Study design and participants

The PEBRA trial was a cluster-randomized, open label, pragmatic clinical trial conducted at 20 rural nurse-led health facilities (clusters) in Butha-Buthe, Leribe, and Mokhotlong districts of Lesotho, Southern Africa. The 20 health facilities serve a rural population in a mountainous area with poor infrastructure. We decided to conduct this trial only across nurse-led clinics in rural areas because these are the most common health facilities across Lesotho, they provide standardized services (e.g., important for the control arm comparison) and the population they serve reports generally worse HIV outcomes compared to the population served at the district hospitals [12]. Recruitment for the trial lasted from November 6, 2019 until February 4, 2020. A detailed study protocol has been published previously [13].

Eligible clusters were public or missionary nurse-led clinics (not hospitals) offering ART services to a rural population, situated in an area with stable cell phone signal, willing to participate, and with a PE who passed the study-specific training assessment. Young people were eligible for inclusion if they were living with HIV, registered for HIV care at one of the eligible clusters, aged 15 to 24 years, taking ART, and able to provide informed consent.

Randomization and blinding

Randomization events with all health facilities and the District Health Management Team were conducted in each district between October 25 and 30, 2019. At these events, health facility representatives drew opaque, sealed, equally sized envelopes containing the group allocation (control or intervention) from a Mokorotlo (traditional Lesotho hat), and disclosure took place only once all facilities had drawn their envelope. Additionally, to minimize potential selection bias, the sequence of drawing was randomly selected in advance by an independent person drawing from a second pile of opaque, sealed envelopes containing the names of the facility. Randomization was stratified by district (Butha-Buthe versus Leribe versus Mokhotlong), using 1 block per strata. Participating clinics were assigned (1:1) to either offer standard of care or the PEBRA model. This was an open-label trial; however, laboratory staff who assessed the primary endpoint were blinded. We used non-site-specific study ID numbers on all laboratory and data collection forms to maintain masking.

Procedures

During the recruitment period, the PE actively screened all young people attending their health facility for inclusion on a rolling basis and obtained written informed consent. The PEs recruited from Monday to Saturday on a strictly rolling basis and followed a screening log file. Recruitment happened at all 20 health facilities concurrently. Illiterate participants provided a thumbprint and chose a literate witness (independent of the trial and chosen by the participant) to co-sign the form. In order to minimize selection bias, the ethics committees agreed to waive parental consent for the 15 to 17 years old study participants. If eligible, the PE administered a questionnaire that included sociodemographic and socioeconomic data, medical history, adherence to ART, quality of life (QoL), HIV/AIDS-related knowledge, and satisfaction with care. If no viral load (VL) within the previous 12 months was available, the participant was sent to the nurse for VL measurement.

Intervention and control

At health facilities randomized to the intervention arm, the participants were offered the PEBRA model. PEBRA is a DSD model coordinated by the PE, delivered using an eHealth application (PEBRApp) and based on regular service preference assessments in 3 domains: (1) medication pick-up; (2) SMS notifications; and (3) psychosocial support. An overview of the PEBRA model is presented in Fig 1 and details to each intervention component in S1 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Description of the PEBRA model and clusters. ART, antiretroviral therapy; PEBRA, Peer Educator-Based Refill of ART.

https://doi.org/10.1371/journal.pmed.1004150.g001

First, the PE explained all options within each domain to the participant (Fig 1). Second, the participant picked his/her preferred choice within each domain. Third, the PE systematically assessed the feasibility of the options chosen, as not all options are available to everyone all the time, e.g., no nearby Village Health Worker available who could dispense ART, or no community youth club established in the participants’ community, or home-delivery by the PE not feasible. Finally, the compromise between preference and feasibility was delivered. This process was guided by the PEBRApp, an android-based application, installed on the tablet of the PEs.

The PEBRA model and the PEBRApp were developed in close collaboration with PEs, young people living with HIV, youth activists, clinicians, district Ministry of Health authorities, Sentebale youth leaders, SolidarMed and Swiss Tropical and Public Health Institute research staff as well as local and international mobile application developers. During a series of workshops, these stakeholders defined the current main challenges in adolescent HIV care in Lesotho, possible strategies within the PEBRA model to overcome these challenges building on existing resources at the study health facilities, the role of the PE within the PEBRA model, and the scope and design of the PEBRApp. Due to limited access to smartphones in rural Lesotho, the consensus from the formative workshop series was to build an application for the PEs only, with a communication channel between PE and participants based on SMS technology. As such, the PEBRApp helped the PE not only to collect data, regularly assess the preferences of their peers, and to keep track of the ART refill and next assessment dates, but also ensured regular contact between the PE and the participant. More details and screenshots about the PEBRApp are provided in the published study protocol [13] and the code is open-source available on github (https://github.com/chrisly-bear/PEBRApp). Any selected SMS notifications were sent out automatically according to chosen frequency and content and always included a free-of-charge call-back option to the PE’s phone number. These service preferences were assessed at enrolment and thereafter every month for participants with unsuppressed VL (≥1,000 copies/mL) and every 3 months for participants with suppressed VL.

Before starting the trial, the PEs completed a 1-week training camp that covered obtaining informed consent, administering study-specific questionnaires, the use of the PEBRApp, and how to deliver PEBRA model. The PEs were recruited from Sentebale’s longstanding PE program in collaboration with the Ministry of Health and the SolidarMed study coordinator, received a monthly stipend and were closely supervised by the attached health facility staff and the study coordinator.

At health facilities randomized to the control arm, the study participants received the standard of care offered at nurse-led rural health facilities in Lesotho, i.e., ART refill at the clinic and nurse-led support (Fig 1).

Data collection

In intervention arm, data were collected by the PE and in control arm by trained young study staff members. In both arms, data were entered into the password-protected PEBRApp. However, the PEBRApp in control arm was limited to baseline and follow-up data collection and thus did not allow any DSD. The randomization assignment of the health facilities was preloaded into the PEBRApp, and unique individual identifiers were automatically generated. The PEBRApp was connected to the routine VL laboratory platform that includes all VL measurements from the 2 study districts. An automatic synchronization ensured regular download of VL measurements of study participants as well as anonymized data upload to a password-protected, secured database that could be accessed by the data manager. The Satisfaction with Care questionnaire including satisfaction with the PEBRA model was administered by the study staff, all other questionnaires by the PEs. Data integrity checks were programmed into the PEBRApp and the data manager monitored the uploaded data on a regular basis. Data closure was on April 7, 2021.

Outcomes

The primary outcome was viral suppression at 12 months, defined as the proportion of all participants in care with a VL below 20 copies/mL at 12 months (range 9 to 15 months) after enrolment. This also included participants who transferred out to another health facility and had a documented VL from the new facility. Blood draw for VL measurement for all study participants was conducted at the clinics using full blood, and the analysis was performed at the corresponding laboratories of the study districts using COBAS TaqMan HIV-1 Test, v2.0 (Roche Diagnostics).

The secondary endpoints were viral suppression using the by then valid WHO threshold of less than 1,000 copies/mL at 12 months, engagement in care at 6 (range 5 to 8) and 12 months, transfer out to another health facility, loss to follow-up (LTFU) (defined as more than 2 months late for a scheduled visit or medication pick-up and no information found about the participant), all-cause mortality, perfect self-reported adherence to ART in the past month (defined as no missed doses according to self-reporting), physical and mental QoL measured using the Short Form 12 (SF-12) questionnaire, satisfaction with care (based on a setting-validated HIV service satisfaction questionnaire [14]), and satisfaction with their PE (in intervention arm only). Details about the endpoints are provided in the published study protocol [13].

Statistical analysis

According to routine cohort data from the study districts [12], we estimated to recruit 10 to 20 eligible participants per site. For the control arm, based on the same routine cohort data, we assumed a viral suppression rate of 70% in Butha-Buthe and Leribe districts, but significantly lower in Mokhotlong district. We hypothesized that the PEBRA model would increase the proportion with viral suppression by 20 percentage points in the intervention arm. Using a power of 90%, assuming a type 1 error of 0.05 and a conservative intracluster correlation coefficient (ICC) of 0.05 (design effect of 1.7) based on similar studies [15,16], a sample size of 300 participants in 20 clusters (10 per arm) was needed.

The analysis followed an intention-to-treat approach, including all enrolled participants as randomized. Health facilities (clusters) were the unit of randomization, whereas individuals were the unit of analysis, with viral suppression as a binary outcome. All participants missing their blood draw or having invalid VL results were classified as not meeting the viral suppression endpoints. For the analysis of the primary endpoint, we used a multilevel logistic regression model including cluster (health facility) as a random effect, and arm allocation and the randomization stratification factor (district) as a fixed effect. According to the procedure outlined in our statistical analysis plan, we first assessed all baseline tables for clinically important random imbalanced factors. We noted a substantial random imbalance in sex and thus decided to adjust for it in all analyses.

For the secondary endpoints, we used the complete case set. The secondary endpoints of alternative viral suppression threshold, engagement in care, transfer out, LTFU, and mortality were analysed using a logistic regression model with the same explanatory variables as the primary analysis model. To compare perfect adherence between the 2 arms, we used the same model, but additionally adjusted for self-reported perfect adherence at enrolment and the treatment taken at the time point under consideration (dolutegravir (DTG) versus non-dolutegravir). Differences of QoL between the 2 arms were assessed using multilevel mixed effect linear regression models, adjusted for sex and additionally the respective QoL scores at enrolment. Similarly, satisfaction with care was assessed using a multilevel mixed effect logistic regression model with outcome “very satisfied with care” at 12 months, adjusted for sex and additionally the respective satisfaction at baseline. From the logistic regression models, results are presented as odds ratios (ORs) with 95% confidence interval (CI) and from the linear regression model as beta coefficients with 95% CI. We performed prespecified sensitivity analyses for the primary endpoint: adjustment for baseline VL, using a wider primary endpoint visit window (9 to 18 months), and restricting to the individual per protocol set (participants who attended both the 6-month and 12-month study visit). For the primary outcome, we also assessed effect modification by prespecified variables (age groups, sex, marital status, occupational status, time of ART exposure, DTG at time of endpoint). All analyses were done using R, version 4.0.3 (2020-10-10).

Ethics statement

This trial was approved by the National Health Research and Ethics Committee of the Ministry of Health of Lesotho (118–2019; June 3, 2019) and the ethics committee in Switzerland (Ethikkommission Nordwest- und Zentralschweiz; 2019–00480; June 14, 2019). The trained study staff obtained the individual written informed consent from the participants before inclusion into the PEBRA trial. To minimize selection bias, the ethics committees agreed to waive parental consent for the 15 to 17 years old study participants as outlined in the approved PEBRA study protocol [13]. Illiterate study participants provided a thumbprint and a literate witness (independent to the trial and chosen by the participant) co-signed the form. The informed consent was provided in the local language, Sesotho, and the participant received a copy of the consent form. Participants were not compensated for participation. The PEBRA trial is registered with ClinicalTrials.gov (https://clinicaltrials.gov/ct2/show/NCT03969030; prospectively registered on May 31, 2019) and reported as per the CONSORT extension for Cluster Trials guidelines (S1 CONSORT Checklist).