Investigating interventions to increase uptake of HIV testing and linkage into care or prevention for male partners of pregnant women in antenatal clinics in Blantyre, Malawi: study protocol for a cluster randomised trial

This is a phase II, adaptive, MAMS CRT using ANC day as the unit of randomisation. As a phase II trial the study is intended to investigate efficacy relating to uptake of testing and subsequent HIV services by the male partner, safety outcomes, and to provide an estimate of acceptability to the pregnant woman. The trial will have one interim analysis during which pre-planned adaptations will be made as described below, followed by final analysis at the end of the second stage (two-stage MAMS design). The first stage will have six arms with one SOC and five intervention arms (Fig. 1). At the end of the first stage, a 3-point criteria will be considered by an independent Data Safety and Monitoring Board in order to drop intervention arms. Each intervention versus the SOC yielding p > 0.2 will be considered to be dropped; safety concerns; and costs will guide recommendations to drop or retain an intervention arm at the end of the first stage. The trial will not stop for efficacy at the interim analysis.

Analysis of second stage (end of trial) data will potentially lead to a definitive study (phase III) involving arms that show promise at the end of the second stage. A sample size of 36 ANC days (six per arm) will be required for the first stage (see the ‘Sample size considerations’ section below for justification). In order to control the family wise error rate (FWER) at the specified significance level (α = 0.2) for the five comparisons with the SOC arm, Dunnett’s test [29] will be applied.

A simulation study was set up to compute the overall probability of the trial to find at least one intervention whose efficacy is different compared to the SOC arm at the end of stage 2 (minimal power). The simulation was conducted under the alternative hypothesis with an assumed effect of 29–40% versus 25% in the SOC arm (Table 1). Additional key assumptions included that each ANC day (cluster) would have 20 participants with power set at 80% for each stage (Table 1). The simulation showed that at least two interventions would significantly improve the trial primary outcome (Table 2). These interventions would then be further tested in a potential phase III trial.

The study will recruit participants from Ndirande, Zingwangwa and Bangwe primary health centres (PHC) in urban Blantyre, Malawi. All women attending for antenatal care for the first time at these PHCs and their male partners will be eligible for participation (see Additional file 1). Women and their male partners will be excluded if they received couple or partner HIV testing in the current pregnancy; if either are aged below 18 years of age; if the male partner is reported to be HIV positive by the pregnant woman; if already recruited in this trial; and if not urban Blantyre resident. Malawi has recently (September 2016) implemented the test and treat approach where everyone diagnosed with HIV starts antiretroviral treatment (ART) immediately.

Each ANC day was randomised to any one of the six trial arms using a randomised permuted block design in a ratio of 1:1:1:1:1:1 (Fig. 2). All three PHCs are of comparable size and serving comparable catchment populations, and, therefore, stratification was not deemed necessary. The allocation sequence was generated by an independent statistician using computer-generated random numbers [30]. The file containing the complete randomisation sequence will only be accessible to the independent statistician.

Field workers will enrol women into one of the six study arms on the morning of each ANC day after receiving the randomisation allocation for that ANC day (see Additional file 1). Women will receive study information in a group while in the ANC waiting area followed by one-on-one eligibility assessment and subsequent recruitment. Study information given during group sessions will not reveal details of arm-specific procedures to avoid potential non-participation associated with knowledge of procedures for each arm. All women who show interest at this stage will provide unwritten consent to participate by show of hands followed by written or witnessed thumb-print consent.

In the SOC arm women will receive a personalised letter only addressed to their male partner inviting him to go to the male-friendly clinic (MFC) to have an HIV test, receive HIV care or prevention and pregnancy-related education. The MFC is being implemented by the trial, and will offer men attending confirmatory HIV testing, facilitate linkage to HIV care or voluntary male medical circumcision (VMMC), and pregnancy-related health education.

In all five intervention arms, the woman will receive self-test instructions and two self-test kits to take home. The test kits, test instructions and a personalised letter will be delivered to the male partner by the woman in order to initiate dialogue for him to test and link to the MFC for HIV care or prevention as appropriate.

The five intervention arms differ with respect to financial incentives, participation in a lottery, and phone call reminders received by the male partner. Women in the first intervention arm will only receive the letter and the two self-test kits. In the two fixed financial incentive (FI) arms, male partners who self-test (test) and link into the MFC will receive an equivalent of US$3 or US$10 in the low- and high-FI arms, respectively. In the lottery FI arm, male partners who test and link into the MFC will have a 10% chance of winning US$30. In the final intervention arm, male partners will receive a phone call, through a number given to the study team by the woman at enrolment, to remind him to test and link into the MFC. All FIs will be disbursed as cash through mobile money in the trial in order to safeguard the safety of staff and are conditional on the male partner linking into the MFC.

The primary outcome is the proportion of male partners of ANC attendees who test for HIV and link into HIV care or prevention within 28 days of enrolling the woman (Fig. 3). Thus, the primary outcome is defined as presentation of the male partner at the MFC with a used self-test kit (if in the intervention arm) or undergoing spot HIV testing with a study HIV counsellor within 28 days AND being referred for HIV care if HIV positive or VMMC if HIV negative and uncircumcised. There are four secondary outcomes: the proportion of male partners who test for HIV within 28 days (as reported by the woman); the proportion of women who accept to participate in their allocated trial arm; risk of serious adverse events (SAEs) in men and women in the study; and the total cost of implementing each trial arm. All outcomes will be analysed at cluster level (see the ‘Statistical analysis’ section).

All male partners who present at the MFC in the SOC arm will be offered a single finger-prick HIV test with Determine 1/2™ as per Malawi national testing algorithm. An HTS counsellor will re-read a used self-test kit if the participant returns one as evidence of self-testing in the intervention arms. Participants who return with unused self-test kits or without self-test kits will be requested to self-test in the presence of the counsellor. All HIV results will be recorded on a data form followed by confirmation of HIV-positive results in parallel using Determine 1/2™ and Uni-Gold, with facilitated linkage to HIV care. All men who test HIV negative and report to be uncircumcised will be offered VMMC to be conducted by Population Services Internal (PSI). Thus, measurement of primary outcome includes evidence of an HIV test, confirmatory testing, and referral to HIV care or VMMC as appropriate within 28 days of the woman being recruited.

The secondary outcome of HIV testing among male partners will also be measured though proxy reporting by the woman using audio computer-assisted self-interview (ACASI) during her next ANC visit 4 weeks later. Participation in the allocated trial arm will be measured by computing the proportion of women who accept to participate after receiving trial-arm-specific information using the denominator of the total number of women who are eligible. All women will be asked to report any adverse events through ACASI at their next ANC visit while men who present to the MFC will be asked to report any adverse events. A costing tool validated in urban Blantyre [31] will be used to capture the costs associated with providing the service in each trial arm. The cost and outcome data will be used to estimate the cost per male partner tested for HIV, and cost per HIV-positive male identified through all SOC and intervention arms.

A modification of the formula for sample size calculation for a MAMS design for binary outcomes [32] was made based on the methodology for CRTs [33] to identify each stage of the trial. We assume that each ANC day will have at least 40 women attending for the first time, 90% will satisfy the eligibility criteria and at least 60% will consent to participate, so having a cluster size of at least 21. We also assume that in the SOC arm 25% of men will satisfy the definition of the primary outcome [34]. For the first stage, six ANC days per arm (36 days in total) would be needed to detect an absolute difference of 15% in linkage compared to 25% in the SOC arm using a FWER of 0.2 with 80% pair-wise power and a coefficient of variation (k) of 0.10 [21]. In general, under the stated assumptions the trial has 80% chance of detecting a 1.6-fold increase in testing and linkage within 28 days compared to SOC at 5% significance level. Sample size for the second stage will be re-calculated based on empirical estimates at interim analysis with FWER of 0.1 and 80% power.

Very little clustering within ANC days is expected, hence we have assumed that k = 0.10 (intraclass correlation coefficient = 0.003). The simulation study assumed that k = 0.2 and differs with the final design in this respect resulting in six and seven clusters per arm in the final design and the simulation, respectively. A larger than conventional (0.05) FWER of 0.2 and 0.1 for stage 1 and stage 2, respectively, may lead to erroneously taking forward an ineffective intervention. However, given that this is a phase II trial this is not a major concern as it guards against dropping interventions that may otherwise prove to be effective in a larger, phase III trial. A more conventional pair-wise power of 0.8 was chosen to ensure that there is a high chance of taking forward most of the efficacious interventions from stage 1.

Analyses will be done in R [35] and Stata 14.0 (Stata Corp., TX, USA). Baseline characteristics will be computed as proportions or median (interquartile range (IQR)), as appropriate, by arm in each of the two stages of the trial. Any variables that show imbalances will be adjusted [33] for when analysing the trial outcomes at the end of the second stage. We will assume that the two stages of the trial are independent [26] and will proceed to carry out a test of the null hypothesis of no difference in effectiveness of each intervention compared to the SOC. We will do this by analysing data from the first stage first followed by interim decisions to drop arms; then we will conduct and analyse data from the second stage (no overlap of participants from the first stage). Analysis of the whole trial will then be based on combined p values from both the first stage and the second stage using the weighted inverse normal (WIN) method [36] for arms that are not dropped at interim. A weighted average of the log (risk ratio (RR)) will be computed for the whole trial using estimates from each trial stage. All analyses will be by intention-to-treat taking as the denominator the number of women who were eligible and take into account the clustered design.

Given the small number (six) of clusters per arm in the first stage, analysis will be by cluster-level summaries using mean of proportion of male partners per clinic day who link to care or prevention in each arm [33]. The proportion of male partners who link into care or prevention will be computed per clinic day for each arm with number of men achieving the primary outcome and the number of women eligible and recruited in ANC on enrolment day as denominator. A log transformation of the clinic day proportions will be applied if a positive skew is observed [33]. The geometric mean of clinic day proportions in each of the five intervention arms will be compared to the SOC arm using an unpaired t test [33]. An estimate of the RR and a 95% CI will also be computed for each comparison by dividing the geometric mean of proportions in each intervention arm and the geometric mean of proportions in the SOC arm [33].

This analysis involves more than two comparisons with a single control arm which can lead to higher than the specified FWER or significance level. Therefore, Dunnett’s test [29] will be applied to the t-statistics generated from the unpaired t test to control the stage-wise FWER. Final decision-making at interim analysis will compare the Dunnett-corrected p values to stage 1 FWER of 0.2. The first of the three-part criteria for dropping trial arms will then be considered after examining final p values at the end of the first stage (interim analysis). Although sample size will be re-calculated at the end of stage 1, the total number of clinic days per arm is still presumed to be small for stage 2. Since the two stages are assumed to be independent, cluster-level summaries approach analogous to stage 1 analysis will also be followed in stage 2 comparing intervention arms that proceed to stage 2 with the SOC arm. A detailed analysis plan will be developed to guide analysis of the trial.

Interim analysis at the end of stage 1 will assess whether any of the five intervention arms should be dropped as recommended by an independent Data Monitoring and Safety Board (DSMB) based on a three-part criteria. First, an arm whose statistical comparison to the SOC arm yields a p value > 0.2 will be considered for dropping for futility. Second, any intervention arm with a high incidence of SAEs, i.e. grades 3, 4 or 5 (Table 3) compared to SOC will be considered for dropping. It is at the discretion of the DSMB to decide, based on absolute number of SAEs in each intervention trial arm, whether they are high or not. Such an observation and recommendation will then be shared with the investigators who will make the final decision. Thirdly, an arm may be maintained after taking into account the costs associated with providing the service in light of the p value from statistical analysis. For this cost analysis, we will provide the DSMB estimates of the incremental cost per male partner tested, and incremental cost per HIV-positive male identified through the intervention arms in comparison to the SOC arm. The investigators will access the first-stage data only after the last follow-up visit for participants has occurred in order to perform interim analysis. The CRT extension to the Consolidated Standards of Reporting Trials (CONSORT) [37] will be followed when reporting the data.