Introduction

The primary goal outlined in the 2021 U.S. National HIV/AIDS Strategy (NHAS), supported by the Ending the HIV Epidemic in the U.S. (EHE) initiative, is to reduce the number of new HIV infections by 75% by 2025 and by 90% by 2030 [1, 2]. HIV testing is one of the cornerstones of HIV prevention efforts and a gateway to accessing pre-exposure prophylaxis and antiretroviral therapy [1]. The U.S. Centers for Disease Control and Prevention (CDC) recommends HIV testing at least once for all Americans aged 13–64 years [3], and at least annually for sexually active gay, bisexual, and other men who have sex with men (MSM) [4], a group disproportionately affected by HIV [5]. However, many MSM in the U.S. are not being tested for HIV as recommended [6], which contributes to delayed testing in this population [7]. For example, a national survey reported that the median interval since last HIV test was 3.0 years, and 1.3 years among sexually active MSM [7].

Despite concerted efforts to scale up HIV prevention activities for MSM, barriers to HIV testing persist. National HIV Behavioral Surveillance data indicate that perceived low risk was one of the main reasons reported by MSM who did not test in the past year [8]. There are several barriers to HIV testing among MSM, including discrimination, racism, homophobia [9], and anticipated stigma [10]. HIV self-testing offers a unique opportunity for MSM who might be experiencing challenges using traditional testing services to learn their HIV status, and it has been found to be highly acceptable and feasible among diverse populations, including MSM [11,12,13]. In our randomized controlled trial (RCT), the Evaluation of Rapid HIV Self-testing Among MSM Project (eSTAMP), we found that 76.6% of participants in the intervention arm, who received HIV self-tests (HIVSTs) by mail, reported testing three or more times during the study year, compared to 22.0% of participants in the control arm, who only received HIV prevention information and resources to locate local HIV testing services. Additionally, the cumulative number and percentage of newly identified HIV infections was higher among participants in the intervention arm (25/1325; 1.9%) than in the control arm (11/1340; 0.8%) [14].

Social and sexual networking strategies have been used as an approach to identify persons at high risk for HIV infection [15]. Peer-driven networking strategies [16] have also been used to leverage the potential of HIV self-testing [16,17,18]. Specifically, peer-based distribution of HIVSTs (also called “secondary distribution”) to members of one’s social network (e.g., family, friends) or sexual network (e.g., main partners, casual partners) may be an effective strategy to increase testing coverage among populations at an elevated risk for acquiring HIV, including MSM. In one study, MSM and those who were willing to use HIVSTs themselves reported high levels of willingness to distribute HIVSTs to their friends, sexual partners, family members, and other acquaintances (collectively referred to as social network associates, SNAs) [19]. Few studies have examined the demographic and behavioral characteristics associated with the distribution of HIVSTs by MSM to their SNAs. Understanding how these characteristics impact the decision to distribute HIVSTs could help public health planners determine whether the secondary distribution of HIVSTs would work better in some subpopulations than in others (e.g., people who live in certain geographic regions, represent a specific age group, or those who have never tested for HIV). In addition, understanding the reasons why people at risk for HIV refrain from distributing HIVSTs to their SNAs (e.g., concerns about negative reactions) offers insight into the barriers of distributing HIVSTs. Research on the distribution of HIVSTs to SNAs has shown promise in identifying preliminary positive test results. In an HIV self-testing distribution study of persons with HIV, 90.0% reported giving at least one HIVST to an SNA [20]. Of the 31 HIVST SNA results provided by the study participants, 2 (6.5%) were positive [20].

The eSTAMP trial evaluated the effect of providing HIV self-tests on frequency of testing and identification of HIV infection [14]. In the intervention arm of the eSTAMP study, participants who distributed HIVSTs reported demographic information about the SNAs to whom they gave these HIVSTs, and the results of the tests performed with these kits, to the extent that they knew this information. In this analysis we describe: (1) the likelihood of distributing HIVSTs by demographic and behavioral characteristics of study participants; (2) the reasons for not distributing HIVSTs to SNAs; and (3) the likelihood of SNAs testing positive for HIV by their own demographic characteristics and by the distributor’s demographic and behavioral characteristics.

Methods

A detailed description of eSTAMP has been published previously [14]. Briefly, from March 2015 through November 2016, 2665 internet-recruited MSM participated in a 12-month longitudinal, 2-arm RCT. Eligibility criteria included: male sex at birth, identifying their gender as male, being at least 18 years of age, residing in the U.S., having had anal sex with at least one man in the past 12 months, and reported having tested HIV-negative or being unaware of their HIV status. This analysis is limited to the 1325 participants in the intervention arm of the eSTAMP study.

Participants in the intervention arm had access to online resources to locate local HIV testing services and were asked to complete a baseline survey and quarterly surveys during a 12-month period (i.e., at 3, 6, 9, and 12 months). After completing the baseline survey, participants in the intervention arm were mailed 4 HIVSTs that had at least 6 months remaining prior to their expiration date: two oral fluid tests (OraQuick® In-Home HIV Test, OraSure Technologies, Inc.) and two finger-stick blood tests (SURE CHECK® HIV 1/2 Assay, Chembio Diagnostics, Inc.; used under an investigational device exemption from FDA). Participants were not prompted or encouraged to distribute the HIVSTs to SNAs, and after completing the 3-, 6-, and 9-month surveys, participants could request replacement HIVSTs for those they reported having used themselves or distributed to SNAs. The protocol was approved by the Emory University Institutional Review Board in Atlanta, Georgia.

Among eSTAMP participants who responded to the questions about distributing HIVSTs in any follow-up survey, we computed the proportions who reported distributing one or more tests, overall and stratified by various participant characteristics, along with Clopper–Pearson exact 95% confidence intervals (CIs). We also computed p-values using Fisher’s exact test for participant characteristics with two categories and using a permutation test for participant characteristics with more than two categories. In addition, we computed the pairwise differences in proportions, along with exact (score) 95% CIs and Barnard’s exact test p-values. Furthermore, we used McNemar’s test to compare the proportions of participants who reported distributing any OraQuick or Sure Check tests.

Participants who responded that they had not distributed HIVSTs since the previous survey were asked why they had not distributed tests to SNAs. They could select from nine prespecified reasons or write a free-text response. In each follow-up survey, participants could provide multiple reasons and could repeat reasons on subsequent surveys. We reviewed all responses, created new reasons based on the most common free-text responses, and then selected the top nine reasons, each of which were reported by at least 5% of all responding study participants. These top nine reasons consisted of five prespecified reasons and four composite reasons. The composite reasons resulted from combining similar prespecified or new reasons. We then computed the numbers and observed proportions of reasons that participants gave for not distributing the HIVSTs, overall and at each follow-up survey, along with Clopper–Pearson exact 95% CIs (see Table 2).

To determine whether participants distributed HIVSTs to the same SNAs or not, we matched SNAs using information provided by the participants (e.g., initials, name, or nickname; age; gender; and race/ethnicity) within each follow-up survey. SNAs who had a positive test result from any of the HIVSTs within a particular survey were classified as having an “identified HIV infection.” In turn, SNAs who had a positive result within a particular survey were classified as having a “newly identified HIV infection” when the participant reported either that (a) the SNA did not already know that he/she had HIV or (b) the participant did not know whether the SNA already knew that he/she had HIV. See the Supplementary Notes for more details.

Among the unique SNAs for whom one or more test results were reported, we computed the proportions of SNAs who had an identified HIV infection and who had a newly identified HIV infection, both overall and stratified by various participant and SNA characteristics. We computed adjusted proportions, along with 95% CIs, using generalized estimating equation (GEE) methods (based on the binomial distribution with the logit link function) with an exchangeable correlation, which accounts for the correlation between SNAs within participants (“clusters”). Additional details on the GEE weighting methods to deal with categories with no positive HIVST results and the sensitivity analyses to assess the impact of weighting choices can be found in the Supplementary Notes. Due to small sample sizes and unstable estimates for transgender persons, the GEE analysis of SNA gender was restricted to male and female SNAs. We also computed p-values using generalized score tests (GSTs). All calculations were performed using SAS software, version 9.4 (SAS Institute, Inc., Cary, NC).

Results

Among the 1325 intervention arm participants, 995 (75.1%) responded to the questions about distributing HIV self-tests in any follow-up survey. Most (73.7%) of the respondents were < 35 years of age, 61.3% were non-Hispanic White, 8.7% were non-Hispanic Black, and 21.2% were Hispanic. One-third of participants (n = 328; 33.0%) reported not distributing any HIVSTs on the surveys that they answered, whereas 667 (67.0%) participants reported distributing tests as follows: only OraQuick tests (n = 127; 12.8%); only Sure Check tests (n = 63; 6.3%); both types of tests (n = 477; 47.9%). Comparing the percentages of participants who distributed any OraQuick tests and who distributed any Sure Check tests, more participants distributed OraQuick tests (n = 604 = 127 + 477; 60.7%) than Sure Check tests (n = 540 = 63 + 477; 54.3%) (difference = 6.4%; approximate 95% CI 4.0%, 8.9%; McNemar’s test p-value < 0.001). Collectively, the 667 participants who reported distributing HIVSTs gave away a total of 2862 tests (1546 OraQuick and 1316 Sure Check tests).

In every demographic and behavioral category considered in this analysis, more than 60% of participants reported distributing HIVSTs to their SNAs (Table 1). The percentages of participants who reported distributing HIVSTs vary somewhat by their demographic characteristics, including age group [ranging from 62.7% (18–24 years) to 70.9% (45–54 years)], race/ethnicity [66.4% (Hispanic/Latino) to 74.4% (other/mixed)], education [66.4% (greater than high school) and 71.0% (high school/equivalent or less)], and geographic region [64.3% (Midwest) to 69.1% (West)]. The percentages of participants who reported distributing HIVSTs also vary somewhat by participant risk behaviors in the 3 months prior to study enrollment, including whether the participants had any condomless anal sex with a male partner [62.8% (No) and 68.3% (Yes)], any condomless anal or vaginal sex with a female partner [66.6% (No) and 83.3% (Yes)], any non-prescription injection drug use [66.8% (No) and 100.0% (Yes)], and any non-injection drug use [66.0% (No) and 69.8% (Yes)]. In addition, the percentages vary somewhat by HIV testing history prior to enrollment [63.6% (tested over 12 months ago) to 68.3% (tested in past 12 months)]. None of the differences in the percentages of participants distributing HIVSTs by participant demographic and behavioral characteristics were statistically significant (all Fisher’s exact test p-values > 0.10, all Barnard’s exact test p-values > 0.08; see Supplementary Notes).

Table 1 Numbers and percentages of eSTAMP intervention arm participants who reported distributing HIV self-tests (HIVSTs) to social network associates by baseline participant demographic and behavioral characteristics, United States, 2015–2016
Full size table

Of the 995 participants, 674 gave one or more specific reasons for not distributing HIVSTs during at least one of the four follow-up periods (Table 2). The nine most common reasons that participants gave for not distributing HIVSTs were: (a) they wanted to use the HIVSTs themselves (n = 505; 74.9%); (b) they thought their SNAs would get angry or upset if they offered them HIVSTs (n = 84; 12.5%); (c) they didn’t know that they could give the HIVSTs away (n = 76; 11.3%); (d) they were afraid that others would think that they had HIV (n = 75; 11.1%); (e) the HIVSTs expired (n = 74; 11.0%); (f) they did not know anyone who was interested in the HIVSTs or did not have the opportunity to distribute them (n = 72; 10.7%); (g) they were concerned that it might affect their relationships, or be awkward or embarrassing (n = 69; 10.2%); (h) they would rather their SNAs talk to a counselor when testing, or SNAs are already testing or prefer clinics (n = 55; 8.2%); and (i) they were concerned about test accuracy or that their SNAs wouldn’t perform or read the HIVSTs correctly (n = 39; 5.8%). Over the course of the study, the frequencies of reasons (a), (c), and (d) each decreased by at least 5% points. By contrast, the frequencies of reasons (b), (e), (f), (h), and (i) remained roughly constant, whereas the frequency of reason (g) increased slightly. Note that (a), (c), and (e) pertained to logistical reasons for not distributing, (b), (d), and (f) pertained to anticipated reactions and dynamics between the participants and their SNAs, (g) pertained to participants’ perceptions about SNAs’ testing preferences, and (h) pertained to participants’ concerns about self-test performance and result interpretation by SNAs.

Table 2 Frequencies of reasons given by eSTAMP intervention arm participants for not distributing study HIV self-tests (HIVSTs), ever and on four follow-up surveys, United States, 2015–2016
Full size table

Of the 667 participants who reported distributing HIVSTs, 658 participants (98.7%) responded to questions about the 2301 SNAs to whom they distributed HIVST kits, corresponding to a total of 2737 test results (1467 OraQuick and 1270 Sure Check test results). These participant-reported SNA HIVST results were negative (n = 1842; 67.3%), positive (n = 57; 2.1%), invalid (n = 21; 0.8%), and “I don’t know the result of the test” (n = 817; 29.9%). Overall, based on all the HIVST results, 52 SNAs (adjusted percentage: 2.2%; 95% CI 1.5, 3.1) had an identified HIV infection and 34 SNAs (adjusted percentage: 1.3%; 95% CI 0.9, 2.1) had a newly identified HIV infection (Table 3). There were few statistically significant differences in the percentages of SNAs who had identified infections and newly identified HIV infections by distributor demographic and behavioral characteristics (most GST p-values > 0.05). Specifically, there only were statistically significant differences in the percentages of SNAs who had identified infections and newly identified HIV infections by distributor education level (GST p-values = 0.028 and 0.006, respectively) and in the percentages of SNAs who had newly identified HIV infections by distributor non-injection drug use in the past 3 months (GST p-value = 0.002).

Table 3 Numbers and proportions of unique social network associates (SNAs) who had identified HIV infections and newly identified HIV infections by distributor demographic and behavioral characteristics at baseline, eSTAMP, United States, 2015–2016
Full size table

Regarding the SNA demographic characteristics (Table 4), the adjusted percentages of SNAs who had an identified HIV infection vary narrowly by SNA age group (ranging from 0.0 to 5.3%; GST p-value = 0.081), race/ethnicity (1.1–3.6%; p-value = 0.398), and gender (2.4%, males vs. 0.0%, females; p-value < 0.001). Likewise, the adjusted percentages of SNAs who had a newly identified HIV infection vary narrowly by SNA age group (ranging from 0.0 to 2.5%; p-value = 0.465), race/ethnicity (0.9–2.0%; p-value = 0.768), and gender (1.4%, males vs. 0.0%, females; p-value = 0.021).

Table 4 Numbers and proportions of unique social network associates (SNAs) who had identified HIV infections and newly identified HIV infections by their demographic characteristics, eSTAMP, United States, 2015–2016
Full size table

Discussion

In this study of sexually active MSM, two-thirds of participants in the intervention arm of eSTAMP reported distributing HIVSTs to their SNAs, even though the study team did not provide any messaging, training, or guidance on the distribution of HIVSTs. This unprompted distribution of HIVSTs by participants to people in their social networks resulted in 52 HIV infections being identified among SNAs overall, and of these, 34 were classified as newly identified HIV infections [14]. Although the percentages of those with an identified HIV infection and those with a newly identified HIV infection vary somewhat by demographic and behavioral characteristics of participants, most of the differences were not statistically significant. Notably, more than half of participants in nearly every socio-demographic category chose to distribute HIVSTs to their SNAs suggesting robust willingness and feasibility of secondary HIVST distribution. These findings support the inclusion of secondary HIVST distribution as part of an HIV prevention program for sexually active MSM.

Many reported not distributing tests on at least 1 of the 4 follow-up surveys. Understanding the reasons for not distributing tests will allow program planners to overcome barriers in implementing HIVST distribution (secondary distribution) interventions. The main reason for not distributing tests was a logistical one (e.g., “I wanted to use the HIVSTs myself”). However, the proportion citing this reason decreased over the follow-up survey periods, which indicates that participants might have become more aware of the opportunity to order more self-test kits, and thus were not saving kits from each order for themselves. Some reasons reported for not distributing HIVSTs to SNAs (e.g., “I was afraid that others would think I have HIV” and “I was concerned it might affect my relationships”) pertain to perceived reactions and dynamics between the participants and the SNAs, which may reflect an uncertainty on how to approach a conversation on HIVST distribution. Instructions on and tools for the secondary distribution of HIVSTs, including communication guidance on how to address negative reactions or perceptions from SNAs, might allay these concerns and aid in skill-building to support the distribution of HIVSTs. In addition, qualitative studies might be needed to help contextualize the relationships between distributors and their SNAs to better understand the barriers and facilitators of peer-based distribution. Finally, some participants reported concerns about HIVST performance, that the SNAs would not perform the test correctly, or that the SNAs would misinterpret the result. In addition, some participants preferred that their SNAs talk to a counselor when getting an HIV test, or they believed that their SNAs were already testing or preferred clinic-based testing, which suggests that HIV self-testing should not replace facility and community-based testing strategies but should instead complement them.

HIV self-testing is one strategy to increase availability of testing to unreached populations under the EHE initiative [21], and HIVST distribution may be a sound strategy for increasing the frequency of screening for sub-populations with high risk for HIV infection [22]. A cost-effectiveness analysis of eSTAMP, based on the intervention cost of $449,510, estimated that HIV self-testing potentially averted 3.34 transmissions and thus saved 14.86 quality-adjusted life years and nearly $1.6 million in lifetime HIV treatment costs [23], which provides further support for implementing HIV self-testing strategies more widely in the U.S.

Since the novel coronavirus disease 2019 (COVID-19) pandemic, there has been an increased need to promote and implement HIV self-testing, as traditional in-person HIV testing has been disrupted. Although the data from this study were collected prior to the pandemic, the findings support peer-based distribution of HIVST as a modality to increase the reach of HIV self-testing. CDC has encouraged health departments and community-based organizations to consider implementing HIV self-testing programs [24]. The CDC’s HIVST distribution program provided an option to fill the need for increased access to HIV testing. In this program, persons over the age of 17 could order two HIVSTs online, every 90 days, but were not given instructions to distribute the HIVSTs. Program participants who had been previously diagnosed with HIV or were taking HIV pre-exposure prophylaxis (PrEP) medications were encouraged to distribute these HIVSTs to their SNA(s) [25]. HIVST distribution provides an opportunity to increase the frequency and reach of HIV testing among MSM [14] and can be a key strategy in a comprehensive program to promote HIV testing among MSM and their SNAs.

There are several limitations to consider. First, about a quarter of participants did not respond to any follow-up survey, and it is unknown whether the experiences of non-respondents might have been different than the experiences of respondents. Thus, we do not know if the observed percentage of participants who reported sharing the HIVSTs with SNAs is an underestimate, and hence there could be more distribution of HIVSTs than captured through the surveys and, in turn, more SNAs who tested positive. Second, distributors of the HIVSTs reported on the characteristics of their SNAs to the best of their knowledge, so information on SNAs may be imprecise. Specifically, there could be under-reporting of positive results because the SNAs had to tell the study participant that they used the HIVST and their test result; indeed, nearly a third of results were not known. Positive results might have been differentially under-reported to study participants. This limitation reduces the accuracy of our estimates and likely leads to underestimating the true proportions of SNAs who obtained a positive test result. Third, it is very likely that the matching process used to estimate the 2301 unique SNAs may have over- or under-matched SNAs, due to inconsistent fidelity in reporting by participants on information about their SNAs or HIVSTs, and due to methods (as stated above) of only trying to match SNA information within a single survey time and not across surveys. Finally, the nature of the reporting of the SNAs’ HIVST result, along with the modest sample sizes, may have hampered our ability to detect meaningful differences in the proportions of SNAs who had an “identified HIV infection” or “newly identified HIV infection,” by either distributor or SNA characteristics, so it is possible that we missed important differences.

Conclusion

To meet the goals of NHAS and overcome disruptions in face-to-face services because of the COVID-19 pandemic, new strategies are needed to prevent HIV infections in the U.S. The high percentage of participants who distributed HIVSTs to their SNAs in the current study, even in the absence of specific instructions to do so, is promising. The willingness of study participants to share HIVSTs with their SNAs indicates that this may be an effective strategy to increase awareness of HIV status in the broader community, which in turn may contribute to ending the HIV epidemic in the US. Programs that distribute HIVSTs to MSM could provide multiple kits at one time and encourage secondary kit distribution [24,25,26]. Providing written instructions or a how-to video guidance may help facilitate the distribution of HIVSTs to their SNAs. Peer-based distribution programs of HIVSTs show promise for increasing awareness of HIV status among social network associates of sexually active MSM.