The HOSENG (HOme-based SElf-testiNG) trial will be conducted in the districts of Butha-Buthe and Mokhotlong, in Northern
Lesotho, Southern Africa, in the catchment areas of 22 health facilities. Both districts
are characterized by mostly rural settings with an estimated population of 220,000,
mainly subsistence farmers and mine workers as well as construction or domestic labourers
who work in neighbouring South Africa. Each district has only one mid-size town: Buthe-Buthe
with ca. 25,000 inhabitants, and Mokhotlong with ca. 10,000 inhabitants. The remaining
population lives in villages scattered over a mountainous area of 5,842 km2. According to the recent household-based national survey from 2016-2017 the adult
HIV prevalence is 17.8% in Butha-Buthe and 26.1% in Mokhotlong.21
The HOSENG trial is a cluster-randomized controlled, superiority trial in a resource-limited
setting. The rationale for a cluster-randomized design at village level is a) the
reliance of the trial on the village health workers (VHWs) and b) the high risk of cross-contamination between the study arms if randomization
would be done at individual or household level. The
HOSENG trial with its home-based HIV testing campaign provides a recruitment platform
for another trial, the VIBRA (Village-Based Refill of ART) trial, and thus they are based on the same cluster-randomization and run in parallel.
Together, HOSENG and VIBRA constitute the GET ON (GETting tOwards Ninety) research project. To ensure a balance in the exposure to the HOSENG intervention in
the two VIBRA arms, the clusters (villages) will be randomized into 4 potential groups
in a 1:1:1:1 allocation.
Cluster sampling and randomization
Figure 1 summarizes the cluster sampling and randomization process. A list of all
villages with their corresponding VHWs in the study districts was provided by the
Ministry of Health and the local District Health Management Teams. Two local members
of the research team cross-checked the village list for accuracy. They defined the
village size (≥30 versus <30 households) and access to the nearest health facility
(easy versus hard to reach, defined by needing to cross a mountain or river or >10
km away from health facility) by contacting the relevant VHW coordinators. Each village
was considered a cluster, except villages which do not have their own VHW and therefore
form one cluster with neighbouring villages served by the same VHW. It is not feasible
to visit all clusters (358 villages in Butha-Buthe and 290 villages in Mokhotlong)
nor to include >2 VHWs per cluster. Therefore, a random sample of 180 clusters, stratified
by district, village size and access, was taken. If a cluster had more than two VHWs,
randomly two VHWs were selected. After excluding clusters who did not meet the eligibility
criteria (see next section), the remaining 159 clusters were randomized into the 4
groups in a 1:1:1:1 allocation ratio with block sizes of 4, stratified by the same
stratification factors. All random sampling processes as well as the randomization
were performed by an independent statistician. The first 103 clusters were provided
to the research team and will all be enrolled for the campaign. If needed, further
clusters will be released from the randomization list.
Table 1 outlines the eligibility criteria for clusters and households. No individual
eligibility criteria apply for the HOSENG trial as HIV testing coverage is a population-level
outcome including every individual in the surveyed area.
Before randomization, all VHWs from the 175 randomly selected villages attend a one-day
refresher training focusing on HIV testing and counseling (HTC). Every VHW takes a
short pre- and post-training assessment. The assessment includes questions about a)
HIV knowledge and HIV stigma using a validated questionnaire22, b) HIV testing knowledge, and c) one open question (“What are important qualities
of a counselor?”). The assessment will be used to evaluate the baseline HIV/AIDS-related
knowledge and stigma of the involved VHWs and to evaluate their eligibility for participation
in the trial (Table 1).
Prior to trial start, the research team met with all relevant village chief councils
in order to inform them about the trial and to obtain their consent.
Two to three specifically trained teams, each consisting of 5-6 lay-counsellors, 1
campaign organizer and 1 supervising study nurse will visit all households in selected
villages, going from door to door. The teams will propose HTC and multi-disease screenings
and prevention. The population of the selected areas are informed about the campaign
beforehand by their village chief. Specifically, the campaign team will arrive in
the village in the morning, meet with the village chief and then start systematically
visiting all households in the pre-defined area by going from door to door. At the
household, the teams will proceed as follows:
The counselor will introduce him-/herself and the purpose of the testing/screening
campaign
The counselor will ask the head of household or their representative (aged 18 years
and older) for written informed consent to the testing/screening campaign, and to
obtaining data about present and absent household members for study purpose
If the head of household or the representative refuses his/her household participation,
the team will leave the house, record the refusal reason, and proceed to the next
household
The counselor will assess the total number of present and absent household members
Definition of household member: 1) is acknowledged by the household head or the representative
as part of the household and 2) sleeps in the household regularly (at least once a
month) and 3) if absent during the campaign: returns to the household no later than
three months after the date of home-visit
The counselor will provide information about HIV and testing, prevention aspects,
and the other disease screenings (see below)
The counselor will assess the HIV status of all household members and screen the patient’s
health booklets (“bukana”)
Household members who are eligible for and consent to testing (by filling in the Lesotho
national informed consent form for HIV testing) undergo HIV testing by the counselor
according to specific cluster arm allocation procedure and according to national HIV
testing guidelines.20 For absent or declining household members, the specific procedures differs between
the two cluster arms, and are described in detail below in the section on intervention
clusters. Figure 2 presents all possible HIV testing scenarios during the campaign.
Once the HIV test is done, the counselor documents the result in the patient’s health
booklet and provides post-test counseling. If the HIV test is confirmed positive,
the counselor contacts the study nurse, who then provides further counseling and assesses
the participant for enrolment into the inter-linked follow-up study (VIBRA trial).
The HIV testing campaign is combined with additional multi-disease screening. The
campaign screens for tuberculosis (TB) according to national guidelines using the clinical symptom screening
tool.23 If TB is suspected, one sputum sample is collected on the spot, transported to the
health facility the same day by the campaign team for testing with GeneXpert. Another
sputum bottle for collecting a morning sample is left behind, the individual is instructed
how to use it and a follow-up plan for collection (i.e. collection by VHW) is agreed
upon. Further campaign services include alcohol abuse screening using the CAGE questionnaire24, information and referral of males aged 15-50 years old (HIV-negative and -positive
due to stigma reasons) for voluntary medical male circumcision (VMMC) according to
WHO recommendations25, and promotion and provision of male condoms. Figure 3 presents the overview of the HIV and multi-disease campaign and its algorithm.
Figure 4 summarizes the procedures in the HOSENG clusters and Figure 5 the SPIRIT
flow diagram. In the intervention clusters, the campaign team will leave an oral HIVST
kit (OraQuick® ADVANCE HIV I/II, second generation serology assay with a sensitivity
of >93%, a specificity of >9926–29) for every household member 12 years or older who is absent or declined HIV testing
on the day of the campaign. We chose 12 years as threshold because adolescents are
an important risk group and this is the legal age for providing HIV testing consent
in Lesotho.23 The household can refuse to have an oral HIVST left behind for their absent household
members. The oral HIVST kit is prepacked, includes a written and pictoral instruction
for use in the local language, Sesotho, and the package is labelled with a written
request to consult the VHW within 2 weeks after use of the test – irrespective of
the result. In cases where more than one trained VHW serves the village, the household
will be asked for the VHW of choice. The team will label the kit with the name of
the absent member before dispensation.
One household member – the one with the closest relation to the absent person(s) –
will be tested and trained using the oral HIVST. All other present household members are tested using the standard blood-based point-of-care
HIV test according to Lesotho national guidelines.20
The VHWs in the intervention custers will receive a list of all household members
for whom an oral HIVST was dispensed. The VHW will visit all households 2-4 weeks
after the campaign to collect the oral HIVST in case the oral HIVST is not returned
to him/her. If an oral HIVST is reactive, the VHW will either provide further blood-based testing
on the spot or organize referral to the nearby health facility for confirmatory testing.
All VHWs from intervention clusters will receive a second extensive training session
about oral HIVST, handling disclosure and stigma, and data entering on paper-based
study forms and the patient’s health booklet.
Following standard of care during home-based HIV testing, every absent household member
or those declining HIV testing will be encouraged to get an HIV test done by the VHW
or the nearby health facility. The research team will screen the registers at the
health facilities in order to assess if these individuals came for testing in the
timeframe of the primary endpoint window.
The primary endpoint is HIV testing coverage among individuals aged 12 years or older
in the surveyed area within 120 days after the home visit, defined as the proportion
of all individuals 12 years or older living in a household of the surveyed area with
a confirmed HIV test result.
We define a confirmed HIV-negative result as either being tested HIV-negative as per
the algorithm defined in Figure 2, or being tested HIV-negative within the last 4
weeks with proof of documentation (i.e. documentation in patient’s health booklets).
We define a confirmed HIV-positive result as a) tested HIV-positive as per the algorithm
defined in Figure 2, or b) being tested HIV-positive but not yet on ART with proof
of documentation, or c) being on ART with proof of documentation.
The rationale for the time-point of 120 days is that we leave oral HIVST kits at the
households for all absent household members aged 12 years or older who return within
a maximum of 3 months (see definition of household member above in section Procedure).
An interval of 120 days after home visit allows sufficient time for absent members
to return to their households, conduct self-testing, and be followed up by the VHW.
The secondary endpoints are outlined in Table 2.
Additional research within the project
For the entire GET ON project we will collect cost data, see more details in the VIBRA
trial protocol published elsewhere. Specifically for HOSENG trial, first, direct costs
of the intervention will be assessed: Staff costs (campaign team, VHWs, clinic staff),
personnel training costs (VHWs), cost of equipment (HIV tests, consumables, logistics),
as well as non-medical costs to the participant (i.e. cost of transportation to ART
service). These data will provide the cost per participant achieving the primary endpoint
within 120 days in each cluster arm (‘per participant tested cost’). Secondly, a cost-effectiveness
analysis will be performed with respect to the primary endpoint. The cost-effectiveness
ratio will be presented as incremental cost per additional confirmed HIV test result.
Data to assess patient level costs will be collected from a randomly selected sub-sample
of study participants from each cluster arm. Costs will be reported in local currencies
and U.S. dollar and International dollar. The robustness of results will be tested
with probabilistic sensitivity analysis and deriving cost-effectiveness acceptability
curves to capture the uncertainty around the probability that the intervention is
below the relevant cost-effectiveness thresholds.
A nested study (ADORE study: “ADolescent ORal sElf-testing”) will explore the effectiveness
and acceptability of oral HIVST among adolescents and young adults with quantitative
methods (testing coverage among adolescent and young adults, defined as the proportion
of all 12-24 years old individuals living in a household of the surveyed area with
a confirmed HIV test result) and qualitative methods (case-control study). Cases are
those who refused testing through oral HIVST and controls are those who accepted testing
through oral HIVST. We plan to conduct at least 10 interviews per group, stratified
by two pre-defined factors (male vs female; age 12-15 vs age 16-24), following the
concept of saturation. Data will be collected by a trained study member, who was part
of the HIV testing campaign, using a piloted interview questionnaire (KoboToolbox;
www.kobotoolbox.org), conducted in the local language (Sesotho). Qualitative data will be recorded, transcribed,
translated into English and coded and analyzed using the Framework Method30.
Data collection and management
The campaign team will capture all data collected during the campaign using a tablet-based
application and platform (MACRO, Elsevier). The randomization assignment of the villages
is pre-loaded into the program and a unique household identifier is automatically
generated. Before leaving the household, the completed questionnaire will be checked
for mistakes and completeness. Data from the tablet devices will be uploaded regularly
via secure electronic transfer and stored on a secure server at the Swiss Tropical
and Public Health Institute (Swiss TPH). After the follow-up period, all confidential
information of the study participants (i.e. names) will be deleted from the database
and only the anonymous study-ID will be kept. The informed consent forms will be stored
in a secure way in the headquarter of the study center (SolidarMed Office in Butha-Buthe,
Lesotho). Participant files will be maintained in storage for a period of at least
10 years after completion of the trial.
The VHWs enter data for following up the dispensed oral HIVST kits on standardized
paper study forms (Case Report Forms), that act as source documents. These data will
be collected regularly by the study team and entered into the above mentioned platform.
A study data manager will monitor data quality and completeness on a weekly basis.
Queries about the data will be sent to the local principal investigators for follow-up
and correction, as needed. Data integrity checks will be written into the database
to limit missing fields or the entry of incorrect data. The type of activity that
an individual user in the online study database may undertake, will be regulated by
the privileges associated with his/her user identification code and password.
The sample size is driven by the VIBRA trial. We assume to achieve the VIBRA trial
sample size (minimum 262 HIV-positive individuals not on ART) by testing about 10’000
individuals in 100 villages, based on a previous trial.31 According to a previous home-based HIV testing campaign32, we estimate an HIV testing coverage rate of individuals 12 years or older in rural
villages of 63%. We consider a 15% increase in coverage as relevant from a policy
perspective. Using an intra-cluster correlation (ICC) of 0.028, with a sample of 100
clusters, we will require a sample size of 450 in order to have >90% power to detect
an increase in coverage rates to 78% or higher. This corresponds to an increase of
1.17 (variance inflation factor VIF) from the 386 individuals required in the absence
of clustering. If we vary the ICC for villages and include an additional ICC for clustering
at the household level33, the maximum VIF would be 8.3 corresponding to a maximum sample size of 3204 needed
to ensure 90% power to detect a 15% difference, well below the expected 10,000 individuals
to be included. Table 3 displays the sample size calculations considering all relevant
factors under different scenarios.
Analyses will be reported following CONSORT guidelines for cluster-randomized trials.34 Clusters will be set as the unit of randomization (stratified by district, size of
village, and village access to the nearest health facility), whereas individuals are
set as the unit of analysis. An intention-to-treat set will be used, i.e. all study
participants will be evaluated according to arm assignment at randomization. The primary
analysis will use multi-level logistic regression models including village and household
as random effects to assess the difference between HIV testing coverage in the intervention
versus control arm. These models will be adjusted for the pre-specified randomization
stratification factors and relevant baseline factors (age groups, gender, education
status, employment status, HIV testing history) that may be randomly unbalanced between
intervention and control clusters.35,36
Baseline characteristics will be presented according to randomized groups; no formal
testing will be performed. Categorical variables will be described with absolute and
relative frequencies and continuous variables with medians and interquartile ranges.
As with the primary analysis, secondary endpoints will be analyzed with multi-level
logistic regression models including village and household as random effects. All
results will be presented as odds ratios and 95% confidence intervals. The potential
effect modification of sociodemographic determinants (age groups, gender, education
status, employment status, HIV testing history) on all endpoints will be assessed
by including interaction terms in the models. If the Wald test for the interaction
terms are significant, intervention effects will be presented separately by the levels
of these factors. Sensitivity analyses will be conducted in order to provide evidence
that the result seen from the primary analysis are robust. In order to assess the
reliability of the model fit, we will perform a quadrature check. In case of unreliable
model fit, we will use generalized estimating equations to fit our model which will
provide population-averaged odds ratios and 95% confidence intervals. However, this
model would not allow inclusion of more than one random effect although including
the highest level of clustering is suggested to be sufficient.37 All analyses will be done using Stata (version 14, Stata Corporation, Austin/Texas,
USA). For all tests, we will use 2-sided p-values with alpha 0.05 level of significance.
Monitoring, auditing, and data safety and monitoring board
At least one external monitoring visit will assess adherence to the approved trial
protocol, accuracy of completed study forms, and the electronic dataset. The Principal
Investigator agrees to allow inspectors from regulatory agencies to review records
and will assist the inspectors in their duties, if requested.
The HOSENG trial represents implementation research and the oral HIVST is a well-established
diagnostic tool. Thus, we do not expect serious adverse effects (SAE) on patient’s
health from this intervention. However, for the purpose of this trial, we will try
to capture the following SAEs: a) Death due to any reason (especially within 30 days
of a positive HIVST results), b) Hospitalisation due to self-inflicted injuries within
30 days of a positive HIVST results, and c) Hospitalisation resulting from violent
assault by others (intimate partner violence, assault by family or community members)
within 30 days of a positive HIVST result. The campaign team members all have several
years of experience in HTC and received an additional study-specific one-week training
in order to handle adverse events related to testing stigma. A separate, detailed
safety monitoring plan has been developed to handle these SAEs in line with Swiss
and Basotho ethics regulations. It is not planned to establish a data safety and monitoring
board.