Menu
Menu
Menu
Menu
Mother-to-child transmission of HIV remains a significant concern in Africa despite earlier progress. Early infant diagnosis (EID) of HIV is crucial to reduce mortality among infected infants through early treatment initiation. However, a large proportion of HIV-exposed infants are still not tested in Kenya. Our objective was to investigate whether weekly interactive text-messages improved prevention of mother-to-child transmission (PMTCT) of HIV care outcomes including EID HIV testing. This multicentre, parallel-group, randomised, open-label trial included six antenatal care clinics across western Kenya. Pregnant women living with HIV, aged 18 years or older, with mobile phone access, were randomised in a 1:1 ratio to weekly text messages that continued until 24 months postpartum, asking “How are you?” (“Mambo?”) to which they were asked to respond within 48 h, or a control group. Healthcare workers contacted participants reporting problems and non-responders by phone. Participants in both groups received routine PMTCT care. The prespecified secondary outcome reported in this paper is EID HIV testing by eight weeks of age (blinded outcome assessment). Final 24-months trial results will be published separately. We estimated risk ratios using Poisson regression with robust standard errors. Between June 2015–July 2016, we screened 735 pregnant women, of whom 600 were enrolled: 299 were allocated to the intervention and 301 to the control group. By eight weeks of age, the uptake of EID HIV testing out of recorded live births was 85.5% in the intervention and 84.7% in the control group (71.2% vs. 71.8% of participants randomised, including miscarriages, stillbirths, etc.). The intention-to-treat risk ratio was 0.99; 95% CI: 0.90–1.10; p = 0.89. The proportion of infants diagnosed with HIV was 0.8% in the intervention and 1.2% in the control group. No adverse events were reported. We found no evidence to support that the WelTel intervention improved EID HIV testing. A higher uptake of EID testing than expected in both groups may be a result of lower barriers to EID testing and improved PMTCT care in western Kenya, including the broader standard use of mobile phone communication between healthcare workers and patients. (ISRCTN No. 98818734. Funded by the European-Developing Countries Clinical Trial Partnership and others).
WelTel PMTCT was a multicentre parallel-group randomised controlled open-label trial carried out at six ANC clinics in western Kenya. The primary trial endpoint was retention in PMTCT care up to 24 months postpartum among women living with HIV and their HIV-exposed infants. This paper is the first trial report on a prespecified secondary outcome to investigate the first step in the postpartum PMTCT cascade, EID HIV testing by eight weeks of age29. The ANC clinics in the study are government-run and part of the Academic Model Providing Access to Healthcare (AMPATH) program. Clinics included Moi Teaching and Referral Hospital (MTRH), Uasin Gishu District Hospital (UGDH), Huruma sub-County Hospital, Kitale County Referral Hospital, Chulaimbo sub-County Hospital, and the Matayos Health Centre (Fig. 1). MTRH, UGDH and Huruma sub-County Hospital are located in Eldoret, the fifth most populated town in Kenya. The clinics in Kitale, Chulaimbo and Matayos are located in different parts of western Kenya. Overall, the ANC clinics involved in the trial were spread over a large geographical area and serve urban as well as rural populations. All participants provided informed consent and ethics approval was obtained from the Institutional Research and Ethics Committee at Moi University, Kenya (FAN: IREC 1292) and the Regional Ethics Committee, Stockholm, Sweden (2018/742-31/1). All trial procedures were performed in accordance with relevant guidelines and regulations. The study was registered at the international standard randomised controlled trial number (ISRCTN) registry (ISRCTN98818734, registration date: 09/12/2014) and the trial protocol has been published elsewhere24. Location of study sites in western Kenya. Between June 25th, 2015 and July 5th, 2016, pregnant women living with HIV aged ≥ 18 years and presenting to their first ANC visit in their current pregnancy were invited to participate in the trial. HIV diagnosis was based on two repeated Determine™ or Colloidal Gold blood tests, or referral from a HIV comprehensive care clinic for those with known HIV infection. To be eligible, women had to have access to a mobile phone, be able to respond or have someone in close contact who could respond to the text messages, and be a resident of the ANC clinic catchment area. Women who planned to relocate from the ANC clinic catchment area, and who were not willing to be followed up until 24 months postpartum were excluded. Infants born to trial participants were also included in follow-up (Fig. 2). Flow chart of screening, enrolment, reasons for non-participation and randomisation of participants. aOne participant withdrew from receiving text messages before the infant was born, after having received 31 messages. The participant agreed to continued follow-up and was included in intention-to-treat analysis as a participant of the intervention group. bWelTel intervention was stopped in case of miscarriage, stillbirth, infant death, or maternal death. Participants were interviewed face-to-face by a trained research assistant in Kiswahili or English upon trial enrolment. Participants were thereafter randomly allocated by the research assistant to the intervention or control group using a 1:1 allocation ratio. Randomisation was performed at each ANC clinic separately using individual, opaque, sealed envelopes. To ensure balance of participants’ baseline characteristics between the two study groups, a computer-generated non-disclosed permuted-block randomisation scheme with block sizes equal to four was generated by an independent statistician at the Karolinska Institute, Stockholm, Sweden. The randomisation scheme was generated for each clinic separately. The randomisation sequence was created using the ralloc command in Stata (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP). The trial was open-label and unmasked as the intervention required overt participation. The trial was however assessor-blinded to laboratory staff that analysed the EID HIV tests and to the authors of the paper. The randomisation code was disclosed to the research team only after the trial had ended and data management was complete in September 2020. Every Monday morning, a mobile text message in Kiswahili was automatically sent asking “How are you?” (“Mambo?” in Kiswahili) from a digital platform using a generic program number saved to the participant’s phone under a name of their chosing. The participants were instructed to respond within 48 h that they were either doing well (e.g. “Sawa” in Kiswahili) or that they had a problem (e.g. “Shida” in Kiswahili). The participants’ responses (in Kiswahili or English), and instances of non-response were automatically registered in the platform. Reported problems were automatically forwarded to a mobile phone at the participant’s local ANC clinic where they were followed up by a healthcare worker within 48 h. A list of non-responses was made every week by a study coordinator and delivered (by paper or text message) to the local ANC clinic by a research assistant, where they were followed up by a healthcare worker. Healthcare workers contacted the participants by telephone to determine the nature of the problem or reason for not responding and then provided assistance. Outcomes of the follow-up calls were recorded in a log by the healthcare worker. The WelTel intervention continued until 24 months postpartum, except in case of a miscarriage, stillbirth, infant death, maternal death, or participant withdrawal from the intervention, when text messages were discontinued. Apart from weekly text-messaging, participants in the intervention group received the same PMTCT care as the control group. Intervention group participants were informed that the messaging did not replace routine clinic services, that scheduled appointments should be honoured, and that all emergencies should be handled by usual means. Participants in the control group received routine PMTCT care, based on the WHO’s PMTCT Option B+ guidelines7, which included a defaulter tracing outreach program that was initiated by AMPATH in western Kenya in 2005 (i.e. patients who missed one or more scheduled appointments were traced, first by telephone, then by household visit)30,31. EID HIV testing was a secondary trial outcome of the WelTel PMTCT trial24. It was defined as the proportion of HIV-exposed infants who had an early HIV test by eight weeks of age. The definition of an EID HIV test was modified after the study protocol was published but before the statistical analyses were undertaken29. The definition of an EID test in the published study protocol was ‘HIV-exposed infants tested for HIV within eight weeks of birth, measured as HIV-exposed infants with known HIV status at age 10 weeks’. The change was made to be consistent with UNAIDS’ most recent definition of an EID HIV test (i.e. the percentage of infants born to women living with HIV receiving a virological test for HIV within two months of birth)32. During participant follow-up, ANC clinics involved in the trial piloted at-birth HIV testing. All infant HIV tests prior to eight weeks of age (including at-birth tests), were included to define EID HIV testing by eight weeks of age. In case of an at-birth test, the participants were instructed to have a second infant HIV test at around six weeks of age. To exclusively assess EID HIV testing around six weeks of age, we also analysed the proportion of infants tested for HIV between four to eight weeks of age. In addition, we calculated the proportion of infants diagnosed with HIV based on EID HIV testing. Information about EID testing was extracted from patient files, patient registers and the Kenyan National AIDS & STI Control Program (NASCOP) database. Information on socio-demographic and HIV-related characteristics was collected through interviews at trial enrolment, including: age (18–24, 25–29, 30–34, 35–44 years); education (≤ primary schooling, secondary schooling, higher education); married or living with a partner (yes, no); employment status (working outside the household i.e. employed, self-employed, casual labour, farm work, or not working outside the household i.e. unemployed, homemaker, student); time since HIV diagnosis (< 6 months, ≥ 6 months); travel time to clinic (< 1 h, ≥ 1 h); HIV status disclosure to someone (yes/no); and HIV disclosure to a partner (yes/no). HIV status disclosure to a partner excluded participants that were not married or living with a partner. Overall, there were few participants from the Matayos clinic. Therefore, based on similarities between the Chulaimbo and Matayos clinics, participants from these clinics were combined in regression models and subgroup analyses. The sample size of the WelTel PMTCT trial was based on trial’s primary endpoint24. A sample size of approximately 300 in each study group was estimated to have at least 80% power to detect an 11% difference in the primary endpoint using α = 0.05. We expected 30% of participants in the control group to be retained at 24 months. The primary analysis of the effect of the intervention on EID HIV testing was intention to treat, i.e. all women were analysed according to the treatment group to which they were originally allocated. We used descriptive statistics to summarize demographic statistics of the study population. For the outcome of EID HIV testing by eight weeks and for the analysis of infants tested between four to eight weeks of age, we used Poisson regression with robust standard errors to estimate risk ratios (RR), rather than odds ratios as infant testing is not a rare event. Poisson regression with robust standard errors is known to perform well over a wide range of settings, and could thus be pre-specified with confidence in the trial’s statistical analysis plan published prior to data analyses29,33,34. We ran each of these models adjusting for women’s age group at baseline, time from HIV diagnosis to trial enrolment, and ANC clinic of enrolment. We also ran these analyses excluding mother-infant pairs with a record of miscarriage, stillbirth, infant death, or maternal death prior to eight weeks of age, and a record of transfer of care before the infant was eight weeks old, which included recorded transfers of the mother prior to giving birth. We conducted sensitivity analyses to determine whether there was an association between these exclusions and trial group allocation using logistic regression. Exploratory hypothesis-generating subgroup analyses were performed using Poisson regression with robust standard errors. All p-values are unadjusted, and all models are reported. Analyses were performed using StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC.
N/A
