Trial design

We previously described the study methods [24]. Briefly, we conducted a parallel group, standard care-controlled randomized trial evaluating infant feeding practices at four follow up visits, among 276 mother-child pairs. We randomly assigned participants to either weekly mobile phone text messaging plus in-person motivational interviewing or to standard infant counselling for 24 weeks. The principal investigator used Stata 17 random number generating command (RALLOC), to generate the randomization allocation sequence restricted by permuted block sizes of 2, 4 and 6, with a 1:1 allocation ratio. A research assistant enrolled study participants, collected baseline data using the Research Electronic Data Capture (REDCap) mobile application and uploaded the data to the REDCap online server. The principal investigator assigned study group sequentially, using the locked unmodifiable allocation sequence stored on the REDCap online server. An independent data analyst performed unblinded interim analysis during trial monitoring. Study participants and the principal investigator were aware of the group assignment. A research assistant not administering the study interventions completed outcome evaluation questionnaires at follow up visits, without knowledge of group assignment. This manuscript was written following the Consolidated Standards of Reporting Trials (CONSORT 2010) guidelines for reporting parallel group randomized trial [25].

Setting and study population

Mothers were counselled to exclusively breastfeed for the first 6 months by nurses and trained lay counsellors during routine postnatal clinic visits at their local primary healthcare clinics. Khayelitsha District Hospital provides secondary-level healthcare services. Mothers were informed about the study and invited to participate within 24 h of giving birth at Khayelitsha District Hospital and followed for 6 months at Masiphuhlisane Research Centre in Khayelitsha, Cape Town. We enrolled mothers living with HIV and on cART, initiating breastfeeding, 18 years or older, with a mobile phone, and their infants. We excluded mothers who initiated formula feeding or who were advised by healthcare providers not to breastfeed due to high viral load or other breastfeeding contraindications, gave birth to more than one infant, infant birthweight < 2500 g or gestational age at birth < 36 weeks.

Study interventionsMobile phone text messaging and motivational interviewing

A research assistant sent text message weekly to mothers in the intervention group, encouraging exclusive breastfeeding and inquired if there were any breastfeeding problems. The research assistant contacted mothers who indicated a breastfeeding problem and those who failed to respond within 48 h. Text messages were dispatched weekly throughout the entire 6-month follow-up period, with disruptions only during the Christmas holiday breaks, between 15 December 2022 and 10 January 2023 and between 15 December 2023 and 10 January 2024, for mothers who were actively being followed up during that time. Mothers had face-to-face individual motivational interviews post-delivery at week 2, 6, and 10. During the interviews, the research assistant and the mother discussed breastfeeding practices and problems, potential solutions, reinforced mother’s own self-motivational statements and readiness to correct suboptimal infant feeding practices and affirm the mother’s freedom of choice. The discussions included the importance of exclusive breastfeeding, adherent to ART to maintain a suppressed viral load and risks of mixed feeding. Mothers who had problems with breastfeeding or were concerned about their viral load were advised to seek healthcare services at their primary health facility. Text messaging and motivational interviews were discontinued for mothers who completely stopped breastfeeding before trial conclusion. The mother-child pair who stopped breastfeeding early were followed for the secondary outcomes.

Standard care infant feeding counselling

As part of standard care mothers were referred to a community health worker to support ART adherence, and breastfeeding postdelivery, were expected to attend monthly routine child growth monitoring, receive infant feeding counselling by primary healthcare nurses and trained lay counsellors during routine visits. Maternal viral load monitoring was done at delivery and 6-monthly or every 4 to 6 weeks for mothers with suppressed and unsuppressed viral load, respectively. Babies were tested for HIV at birth, 6, 10 weeks and 6 months. Mothers’ attendance of standard of care routine visits was not monitored during the study.

Sample size, power, and detectable differences

The study was only powered to detect minimum importance difference exclusive breastfeeding rates between study groups. We expected exclusive breastfeeding rate of 8% from birth through 6 months in the standard care group [9]. To detect a difference of 15% between the intervention and standard care groups (i.e., 23% vs. 8%), 182 mother-infant pairs were required for a two-sided test traditional fixed sample size computation. Adjusting the sample size for two planned analyses using the O’Brien-Fleming inflation factor 1.01 at the 0.05 significance level and 80% power, we revised the sample size to 182 × 1.01 = 184 and further inflated the sample size by 33% to 275 mother -infant pairs, accounting for loss to follow up.

Study measurements and procedures

The study included an enrolment maternal interview at Khayelitsha District hospital, four in-person follow-up visits at week 2, 6, 10, and 24, at Masiphuhlisane Research Centre, child medical record review, and child length and weight measurements. Baseline sociodemographic and clinical characteristics were obtained by interviewing the mother and abstracting medical records data. At each visit, a research assistant interviewed the mother using a questionnaire of food items given to the child in the last week or 24 h preceding the inquiry. Mothers who completely stopped breastfeeding where asked an open-ended question on reasons for stopping breastfeeding. Infant death, hospitalization, safety events and other study related data were obtained by interviewing the mother.

Baseline clinical characteristics

We classified the time of starting cART as prior to conception, during pregnancy or at delivery according to mother self-report or abstracted from the medical records. We abstracted from the medical records the most recent viral load and CD4 count at delivery. Mother’s disclosure of HIV status was classified as yes or no according to mother self-report. Baseline socio-demographic characteristics were obtained by interviewing the mother.

Study endpointsPrimary endpoints

The primary endpoints included exclusive breastfeeding and any form of breastfeeding from birth through week 24, based on the consecutive 24-hour food recall interviews. Exclusively breastfeeding was defined as a child who had only breastmilk and no other liquid or solid foods. Any form of breastfeeding was defined as a child who had breastmilk only or breastmilk and other liquid or solid foods. The infant feeding questionnaire was based on the WHO standardized instrument [26].

Secondary endpoints

Infant death or hospitalization for any cause, and non-routine or sick-clinic visits occurring within study duration, were obtained by interviewing the mother. Infant weight and length were measured by the research assistants.

Safety outcomes

Potential safety events of socially unintended consequences included relationship conflicts with the partner due to study participation, reduced child monitoring due to exaggerated perception of the benefits of breastfeeding (assessed infant immunization history as proxy), and inadvertent disclosure of participant’s HIV status.

Statistical methods

We summarized baseline characteristics using descriptive statistics. We reported effect sizes with 95% confidence intervals. Statistical significance was set at p < 0.05.

Primary analysis of primary endpoints

We conducted a complete-case analysis, including participants who completed all study follow-up visits for our primary analyses. A proportion test was used to compare differences in exclusive breastfeeding rates and any form of breastfeeding rates between study groups.

Secondary analysis of primary endpointsBinary endpoints

We conducted a binomial regression analysis on exclusive breastfeeding and any form of breastfeeding endpoints. Logistic regression was conducted to assess consistency of the effect.

Analysis of secondary endpointsTime to event endpoints

Analysis of time to stopping any form of breastfeeding and time to first all-cause hospitalization or death outcomes was compared between study groups using the log-rank test. Participants were censored at 24 weeks, or at last completed visit.

Binary endpoints

We compared the number of child hospitalization or death between study groups using chi-squared test and a logistic regression.

Continuous endpoints

WHO standardized weight-for-age, length-for-age and weight for-length z-scores were computed adjusting for gestational age at birth, using WHO Anthro Stata macro. We excluded z-scores below − 5 and above 5 from analysis. Mean weight-for-age, length-for-age and weight for-length z-scores were estimated and compared between study groups using random slope linear mixed models. The mixed effects model included study group, time and study group-time interaction as fixed effects and participant and time as random effects.

Additional analysis imputing missing outcome data

We used multiple imputation to compute missing exclusive breastfeeding and any form of breastfeeding outcome data. We used birthweight, marital status (married versus not married), educational status (primary or no schooling versus secondary or tertiary) as predictor variables. We created 20 imputations using the Stata mi estimate: logit command that combines estimates using the Rubin’s pooling rules. Multiple imputation was done separately by study group.

Interim analysis

The interim analysis was performed when 103 of the planned modified intention to treat sample size of 184 mother-child pairs completed the study. Of these, only 60 had complete data on the primary outcomes across all visits. The computed z-statistic based on 60 was 0.39, not exceeding the predefined O’Brien-Fleming stopping boundary value of ± 2.7967.

Ethics considerations

Stellenbosch University Human Research Ethics committee (reference M21/03/010) approved the study. Western Cape Department of Health approved access to Khayelitsha District Hospital (reference WC_202107_007). The DSMB reviewed the unblinded safety data accruing in the study. The study provided transport to and from the study follow up research site; mothers received a R180 (∼ US $10) voucher at each study visit, as an incentive for their participation time.