Study design and participants. This randomized implementation trial was conducted between August and November 2018 in Kisangani, the DRC. Kisangani, the capital city of Tshopo Province, is the third-largest urbanized city in the DRC, with 1.6 million inhabitants and a 2.3% HIV seroprevalence in the general public aged 15 to 49 years [22].
Trained research assistants (physicians or nurses) enrolled participants at four facilities (University Hospital of Kisangani, General Hospital of Kabondo, and the health centres of Neema and Saint Joseph). These facilities were selected because they integrate HIV prevention and care packages, provide free care to people living with HIV, and provide convenient access for those at high risk of exposure in the transmission hotspots in Kisangani. The survey was promoted and made visible by placing posters in the facilities and by informing people in the transmission hotspots in Kisangani and distributing tokens redeemable at the clinics to anyone both during the day and at night.
Participants were eligible for the study if they were between 18 and 49 years old, were at high risk of acquiring HIV infection, did not know their HIV status, lived or worked in Kisangani for at least 6 months before enrolment, and were available and accessible by phone. High risk of acquiring HIV infection was defined as being sexually active with a history of unprotected intercourse with one or more partners of unknown HIV serostatus within the past 6 months, having had new sex partners in the past 6 months, having symptoms of sexually transmitted infections (STIs) in the same period, engaging in commercial sex activities, or being in a known HIV discordant partnership [18].
Randomization procedures. Participants were randomized at a ratio of 1:1 through block randomization (block size 4, 6, 8). Eligible participants were randomly assigned to one of two self-testing groups (Fig. 1), DAH or UH, using sequentially numbered sealed randomization envelopes. Because of the nature of the intervention, the study participants and study staff could not be blinded. However, the study staff and participants were unaware of the assignment until the envelope was opened.
Study procedures and data collection. The blood-based HIVST was performed using the Exacto® Test HIV (Biosynex, Strasbourg, France) self-test kit, which includes simplified pictorial instructions printed in colour in A3 format for use in French, Lingala, and Swahili, as previously reported [8]. After obtaining written informed consent and before randomization, the participants completed a self-administered baseline questionnaire to collect data on their demographic characteristics, sexual behaviour, and HIV testing history, after which they received adequate pre-test HIV counselling.
In the DAH arm, a brief, the participants watched a 10 min face-to-face demonstration of how to use the self-test to familiarize them with the contents of the self-test kit. After this, the participant was asked to perform the HIV self-test in a private room supervised by trained research assistants (supervisors). After completing the test, the participant completed a practicability report using a standardized sheet. The supervisor-interpreted results and the requests for verbal assistance were recorded by the supervisor on a standardized sheet. The supervisors had received rigorous training on how to talk to the participants when they asked for verbal assistance when performing the HIV self-test.
In the UH arm, the participants were asked to perform the HIV self-test at home or in a convenient private location and then read the results guided only by the test’s instructions without the 10 min demonstration or supervision. The participants were instructed to complete the practicability report using the standardized sheet within 10 min of performing the self-test. Furthermore, the participants were invited to return with the test cassette and the standardized sheet (placed in a sealed envelope) to the facility within 12–72 h to re-read the test results and undergo an additional evaluation. Telephone assistance was offered to the participants if needed. The need for assistance from a trusted person was self-reported by the participants and recorded by the investigators.
In each study arm, a confirmatory HIV test using a national rapid test algorithm [21, 22] was performed after HIVST if the test was reactive. If seropositivity was confirmed, the participants were referred to care services. Post-test counselling was provided to the participants if needed. The standardized sheet included questions that asked the participant to confirm the presence of blood in the square well of the test, describe the appearance of the control strip in the self-test, and report the final interpreted self-test result. The final results were recorded as one of three outcomes: (i) may have HIV (preliminary positive), (ii) do not have HIV (preliminary negative), and (iii) test not working (invalid). There was a 24 h helpline for the participants in which anonymity was assured by instructing the participants to introduce themselves using their three-character randomization code. The investigator recorded all information about the provided telephone assistance on a follow-up sheet.
An exit questionnaire was self-administered after completing all parts of the testing process. The questionnaire assessed satisfaction, the willingness to purchase an HIV self-test kit, and the unit purchase price of the test in United States dollars (USD).
Study Outcomes. The primary outcome was the difference in the practicability of the Exacto® Test HIV (Biosynex) self-test kit between the UH and DAH approaches. Practicability was defined as the successful performance of the HIV self-test and the correct interpretation of the HIV self-test result. The successful performance of the HIV self-test was determined by the presence of the control strip. The self-test result was interpreted as negative when the Control line (C) was present and readable and the Test line (T) was absent. The result was positive when the ‘C’ and ‘T’ (clearly or poorly readable) lines were present, and it was invalid when the ‘C’ line was absent regardless of the presence or absence of the ‘T’ line.
Secondary outcomes included the proportion of participants who requested assistance, the retention rate, positivity rate with confirmation HIV testing, linkage to care, and willingness to buy the HIV self-test kit if locally available. The retention rate was defined as the number of included participants who completed the entire evaluation process through the follow-up period. We used the above secondary outcomes as our measures of the effectiveness of HIVST between the UH to the DAH approaches under field conditions in the DRC.
Sample size. A one-sided design to test non-inferiority between the groups was used; specifically, the hypothesis that the practicability of UH is objectively non-inferior to that of DAH was tested. The sample size was estimated using the following formula: n = 1/ΔL2 (Z1-a Ö[pN (1-pN) + pR (1–pR)]) + Z1-b Ö[2pR (1 – pR)])2, where πN and πR are the success proportions for UH and DAH, respectively, with α = 0.05 for a 95% confidence interval, b = 0.2 for a power of 80%, and the non-inferiority limit (ΔL) corresponding to the greatest loss of effectiveness that is possible to consent [23]. The non-inferiority limit was conventionally set at -10% based on previous studies [8, 13, 22]. We assumed that πN=98% and πR=88%. The sample size (n=456) was increased by 10% to account for loss at follow-up, giving a final sample size of 530.
Statistical analysis. MS Excel was used to construct a database to encode the data, and SPSS 20.0 (Chicago, IL) software was used for the analyses. First, descriptive statistics using the mean (standard deviation) or median (interquartile range) for a normal distribution or skewed distribution was computed, respectively. Next, the outcome measures in the two study arms were compared using Pearson’s chi-squared test for categorical data or Student’s t-test for means.
A one-sided Wald asymptotic test was used to assess the non-inferiority of the successful performance of the HIV self-test, correct interpretation of the HIV self-test result, requests for verbal assistance, retention rate, positive rate, linkage to care, and willingness to buy the HIV self-test kit if locally available in the UH arm, and these results were compared to those in the ADH arm. The confidence interval for the difference was based on the Wald asymptotic method, with an alpha level of 0.05 for 95% confidence limits. Non-inferiority was defined as a lower limit > -10 of the 95% CI around the difference in outcomes. To determine the effects of interventions (UH versus DAH) on the primary and secondary outcomes, adjusted risk ratios (aRRs) calculated using Poisson regression were evaluated using two-sided statistical tests with a significance level set at P < 0.05. The participants who did not successfully complete the follow-up were not included in the analyses, as it was not possible to determine the primary and secondary outcomes for them.
Per cent agreement and Cohen’s κ coefficient were used to estimate the agreement between the participant-interpreted results and investigator-interpreted results. The degree of agreement was determined by Landis and Koch ranking (κ=0: poor agreement; κ=0.01–0.20: slight agreement; κ=0.21–0.40: fair agreement; κ=0.41–0.60: moderate agreement; κ=0.61–0.80: substantial agreement; and k=0.81–1.00: almost perfect agreement) [24].
Finally, the satisfaction was assessed using an arbitrary quantitative Likert scale containing four possible responses: 1 (most difficult), 2 (difficult), 3 (easy), and 4 (very easy) [25]. The mean and standard deviation for the Likert scale data were calculated and compared between the two arms using Student’s t-test.
Ethics statement. This study received ethics approval from the ethics committee of the Health Public School of Kinshasa’s University. All participants provided written informed consent. No compensation was provided for participating in this study. The study was conducted by the Research, Teaching, and Care Unit of the Faculty of Medicine and Pharmacy of Kisangani’s University. This trial was retrospectively registered in the Pan African Clinical Trial Registry (www.pactr.org) database, ID number PACTR201904546865585.