study that reported linkage to care from an earlier trial among the general population found mixed observations [2, 25, 46]. The meta-analysis by Jamil et al. [25] was performed among the general population only; that is, priority groups such as FSW and MSM were not considered. They found a moderate heterogeneity and non-significant difference between HIVST and SoC arms. However, Plazy et al. [46] reported that HIVST intervention (home visits and phone calls) significantly increased linkage to care four folds among the general population in South Africa.

Interestingly, Witzel et al. [2] found that the SoC arm significantly linked to care compared to the HIVST arm in their meta-analysis of priority population only (FSW, MSM, and transgender men), with zero heterogeneity. The present review considered the general and priority groups, which may account for the disagreement with Witzel et al. study [2]. Furthermore, this present study included only peer-reviewed published trials, while the mentioned review also considered non-peer-reviewed articles.

It is worrisome that HIVST interventions did not increase linkage to care or ART among newly diagnosed PLWH; also, linkage to PrEP was significantly worse among partners of women living with HIV, according to the only study that reported the effectiveness of PrEP [14]. This finding has implications for the scale-up of HIVST intervention in SSA, where many of the included interventions were conducted. Poor political will is common in SSA [47]; the lack of HIVST effectiveness in increasing linkage to care may meet stiff resistance from political actors who should ensure the scale-up of oral HIVST. It may be difficult to convince stakeholders to allocate resources to scale up HIVST if there is evidence that such interventions do not improve linkage to care, inhibiting an evidence-informed policy.

Another concern worth mentioning is the plausible reasons for the lack of HIVST effectiveness in linkage to care and ART. One major explanation could be the assumption of transferability of intervention. Implementing bodies and researchers should not assume that successful intervention elsewhere guarantees a successful intervention in another location or time; that assumption may be problematic because: (1) the complexity of interventions may be difficult for inexperienced researchers, which can reduce intervention fidelity leading to an ineffective HIVST intervention; (2) sociocultural difference may interact directly or indirectly with an intervention because socio-behavioral influences are difficult to manage, and (3) ethical or legal variations may not allow for proper adoption of an intervention, in the case of priority population for instance [48, 49].

Our subgroup analysis showed that FSW in intervention arms linked to HIV treatment better than the SoC arm, which corroborates the findings of a recent meta-analysis [2]. This finding is not surprising; the reverse would be counterintuitive. FSWs are 30 times more likely to contract HIV than the general population, suggesting their high level of susceptibility to HIV infection [50].

Regarding behavioral change, we found no significant evidence that the number of clients per night and non-clients per month reduces with HIVST intervention. Coincidentally, the two studies that reported the number of clients and non-clients focused on FSWs [40, 44]. The line of no effect was touched slightly in our meta-analysis, making FSWs’ advantage over the SoC non-significant. The evidence shows the promising impact of HIVST on the reduction in contact with the number of clients and non-clients (such as boyfriend), which can reduce the risk of HIV infection among FSWs. However, more interventions tailored towards change in the sexual behaviour of FSWs are needed to make a reasonable conclusion.

CAI by MSM with male or non-concordant partners did not differ between HIVST and SoC; although the data available did not permit data aggregation, this is a direction that needs further interrogation from HIVST researchers. The use of condoms by MSM during sexual acts with a regular male partner appears to increase with HIVST intervention, but we cannot draw inference from the available insufficient data. An earlier review reported similar condomless sex between priority groups (MSM, FSW, and transgender people) that received HIVST intervention compared to others [2]. But the trace of increment in condom use among MSM following HIVST intervention is promising because they are disproportionately at a higher risk of HIV infection in both low and middle income countries and high income countries [51, 52].

Strengths and Limitations

There are several strengths to this present review. This is perhaps the first review to consider only peer-review RCTs, making the findings and conclusion more robust. In addition, the inclusion of only peer-reviewed RCTs makes our findings largely trusted scientific evidence. This review is not without a few limitations. First, many of the RCTs included were conducted in the African region; therefore, caution is required during interpretation, and these findings may not be generalizable. Second, subgroup analyses should be interpreted with caution as they are subjected to false negatives (and positives); insufficient number of participants in subgroup analysis can mask the true effect due to low power, leading to false negative results and false positive results may occur due to multiple comparisons [53]. Third, the duration and endline of these interventions differs and should be considered while interpreting the results. Finally, misinterpretation by self-testers which can lead to a lack of follow-up for confirmation and linkage to care and ART.

Conclusion and Recommendations

The present review found that HIVST interventions do not increase linkage to care, ART, or PrEP. However, FSWs tend to link to care compared to SoC, but no evidence of healthy sexual behavior, such as a reduced number of paying and non-paying clients among this group. Results on CAI among MSM were mixed for both regular male and casual (male and female) partners. The following recommendations are preferred based on the findings of the present review. First, many HIVST interventions have focused on linkage to care or ART, and we found only one RCT that considered linkage to PrEP following HIV seronegative confirmation. This is particularly important among priority groups with a disproportionately higher risk of HIV infection, such as MSM and FSW. Second, more robust RCTs focusing on linkage to care among MSM are urgently needed because most of the available evidence have focused on sexual behavioral change. Third, many HIVST interventions have considered the use of condoms largely among MSM, while FSWs who arguably engage more in transactional sex have rarely been assessed with respect to condom use. Therefore, more empirical trials are warranted to understand how condom use is impacted by HIVST intervention. Finally, there is a need to extend HIVST interventions to other regions besides the African region; this will allow adequate comparison with existing trials in Africa and can birth a more holistic approach to HIVST research.