Description of studies
See: Characteristics of included studies; Characteristics of excluded studies.
For this review update, we identified 354 new references. After screening, we selected 14 new trials for inclusion. (Seventy-nine reports from an updated search in July 2009 have been added to Characteristics of studies awaiting classification.)
The previous version of this review (Sikorski 2002) identified one trial of lay support in progress in Scotland. Further information about this trial was not available in time for this update; however, it has since been published (Muirhead 2006). The completed trial of trained lay breastfeeding counsellors in London identified by Sikorski (Sikorski 2002) has since been published and is included in this review as Graffy 2004. The study by Dennis included in Sikorski 2002, using data from Dennis's thesis, has since been published, and we have made this publication the primary reference for this study in this review (Dennis 2002).
This review has a total of 34 included studies which come from 14 countries. Six studies were conducted in each of the following countries: Canada (Dennis 2002; Gagnon 2002; Lynch 1986; Mongeon 1995; Pinelli 2001; Porteous 2000); USA (Brent 1995; Chapman 2004; Frank 1987; Grossman 1990; Pugh 2002; Wrenn 1997); UK (Graffy 2004; Jenner 1988; Jones 1985; Moore 1985; Morrell 2000; Winterburn 2003). Four studies were conducted in Brazil (Albernaz 2003; Barros 1994; Leite 1998; Santiago 2003); two studies were conducted in Bangladesh (Haider 1996; Haider 2000) and Australia (McDonald 2003; Quinlivan 2003). Single studies came from India (Bhandari 2003), Nigeria (Davies-Adetugbo 1997), Italy (Di Napoli 2004), Iran (Froozani 1999), the Netherlands (Kools 2005), Belarus (Kramer 2001), Mexico (Morrow 1999) and Sweden (Sjolin 1979).
The total number of mother-infant pairs included is 29,385.
There were 42 excluded studies. The main reasons for exclusion were high loss to follow up, evaluation of an educational intervention and lack of data. Full details are available in the 'Characteristics of excluded studies' table.
The main purpose of this review was to analyse the impact of the intervention, extra breastfeeding support, with the purpose of facilitating continued breastfeeding. We included studies if the intervention occurred in the postnatal period alone or also included an antenatal component. We excluded interventions taking place in the antenatal period alone, as well as interventions described as solely educational in nature.
The main outcome measure was the effect of the intervention on duration of breastfeeding to specified points in time. Outcomes were recorded for stopping feeding before four to six weeks and two, three, four, six, nine and 12 months. Other outcomes of interest were exclusive breastfeeding, measures of neonatal and infant morbidity (where available) and measures of maternal satisfaction with care or feeding method.
Personnel and training
The included studies evaluated support provided by a variety of medical, nursing and allied professionals (for example, nutritionists) as well as lay people. Lay support was either voluntary or remunerated. In previous editions of this review, support has been categorised as either 'professional' or 'lay'. A new category, 'lay and professional', has been devised for this update. Nineteen studies used professionals for support (Albernaz 2003; Davies-Adetugbo 1997; Di Napoli 2004; Frank 1987; Froozani 1999; Gagnon 2002; Grossman 1990; Jones 1985; Kools 2005; Kramer 2001; Lynch 1986; McDonald 2003; Moore 1985; Pinelli 2001; Porteous 2000; Quinlivan 2003; Santiago 2003; Sjolin 1979; Wrenn 1997). Nine studies used lay people for support (Chapman 2004; Dennis 2002; Graffy 2004; Haider 2000; Jenner 1988; Leite 1998; Mongeon 1995; Morrell 2000; Morrow 1999) and six studies used a combination of both professional and lay people (Barros 1994; Bhandari 2003; Brent 1995; Haider 1996; Pugh 2002; Winterburn 2003).
Details of those involved in providing support and the interventions used are given in the table of 'Characteristics of included studies'. Eight studies (Albernaz 2003; Davies-Adetugbo 1997; Di Napoli 2004; Froozani 1999; Haider 1996; Haider 2000; Kramer 2001; Leite 1998) used either the 18-hour or 40-hour WHO/UNICEF breastfeeding counselling/lactation management courses as the basis for the training of breastfeeding supporters. A further nine studies reported providing the supporter with extra formal training in breastfeeding support prior to the intervention (Bhandari 2003; Chapman 2004; Dennis 2002; Gagnon 2002; Graffy 2004; Morrell 2000; Mongeon 1995; Morrow 1999; Santiago 2003). Where the length of additional training was reported, this ranged from sessions lasting 2.5 hours to 40 hours.
We also subdivided the studies into broad categories to examine aspects of the interventions, as discussed in the Methods section.
In the majority of studies, the comparison group was reported to have received 'usual postnatal care', which varies both between and within countries. The care at the time of the trials may also differ from that which is offered at the present time. Wherever there were individual study details on care received by the comparison groups, these are given in the 'Characteristics of included studies' table.
Breastfeeding was usually reported as being either partial or exclusive, with no further definitional refinement. Few studies reported both partial and exclusive rates at all time points. Reporting of health outcomes was scanty and inconsistent, allowing little joint analysis. The timing of outcome assessments varied considerably between studies, ranging from two weeks to one year postnatally. Several studies took repeated measurements of breastfeeding rates, and some reported mean duration.
Differences in groups studied
Support was usually offered to women intending to breastfeed, but in three studies (Brent 1995; Morrell 2000; Quinlivan 2003) intention to formula-feed was not an exclusion criterion. In the small study by Porteous (Porteous 2000), support was only offered to those breastfeeding women who identified themselves as unsupported on a self-report questionnaire. In two studies the intervention was targeted at low-income women (Chapman 2004; Pugh 2002), whereas the intervention was only offered to women under the age of 18 years in another (Quinlivan 2003).
In one study (Moore 1985), only women with a personal or partner history of asthma or eczema were selected. Two further trials (Davies-Adetugbo 1997; Haider 1996) studied the effect of support for mothers of sick infants with moderate diarrhoeal disease. One trial (Bhandari 2003) studied the effect of breastfeeding support delivered to communities and included diarrhoea prevalence outcomes. In another trial (Pinelli 2001), the focus of the study was the effect of breastfeeding support to parents of very low birthweight babies.
Effects of interventions
The initial searches of MEDLINE and EMBASE identified 327 references. Twenty-seven references not identified by previous editions of the review were identified by a search of the Cochrane Pregnancy and Childbirth Group Trials Register. Fourteen new trials were finally added to the 20 that featured in Sikorski 2002. The 34 studies included in this review are from 14 countries and include 29,385 mother-infant pairs.
Some studies used professional or lay individuals, or a combination of both. Data were collected regarding the effect of the intervention on breastfeeding duration. Some studies reported exclusive breastfeeding rates, but others were ambiguous and it was difficult to ascertain whether the infant was fed breast milk alone. We collected data on the effect of the interventions on any form of breastfeeding to assess the impact of interventions to enable women to continue breastfeeding.
Types of outcome measures
The main outcome measure was the effect of the interventions on duration of breastfeeding to specified points in time. Outcomes were recorded for stopping feeding before four to six weeks and two, three, four, six, nine and 12 months. Other outcomes of interest were exclusive breastfeeding, measures of neonatal and infant morbidity (where available) and measures of maternal satisfaction with care or feeding method.
Overall effect on any breastfeeding
The main summary outcome measure was breastfeeding at the time of the last study assessment up to six months. There continues to be a beneficial effect on the duration of any breastfeeding up to six months with the implementation of any form of extra support (relative risk (RR) 0.91, 95% confidence interval (CI) 0.86 to 0.96). However, it is noted that there was significant heterogeneity (I2 = 53.6%). Sensitivity analysis using only studies with adequate allocation concealment demonstrated a similar result (RR 0.90, 95% CI 0.83 to 0.98, I2 62.4%).
In order to explore any differential effect of support conditional on the baseline prevalence of breastfeeding in the area in which the trial was conducted, we divided the trials into three categories denoted by high (greater than 80%), intermediate (60% to 80%) or low (less than 40%) initiation rates in the local area. Analysis of the trials conducted in settings with intermediate breastfeeding initiation (Chapman 2004; Dennis 2002; Di Napoli 2004; Gagnon 2002; Graffy 2004; Jones 1985; Lynch 1986; Mongeon 1995; Morrell 2000; Pinelli 2001; Porteous 2000; Pugh 2002; Winterburn 2003; Wrenn 1997) demonstrated all forms of support had a significant benefit on breastfeeding (RR 0.92, 95% CI 0.85 to 0.98), whereas there was no significant effect where there were high rates of breastfeeding (RR 0.91, 95% CI 0.81 to 1.01) (Albernaz 2003; Barros 1994; Bhandari 2003, Froozani 1999; Kramer 2001; Kools 2005; Leite 1998; McDonald 2003; Morrow 1999; Quinlivan 2003). There was no significant effect in areas with low initiation rates (RR 0.88, 95% CI 0.69 to 1.12) (Brent 1995; Frank 1987; Grossman 1990).
Analysis of results at different periods of follow up presented some challenges in interpreting the data. There was variability between the studies regarding the time points when data were collected, therefore caution has to be exercised when interpreting the trends. However, analysis of results at different periods of follow up suggested that the benefit of all forms of support was present at all time points up to nine months.
Overall effect on exclusive breastfeeding
The effect of any support on mothers exclusively breastfeeding is greater than on women continuing any form of breastfeeding (RR 0.81, 95% CI 0.74 to 0.89) (Albernaz 2003; Bhandari 2003; Frank 1987; Froozani 1999; Gagnon 2002; Graffy 2004; Haider 2000; Jenner 1988; Kools 2005; Kramer 2001; Leite 1998; McDonald 2003; Moore 1985; Morrell 2000; Morrow 1999; Porteous 2000; Pugh 2002; Santiago 2003; Sjolin 1979; Wrenn 1997). There is significant heterogeneity in this group of 20 trials (I2 = 92.2%). Those women who receive any form of support are less likely to give up exclusive breastfeeding before five months.
Trials comparing an intervention of extra professional support to usual care in preventing the cessation of any breastfeeding showed professional support to be effective at four months but not at other time points (RR for stopping any breastfeeding before four months in five trials 0.78, 95% CI 0.67 to 0.91) (Albernaz 2003; Frank 1987; Froozani 1999; Quinlivan 2003; Sjolin 1979). However, the overall effect of extra professional support on stopping any breastfeeding did not reach statistical significance (RR for stopping any breastfeeding before last study assessment up to six months in 16 trials 0.94, 95% CI 0.87 to 1.01) (Albernaz 2003; Frank 1987; Froozani 1999; Gagnon 2002; Grossman 1990; Di Napoli 2004; Jones 1985; Kools 2005; Kramer 2001; Lynch 1986; McDonald 2003; Pinelli 2001; Porteous 2000; Quinlivan 2003; Sjolin 1979; Wrenn 1997). There was heterogeneity present among the 16 trials (I2 = 49.8%).
Professional support resulted in a beneficial effect on exclusive breastfeeding (RR 0.91, 95% CI 0.84 to 0.98) (Albernaz 2003; Frank 1987; Froozani 1999; Gagnon 2002; Kools 2005; Kramer 2001; Lynch 1986; McDonald 2003; Moore 1985; Porteous 2000; Sjolin 1979; Wrenn 1997). This is apparent in the first few months (RR before four to six weeks 0.69, 95% CI 0.51 to 0.92; RR before two months 0.76, 95% CI 0.61 to 0.94; RR before three months 0.84, 95% CI 0.72 to 0.99).
Trials that used lay people to deliver the intervention demonstrated a significant reduction in breastfeeding cessation at the time of the last study assessment (RR 0.86, 95% CI 0.76 to 0.98) (Chapman 2004; Dennis 2002; Graffy 2004; Leite 1998; Mongeon 1995; Morrell 2000; Morrow 1999). Significant heterogeneity was present among these studies (I2 = 75.6%). Further subgroup analysis did not reveal a statistically significant effect at any time point up to four months. However, in the studies of lay support which reported exclusive breastfeeding, there was a marked reduction in the cessation of exclusive breastfeeding before the last study assessment (RR 0.72, 95% CI 0.57 to 0.90) (Graffy 2004; Haider 2000; Jenner 1988; Leite 1998; Morrell 2000; Morrow 1999). There was heterogeneity among these studies (I2 = 96.3%). Further subgroup analysis indicated that this effect was significant within the first three months (RR before four to six weeks 0.66, 95% 0.46 to 0.96; RR before two months 0.44, 95% CI 0.26 to 0.73; RR before three months 0.42, 95% CI 0.31 to 0.57).
Combined professional and lay support
Five studies compared combined lay and professional support with usual care (Barros 1994; Bhandari 2003; Brent 1995; Pugh 2002; Winterburn 2003). Overall these showed a significant reduction in cessation of any breastfeeding (RR 0.84, 95% CI 0.77 to 0.92, I2 = 55.7%), especially in the first two months (RR before four to six weeks 0.65, 95% 0.51 to 0.82; RR before two months 0.74, 95% CI 0.66 to 0.83). Two studies (Bhandari 2003; Pugh 2002) demonstrated a significant reduction in cessation of exclusive breastfeeding (RR 0.62, 95% CI 0.50 to 0.77, I2 = 82.2%). However, these results should be viewed with caution as the numbers analysed are small, and there was only one high-quality trial included in this section (Bhandari 2003).
We performed subgroup analyses to test formally for significant differences between the groups offering professional support, lay support and combined professional and lay support. For stopping any breastfeeding there was no evidence of difference between subgroups except for borderline difference at two months (p=0.0468), where the tendency was for combined support to be most effective. For stopping exclusive breastfeeding, there were significant differences for all times tested (three months, four months, six months), and at each time point either lay or combined lay and professional support was most effective.
Differing modes and timing of support
The studies that offered face-to face support showed a statistically significant benefit (RR for giving up any breastfeeding 0.85, 95% CI 0.79 to 0.92) (Albernaz 2003; Barros 1994; Bhandari 2003; Brent 1995; Chapman 2004; Froozani 1999; Jones 1985; Kramer 2001; Leite 1998; Morrell 2000; Morrow 1999; Pinelli 2001; Quinlivan 2003; Winterburn 2003). The overall test for heterogeneity was I2 = 57.4%. In those studies where telephone support was offered, no significant effect was demonstrated (RR 0.92, 95% 0.78 to 1.08) (Dennis 2002; Frank 1987; Grossman 1990; Lynch 1986; Mongeon 1995). Where both telephone and face-to-face support were provided, there was no significant improvement in breastfeeding continuance (RR 1.00, 95% CI 0.91 to 1.09) (Di Napoli 2004; Gagnon 2002; Graffy 2004; Kools 2005; McDonald 2003; Porteous 2000; Pugh 2002; Sjolin 1979; Wrenn 1997).
The effect on stopping any breastfeeding at last study assessment before six months that was measured in studies of interventions containing an antenatal element to breastfeeding support (RR 0.92, 95% CI 0.83 to 1.02) was not significant, whereas the effect in those studies offering postnatal support alone did achieve statistical significance (RR 0.89, 95% CI 0.84 to 0.96). However, effect estimates were similar and the difference between the effect of interventions containing an antenatal element and the effect of interventions offering postnatal support alone was not statistically significant.
There was a highly significant beneficial effect on exclusive breastfeeding two to three weeks after discharge from a healthcare facility in the two studies of support for mothers with sick infants (RR for stopping exclusive breastfeeding before two to three weeks after discharge 8.32, 95% CI 4.94 to 14.01, I2 = 0%) (Haider 1996; Davies-Adetugbo 1997). Three studies (Bhandari 2003; Davies-Adetugbo 1997; Haider 1996) reported on recurrence of diarrhoea. There was a marked short-term reduction in the recurrence of diarrhoea in these trials (RR for recurrence before two to three weeks follow-up (RR 0.70, 95% CI 0.54 to 0.9). There was statistical heterogeneity among these three studies (I2 = 53.8%). In the study by Haider (Haider 1996), eight babies in the control group and two babies in the intervention group had died two weeks after discharge from hospital. The difference in the populations in these trials, when compared to the healthy mother-infant dyads included in other studies, led to their exclusion from the main meta-analysis.
Few trials reported health outcomes and it was not possible to combine these statistically. The PROBIT study (Kramer 2001) found a significant reduction in the risk of one or more gastrointestinal infections and of atopic eczema in the group receiving care from health professionals who had received the WHO/UNICEF Baby Friendly Initiative training. There was no significant reduction in respiratory tract infection. Frank 1987 found no difference in breastfeeding rates in those infants rehospitalised during their study while Froozani 1999 observed a significant reduction in the mean number of days of gastrointestinal illness in the group receiving support but no significant difference in respiratory illness.
Measures of satisfaction
Satisfaction measures were poorly reported. Jones 1985 reported satisfaction with the amount of help received, both at home and in hospital, and found this to be greater in the intervention group. Two studies reported maternal satisfaction with infant feeding. Dennis (Dennis 2002) found no significant differences between the peer and control groups' mean scores on the Maternal Breastfeeding Evaluation Scale (mean scores 53.81 (standard deviation (SD) 5.69) versus 52.98 (SD 5.94), P = 0.26) (Leff 1994). However, significantly more mothers in the control group reported overall dissatisfaction with their infant feeding method. Graffy 2004 reported no difference between intervention group and control group on most measures but found the intervention group were less likely to believe they were not making enough milk.
Socially disadvantaged groups
One study (Jones 1985) reported effects of the supportive intervention in different social groups. In this study, the greatest difference in the proportion of women still breastfeeding at four weeks was in social classes IV and V (86% of social classes IV and V in the intervention group breastfeeding at four weeks versus 58% in social classes IV and V in the control group, P < 0.01). In the UK people are classified into social groupings according to their (or their partner's) occupation, for example, social class IV and V includes women with partners in manual or unskilled occupations.
In two further studies, low-income women from the US were included (Chapman 2004; Pugh 2002), and in another study (Quinlivan 2003) women under the age of 18 years were recruited.
Effect of differing training programmes
Eight trials (Albernaz 2003; Davies-Adetugbo 1997; Di Napoli 2004; Froozani 1999; Haider 1996; Haider 2000; Kramer 2001; Leite 1998) reported using either the 18- or 40-hour WHO/UNICEF breastfeeding training courses. Another trial (Bhandari 2003) used a course based on an adaptation of the WHO Integrated Management of Childhood Illness Training Manual on Breastfeeding Counselling (WHO 1997). Meta-analysis of the six trials using WHO/UNICEF training (Albernaz 2003; Bhandari 2003; Froozani 1999; Haider 2000; Kramer 2001; Leite 1998) showed significant benefit in prolonging exclusive breastfeeding (RR 0.69, 95% CI 0.52 to 0.91) but the trials were statistically heterogeneous (I2 = 97.9%).
Two trials (Chapman 2004; Morrow 1999) used the peer counsellor programme developed by La Leche League, the international lay breastfeeding support organisation and in Graffy 2004 the counsellors were trained by the National Childbirth Trust, a UK-based childbirth and breastfeeding advocacy organisation.
The length of training offered to lay supporters varied from 2.5 hours (Dennis 2002) to 40 hours (Albernaz 2003; Haider 2000). Other studies reported providing some extra training in breastfeeding support prior to the intervention (Dennis 2002; Gagnon 2002; Mongeon 1995; Morrell 2000; Santiago 2003).