Interventions Designed to Promote Exclusive Breastfeeding in High-Income Countries: A Systematic Review

Information sources

Articles were acquired from a search of nine databases: Academic Search Complete; CINAHL with full text; The Cochrane Library: Embase: Health Policy Reference Centre; Health Source Consumer Edition; Health Source—Nursing/Academic Edition; Maternity and Infant Care: MEDLINE with full text; PsycARTICLES; Psychology and Behavioral Sciences Collection; and PsycINFO. The search terms used (Fig. 1) resulted in a total of 952 abstracts, which were reviewed for suitability by one author (M.F.). A full electronic search strategy for the MEDLINE database is shown in Supplementary Figure S1 (Supplementary Data are available online at www.liebertonline.com/bfm).

OPEN IN VIEWER

FIG. 1.

Search terms.

Eligibility criteria and study selection

A search of scholarly peer-reviewed publications was conducted for recent randomized trials published in English between January 2000 and June 2013. Articles were excluded if they were not published in English, had a follow-up period of <4 months postpartum, or referred specifically to developing or low-income countries. Other exclusion criteria were articles specific to adolescent mothers, those not aimed at increasing exclusive breastfeeding specifically, or those specific to breastfeeding in the presence of maternal smoking or a human immunodeficiency virus–positive status. Of the initial 952 abstracts revealed, 30 full-text articles were found to be relevant and were read in their entirety by all the authors. With unanimous agreement across authors obtained by discussion, a further 13 articles were excluded, leaving 17 studies for the current review: 13 randomized controlled trials and four cluster-randomized trials (Figure 2 gives a summary of the systematic search record, and Supplementary Table S1 details excluded studies with reasons).

OPEN IN VIEWER

FIG. 2.

PRISMA systematic search record.

Data collection

Data including authors and country, main aim, main outcomes, sample, design, exclusive breastfeeding definition, and theoretical basis were extracted from articles and entered into Table 1.

Risk of bias

Risk of bias for individual studies was determined by a validity scoring system, modified from the Cochrane Collaboration's tool for assessing risk of bias and used by Gardner et al. ²⁸ This system assesses and scores statistical power, intervention fidelity, blinding of outcome assessors, and intention-to-treat analysis. Scores were awarded on achievement of the required criteria. For statistical power, studies were allocated a score of 2 where power was met and a score of 0 where power analysis was not noted or statistical power was not met. Studies were given a score of 2 when attempt to assess intervention fidelity was noted and a score of 0 when intervention fidelity was not noted. For blinding of outcome assessors, studies were awarded a score of 2 where assessors were blind to group allocations and a score of 0 where blinding was either not implemented or not mentioned in the design of the study. Finally, if no intention-to-treat analysis was used or this was unclear, studies were allocated a score of 0, and they were scored as 2 if it was used. The mean risk of bias score for all included studies was assessed; scores could range from 0 to 8, with lower scores indicating a higher risk of bias.

Data analysis using the GRADE approach

The included studies were also assessed for their overall quality of evidence using the GRADE approach ²⁷ as recommended by the Cochrane Handbook for Systematic Reviews of Interventions. ²⁹ The trials were evaluated for the five factors that decrease the quality of evidence: limitations of the design, inconsistency, indirectness, imprecision, and publication bias. Then the three factors that increase the quality of the evidence—having a large magnitude of effect, plausible confounding variables, and dose–response gradient—are added. An overall determination can then be made about the quality of evidence. Randomized controlled trials have a high rating, and this is either downgraded or upgraded dependent on the GRADE evaluation.

General description

Of all the interventions, five were conducted in the United States,^30–34 three each in Australia^35–37 and the United Kingdom,^38–40 and one each in Turkey, ⁴¹ The Netherlands, ⁴² Denmark, ⁴³ the Republic of Belarus, ⁴⁴ Singapore, ⁴⁵ and Malaysia. ⁴⁶ One study ³² revealed the author used theoretical underpinnings in the development of his intervention, referencing the social cognitive learning theory of Bandura.^47,48 Two of the 17 trials stated explicitly that they adopted the WHO's definition ¹ of exclusive breastfeeding.^42,46 Two studies used the term “full breastfeeding,” which allows for the consumption of vitamins, minerals, water, juice, medicine, and ritualistic and traditional liquids.^36,37 Nine of the 17 interventions defined exclusive breastfeeding as the infant receiving nothing but breastmilk,^{30,31,34,35,38,41,43–45} with the remaining studies not providing a definition.^32,33,39,40

In the reviewed studies, the ages of participants ranged between 16 and 40 years. The ethnicity of participants varied owing to the international sample. Three studies included a sample of approximately 70% white participants.^31,38,39 Two studies included approximately 70% African American participants.^33,41 The study of Bonuck et al. ³⁰ included 57% Hispanic women and 36% African American women. Tahir and Al-Sadat ⁴⁶ reported a predominantly Malay sample. Furthermore, Kronborg et al. ⁴³ reported that all their participants were Danish, and the studies conducted in Turkey and Jordan did not specifically note ethnicity.^36,41

Six studies included only primiparous women in their sample.^{32,34–36,40,41} One study did not note parity percentage, ⁴⁴ whereas another did not mention participant parity. ³³ For all the remaining studies both primiparous and multiparous women were evenly distributed.^{30,31,34,37–39,41,45,46}

Risk of bias

The average risk of bias across all studies was 2.8 out of a possible 8 using the scoring system previously described (higher scores equal lower risk of bias). No study achieved the maximum score of 8. Only one study received a score of 6, ⁴³ seven received a score of 4,^{4,30,35,36,38,39,45} seven received a score of 2,^{31,32,34,37,40,42,46} and two received a score of 0.^33,41

Intervention components

Details of the intervention methodologies (time of delivery [prenatal and/or postnatal], delivery by hospital staff and/or lactation experts, peer support, telephone support, face-to-face support, number of intervention sessions/duration of intervention, follow-up time points, and findings in relation to success of the interventions for any breastfeeding and for exclusive breastfeeding) are presented in Supplementary Table S2. The majority of interventions focused primarily on providing maternal support^{30,33,34,37,43,46} or education.^{35,36,39,41,42,45} Other methods investigated the use of sidecar cribs attached to mothers' beds, ³⁸ mother–infant skin-to-skin contact, ³¹ breastfeeding logs, ³² or midwife education.^40,44

All four interventions implemented during pregnancy focused on maternal education. This was through antenatal classes,^35,39 lactation consultant/peer counseling, ³⁰ and written and oral information regarding breastfeeding practices and milk storage. ⁴² Of these interventions, two continued into the postnatal period. One provided a single in-hospital and a single home postnatal visit to women and made lactation consultants available until 12 months postpartum. ³⁰ The other provided one postnatal home visit; however, the specific timing was not defined. ⁴² Overall, most education interventions focused on training the mothers by providing advice and teaching skills about exclusive breastfeeding practices. This was provided in both a one-on-one^36,41,42,45 and group^35,39 setting.

Six of the interventions that commenced in the postnatal period provided home/telephone support. Two provided face-to-face home visits,^41,43 three provided telephone support,^34,36,46 and one provided both. ³³ One intervention utilized peer support persons, ³³ and another applied physical stimulation via breastfeeding pumps. ³⁴

Several postnatal interventions provided only in-hospital support to mothers through midwife education strategies,^40,44 mother–infant proximity,^31,38 or single educational sessions and educational material. ⁴⁵ In contrast, Pollard ³² focused on increasing the mother's awareness by encouraging self-monitoring via a daily breastfeeding log; this intervention took place within the participant's home and was self-administered.

Effect of interventions on initiation/duration of exclusive breastfeeding

Of the 17 studies included in this systematic review, eight did not significantly increase the duration and rates of exclusive breastfeeding,^30,34–40 whereas nine demonstrated a statistically significant improvement.^{31–33,41–46} Of these significant results, only seven significantly increased exclusive breastfeeding rates to 6 months postpartum.^{31–33,41–44} Furthermore, only Kramer et al. ⁴⁴ assessed their results for clinical significance; they concluded that their findings represented a substantial clinical reduction in an infant's risk of infection.

Successful interventions

Successful interventions tended to be support-based programs providing additional home-based and telephone support by lactation experts.^33,43 These support interventions commenced in the postnatal period and extended over a relatively long period (from approximately 5 weeks to 6 months). Two interventions were education-based: one focused on providing additional face-to-face postnatal breastfeeding education 3 days after birth, ⁴¹ and the other used written and verbal information in both the prenatal and postnatal periods. ⁴² Other successful interventions used novel approaches for breastfeeding promotion. One used unlimited kangaroo care (skin-to-skin contact), ³¹ and another used self-observation via a breastfeeding log. ³² Additionally, a study compared nurses trained in the WHO and UNICEF Baby-Friendly Hospital Initiative ⁹ with those administering usual care. ⁴⁴

Of all successful interventions, only one commenced in pregnancy ⁴² ; however, this study continued into the postnatal period. Two interventions concluded at hospital discharge.^31,44 Other successful interventions spanned a longer period: from a minimum of 3 weeks ³² up to 5 weeks ⁴³ through to 6 months. ³³ The final successful study designed their program so the participant had only one contact with the intervention. ⁴¹

Unsuccessful interventions

Studies that did not demonstrate a significant increase in the duration of exclusive breastfeeding were generally education only and provided one intervention session to mothers.^35,36,39 Only one of these included follow-up phone calls (at 2 and 4 months after birth). ³⁶ Another unsuccessful intervention educated midwives to use a “hands off” approach to breastfeeding care. ⁴⁰ Studies attributed their negative results to the Hawthorne effect,^36,39 existing high standards of maternal breastfeeding care, ³⁴ and under-recruitment of participants. ⁴⁰

Some support interventions did not demonstrate a significant increase in exclusive breastfeeding. Two provided long-term lactation counseling/care to mothers,^30,37 whereas another assessed the effects of telephone support or the use of a breast pump in obese women.^34,46 One intervention comprised a trial of sidecar cribs attached to mothers' hospital beds. ³⁸ Of those interventions that did not demonstrate an effect on exclusive breastfeeding, two were conducted in pregnancy only,^35,39 whereas the remainder were implemented in the postnatal period.^{30,34,36–38,40,46} Furthermore, the majority of unsuccessful interventions were short-term, taking place during participants' hospital stay^34,38,40 or were single-contact interventions.^35,39 Four interventions provided some ongoing support or education following hospital discharge.^30,36,37,46 Of these latter support interventions, one identified the in-hospital nature of the trial implementation as a limiting factor; it suggested that the intervention may have been more successful had it been implemented within the comfort of participants' own homes. ³⁰ Other results were attributed to existing high breastfeeding rates in the study population ³⁷ and increased awareness of breastfeeding practices caused by weekly reporting. ³⁸ The intervention not being implemented as planned ³⁴ was another stated reason, as well as the unknown combination of multiple factors that influence exclusive breastfeeding at 6 months. ⁴⁶

Intervention control and fidelity

Detailed information regarding study power, intervention fidelity, definitions of usual care, stratification, consideration of maternal body mass index, and intention-to-treat analysis is provided in Supplementary Table S3. Of the 17 included studies, seven either did not provide power calculations, or the studies were underpowered.^{33–35,39–41,45} Of these, four did not provide evidence of effectiveness of the intervention being examined in the trial.^34,37,39,40 Furthermore, in the studies where several individuals were involved in the delivery/implementation of the intervention, the researchers did not assess the fidelity of the administration of the intervention program.^{30,31,33–42,44,45}

Demographic factors such as age, education, socioeconomic status, and parity were measured and considered when sampling and analyzing the data. However, only two studies reported maternal body mass index,^34,46 which is a known risk factor that impacts breastfeeding rates.^49–51

GRADE quality rating of evidence

Given that all included intervention studies were randomized controlled trials, there is potential for a high quality of evidence. However, because of the inconsistent results produced by these studies as well as the limitations of design (lack of power, lack of intervention fidelity, use of some unstandardized measures, and varied definitions of both exclusive breastfeeding and “usual care”), the quality of evidence for breastfeeding interventions and exclusive breastfeeding outcomes is moderate according to GRADE guidelines (see Supplementary Table S4).

What are the limitations of current approaches to exclusive breastfeeding promotion?

Of the interventions that did not produce statistically significant results, one was underpowered to detect differences between intervention and control groups, ⁴⁰ and another did not provide a power calculation. ³⁴ Insufficient participant numbers were likely to have had a detrimental effect on the ability of these studies to detect a change if one existed and may in part explain the lack of evidence of effective interventions.

Interventions were inconsistent in their definition of “usual care,” and 10 did not even provide a definition^{32,34,36,38–40,42,44–46} (see Supplementary Table S3). In trials where usual care was defined, there was substantial variation: from one midwife-implemented education session ⁴¹ to accessing a large number of resources as well as postnatal home visits.^33,35,37 This lack of a common definition of “usual care” makes the comparison of the intervention's effectiveness difficult. In an instance where “usual care” is very supportive, the numbers needed to detect a statistical difference would need to be inflated.

In cases where several research team members were involved in intervention implementation, no details were provided of measures undertaken to assess the administration of the intervention program. One study did attempt to stratify participants by midwife experience in an effort to reduce variation. ⁴⁰ However, sample size appears not to have been adjusted for stratification, and whether this intervention was implemented in a consistent way is unknown. Of all interventions included in the current review, 10 were analyzed on an intention-to-treat basis.^{30,34–39,43–45} Of these interventions only three were successful at increasing exclusive breastfeeding practices,^36,43,45 with only two of these significantly increasing exclusive breastfeeding to 6 months postpartum.^36,43 Furthermore, four of the reviewed studies that included more than one service provider did not stratify participant randomization across maternity services.^30,31,36,42

The majority of the studies reviewed here did not report maternal body mass index (see Supplementary Table S3). Evidence indicates that women who are classified as obese have increased difficulty in breastfeeding, including positioning, ¹⁵ and a delay in lactation^49,60 and discontinue breastfeeding sooner than women of normal weight,^47,48 often within 2 weeks postpartum. ⁶¹ This increased difficulty along with the growing concern of global obesity rates complicates the challenge of complying with WHO breastfeeding recommendations. ⁶²

Future research

Our review has highlighted limitations in the research to date that prevent definitive conclusions about intervention efficacy. It is clear there is a need to continue research in this area to develop effective interventions that promote exclusive breastfeeding in women. Furthermore, this research needs to be adequately powered to detect a significant increase in the proportion of women breastfeeding exclusively to a designated time point after birth (preferably to 6 months after birth, in line with WHO recommendations). ¹ The design of new interventions needs to be guided by theory for the greatest chance of a successful intervention.

There are several specific areas where further research is needed. This includes ways to better understand and promote the efficacy of interventions delivered within one's home, which were shown to be effective. Recognizing that maternal body mass index has a measurable effect on exclusive breastfeeding duration, it becomes even more important that breastfeeding interventions are robust and potentially specialized for overweight and obese women. Interventions that focus on overcoming the barriers of breastfeeding to 6 months, including maternal perceptions, embarrassment, and family support, among others, are also very important.