Efficacy of home-based peer counselling to promote exclusive breastfeeding: a randomised controlled trial

Study participants

San Pedro Martir is a periurban community on the southwestern outskirts of Mexico City, compromising three subdivisions that vary in sociodemographic characteristics. In the most established area, families typically live in houses with running potable water, tiled floors, and access to central sewage and garbage removal services. In the most recently settled area, families typically live in crowded conditions, in huts with earth floors, and obtain water from pipe outlets some distance away. The total population is about 30000. Medical care is provided to this population by a government clinic, various private primary-care doctors, and public, private, and military hospitals. The study population has been fully described elsewhere.27 At the time of this study, many hospitals serving the area were changing policies and procedures to support breastfeeding, following international standards known as the Baby Friendly Hospital Initiative.1,19

This randomised, community-based intervention trial included two intervention groups and a control group. Motherinfant pairs in the intervention groups received either six or three home visits from peer counsellors. In the six-visit group, mothers were visited in mid and late pregnancy, in the first week and weeks 2, 4, and 8 post partum. In the three-visit intervention group, mothers were visited in late pregnancy, in the first week, and week 2 post partum. Furthermore, peer counsellors were permitted to respond to occasional requests for additional support that were initiated by intervention-group mothers. Control-group mothers with lactation problems were referred to their own physicians. No other sources of breastfeeding counselling were available in the community. This study was approved by the institutional review boards of the Instituto Nacional de la Nutricion in Mexico City and Eastern Virginia Medical School in Norfolk (Virginia, USA).

Before the study started, San Pedro Martir was mapped into 39 clusters with two to four city blocks each. 13 clusters were allocated randomly to each study group, stratified by subdivision. This randomisation schedule was generated by computer. Clusters rather than individuals were randomised to keep to a minimum the contamination of influences expected if relatives and close neighbours were assigned to different study groups.

Enrolment began in March, 1995, and continued to September, 1996. Study mothers were identified by a semiannual door-to-door census and continuous reporting of new pregnancies in the community by study staff and mothers. For more than 15 years, this method has been used to identify virtually all eligible mothers in the study community;8–11 no record system is available for comparison. All pregnant women residing in the study area were considered eligible, visited at home by a study physician to verify eligibility, and invited to participate in a study of breastfeeding practices. Mothers were provided with an oral and written description of procedures for their study group; other study groups were not discussed. Mothers' consent was obtained in writing. Women were considered ineligible and excluded from the study if they refused participation or moved out of the area before the first postpartum home visit, or if the baby died. Infants were followed up until 3 months of age to assess exclusive breastfeeding and diarrhoea, and 6 months of age to assess duration of any breastfeeding. The study ended in December, 1996.

Design

All data were collected through structured interviews of mothers residing in study households by two experienced staff other than the peer counsellors. The study hypothesis could not be concealed from these staff, but they were trained to adminster all questions in a standard manner, and they undertook an equal proportion of interviews in each study group. Baseline interviews were carried out in the last trimester of pregnancy to examine sociodemographic factors, breastfeeding history and intention, and other factors. A perinatal questionnaire ascertained pregnancy history, the health of mother and infant, and early infant-feeding practices. Five follow-up interviews were scheduled for all study mothers at 2, 4, and 6 weeks, and at 2 and 3 months post partum to record infant-feeding practices in the previous week and any lactation problems experienced. During weeks in which both a counselling visit and a data-collection interview were due, the data-collection interview was scheduled to follow the counselling visit by 2–3 days. During these interviews, mothers were asked whether the infant had experienced diarrhoea since the last interview, and to describe any episodes and whether they had taken the infant to a doctor. An exit interview at 6 months post partum examined the duration of breastfeeding and maternal attitudes towards the peer counsellors.

Three women who had previously worked for the Instituto Nacional de la Nutricion as field data collectors were trained to promote breastfeeding as peer counsellors. Each was a resident of San Pedro Martir, aged 25–30 years, had a high-school education, and had a commitment to breastfeeding, although they did not necessarily have previous personal breastfeeding experience. These peer counsellors were trained and supervised by staff of La Leche League of Mexico and the physician study coordinator (MLG), who was also trained in lactation management. The peer-counsellor training consisted of 1 week of classes, 2 months in lactation clinics and with mother-tomother support groups, and 1 day of observation and demonstration by visiting experts. Finally, the peer counsellors practised in a neighbourhood nearby San Pedro Martir for 6 months before the intervention trial, during which the content of messages and problem-solving skills were refined.

A rapid ethnographic study of infant feeding was done in San Pedro Martir before the intervention trial to guide educational approaches.27 A set of visual aids was developed specifically for this project on the basis of existing materials of La Leche League. Home visits to pregnant women focused on the benefits of exclusive breastfeeding, especially during illness; basic lactation anatomy and physiology; positioning of the infant and “latching on”; common myths; typical problems and solutions; and preparation for birth. Postpartum visits with the mother focused on establishing a healthy breastfeeding pattern, addressing maternal concerns, and providing information and social support. Key family members who could provide support to mothers also were included in these counselling visits.

The primary study outcome was exclusive breastfeeding, defined as giving maternal milk at the only infant food source in the previous week, with no other liquids or food given.28 Secondary outcomes were duration of breastfeeding, the proportion of infants who had an episode of diarrhoea in the first 3 months (cumulative incidence), and maternal satisfaction with counselling. Diarrhoea was defined as more frequent and liquid stools than normal for the infant, as reported by the mother.

Statistical analysis

On the basis of primary hypothesis, we calculated the minimum required sample size to be 120 participants, setting α=0·05, and a one-sided test of hypothesis. This sample size gave 86% power to detect a 20% absolute difference in exclusive breastfeeding in intervention versus control participants (24% vs 4%), if there was no design effect, or 76% power to detect the same difference if there was a design effect of 1·2.29

Data were analysed by means of Stata statistical software (release 5·0). Because randomisation was based on clusters, within-cluster correlation and design effects were assessed. Descriptive analyses included frequencies, contingency tables, and odds ratios. Associations between categorical variables were tested with χ2 or Fisher's exact test. The effect of the intervention on exclusive breastfeeding was analysed by means of generalised estimating equation (GEE) models that accounted for within-individual correlation of outcomes measured at five timepoints from 2 weeks to 3 months post partum. We also used GEE models to compare outcomes at each timepoint, accounting for correlation within clusters. Intervention groups were specified in the models as dummy variables. Maternal, infant, and hospital factors were examined as risk factors, and for potential confounding and interaction effects relative to study outcomes. We also used survival analysis and the log-rank procedure to test differences between groups in time to first failure of exclusive breastfeeding and duration of any breastfeeding. One-sided tests of significance were used to test intervention efficacy, defined as a significant increase in exclusive breastfeeding, increased duration of any breastfeeding, and decreased infant diarrhoea in the intervention groups compared with the control group. Justification of a one-sided hypothesis was based on the need to test programme efficacy (ie, demonstration of significant improvement) and the lack of evidence of potential for harm. Significance was set at p<0·05. All analyses were by intention to treat.

Participants

170 pregnant women were identified in San Pedro Martir and were asked to participate. Of these, only 12 (7%) refused, and 28 were ineligible and excluded (figure 1). Participants and non-participants did not differ by location within study area or by study group, and did not differ in socioeconomic status. 130 mother–infant pairs participated in the study, of whom 125 remained in the study at 3 months. Exit interviews were done in 117 participants: 104 at 6 months, and 13 between 3 and 6 months post partum. Study participants were identified and enrolled in 31 of the 39 geographically defined clusters.

Click to enlarge image

Figure 1. Trial profile

No significant differences were found among study groups in baseline factors, including breastfeeding initiation (table 1). The proportion who initiated breastfeeding within a few hours of delivery was similar in each study group. 35 (27%) mothers were primiparous, and 31% gave birth by caesarean section.

Click to view table

Table 1. Sociodemographic characteristics and perinatal conditions of 130 study mothers and infants

The peer counsellors were assigned approximately equal numbers of study mothers in the two intervention groups. No differences were found between intervention groups in the timing of the first postpartum visit. Furthermore, no differences were found among counsellors in terms of their clients' breastfeeding outcomes.

Confounding and interactions

Sociodemographic, health, and hospital factors (table 1) were analysed in relation to the intervention and exclusive breastfeeding. None of these factors was found to be a significant predictor of exclusive breastfeeding or to modify the efficacy of the intervention; thus, final models specified only intervention groups as independent variables.

Within-cluster correlation values were calculated for each study outcome and interview time and were found to be negligible: -0·011 at the 2-week visit and 0·074 at 3 months. Design effects were calculated and found to approach unity (no effect) for each study group. These results show that the cluster randomisation design achieved the equivalent of individual randomisation.

Main outcome

In the six-visit group, 35 (80%) of 44 women were exclusively breastfeeding at 2 weeks and 28 (67%) of 42 were doing so at 3 months; in the three-visit group, the corresponding numbers were 32 (62%) of 52 at 2 weeks and 25 (50%) of 50 at 3 months (figure 2). In the control group eight (24%) of 34 women were exclusively breastfeeding at 2 weeks, a significantly greater (p<0·05) proportion than in the 1988–91 historical cohort (4%), and four (12%) of 33 were exclusively breastfeeding at 3 months post partum, which did not differ significantly from previously recorded rates.

Click to enlarge image

Figure 2. Proportion of mothers who exclusively breastfed their infants by infant age and study group

Vertical bars represent 95% CI.

Both the six-visit and three-visit intervention groups had significantly (p<0·001) more exclusive breastfeeding over time than controls, as shown by GEE models. GEE models fitted at each timepoint that accounted for within-cluster correlation confirmed this finding. Furthermore, the higher rate of exclusive breastfeeding in the six-visit group compared with the three-visit group was significant (p=0·015 by GEE analysis), but this difference seems to have been establised by 2 weeks, when an 18% absolute difference in exclusive breastfeeding was observed (p=0·028, by χ2).

Some study mothers who breastfed exclusively gave their infants supplementary feedings for a short time, then returned to exclusive breastfeeding, a pattern observed in all study groups, typically associated with perceived illness, stress, or doctor's advice. In the sixvisit group, the prevalence of exclusive breastfeeding varied from 66% at 4 weeks to 75% at 6 weeks and 67% at 3 months post partum. Practice of exclusive breastfeeding at all five measurement times, 2 weeks to 3 months, was found in 21 (50%) of 42 in the six-visit group, 19 (38%) of 50 in the three-visit group, and four (12%) of 33 controls (p<0·001, log-rank test). Thus, 17% of the six-visit group and 12% of the three-visit group exclusively breastfed at 3 months but failed to maintain exclusive breastfeeding throughout follow-up. Examination of the types of supplementary feedings introduced by 3 months shows that higher rates of exclusive breastfeeding in the intervention groups were achieved by less use of formula-milk feedings, solid foods, teas, and water feedings.

Secondary outcomes

To provide sufficient power to detect differences in secondary outcomes, the two intervention groups were collapsed into a single group for comparison with controls. Duration of any breastfeeding was signifiantly (p=0·024, log-rank test) greater among intervention group than control-group mothers (table 2). 21 infants had at least one episode of diarrhoea between birth and 3 months of age: eight of the six-visit group, four of the three-visit goup, and nine controls. Of the 21 diarrhoea episodes, 18 were associated with a visit to a doctor, two with dehydration, and four with vomiting, and four episodes lasted longer than 1 week. Control infants had a significantly greater incidence of diarrhoea than intervention infants (relative risk 2·1 [90% CI 1·11–4·04]; p=0·029).

Click to view table

Table 2. Secondary measures of intervention efficacy

In the exit interview, nearly all intervention-group mothers reported that the peer counsellor was helpful and supportive. The most important source of infantfeeding advice for intervention-group mothers typically was a peer counsellor (66%), followed by a physician (19%), and their mothers (7%). In contrast, among controls, 50% listed a physician as their most important source of infant-feeding advice, 22% listed their mothers, and 2% listed a peer counsellor.