Effect of Neonatal Unit Interventions Designed to Increase Breastfeeding in Preterm Infants: An Overview of Systematic Reviews

Breastmilk contains nutrients to support growth and immune factors that prevent infection [1] and inflammatory disorders [2, 3]. It is, therefore, the recommended source of early nutrition for all infants [4]. For preterm infants (born <37 weeks’ gestation), early breastmilk feeds are particularly important, as they can increase the rate of intestinal maturation, improve feed tolerance, reduce prematurity complications including sepsis [1, 5, 6], necrotising enterocolitis [6‒8], and bronchopulmonary dysplasia [9], and promote neurological development [10, 11].

There are serious challenges to establishing and maintaining breastmilk feeds in preterm infants in neonatal units, particularly those admitted to neonatal intensive care units (NICUs). For all mothers intending to breastfeed, a supply of breastmilk must be established and maintained, facilitated by close maternal/infant contact, yet preterm infants are often separated from their mothers. Establishing supply relies primarily on frequent artificial breast stimulation for a prolonged period, until the infant transitions to direct breastfeeding when developmentally ready. Compounding factors affecting milk production include maternal medical complications, stress, limited lactation education, inadequate support from health care professionals and lack of promotion of kangaroo mother care (KMC) [12, 13].

Therefore, whilst most preterm infants are started on breastmilk feeds, there are high rates of cessation of breastfeeding before and after hospital discharge. For example, a retrospective cohort study [14] based in 10 Australian NICUs showed a breastfeeding rate among preterm infants at hospital discharge of 64%, compared with >90% in term infants [15]. Lower rates of breastmilk feeding have been demonstrated in very preterm [16] and late preterm infants [17]. Further, there is evidence of considerable variations in preterm infant breastfeeding rates both within and between countries internationally [14, 16, 17]. This suggests there are modifiable practices that could be addressed to improve breastfeeding in this high-risk population [18].

There is a substantial systematic review literature, including meta-analyses, reporting effects of interventions designed to encourage and sustain breastmilk feeding in preterm infants [19‒25]. However, to date no review has brought together the body of evidence concerning all available interventions nor undertaken a formal examination of quality of the reviews [26]. Here, we present an overview of reviews reporting pooled effects of interventions designed to encourage breastfeeding and improve breastfeeding rates in preterm infants admitted to neonatal units, at discharge and up to 6 months thereafter.

We followed Cochrane overview methodology [27] and applied the Cochrane Comparing Multiple Interventions Methods Group Editorial Decision Tree to determine that this review was better suited to an overview than intervention review. The overview protocol was prospectively registered (CRD42023424321 [28]).

Systematic reviews published in English from database inception to October 31, 2022, reporting effects of interventions designed to support breastfeeding in preterm infants admitted to NICUs, at discharge or post (up to 6 months thereafter) were included. To be included, a review must have reported a summary relative risk measure with 95% confidence interval (CI) (e.g., odds ratio, risk ratio) for one primary outcome. Reviews including studies in preterm infants and/or infants with a birthweight suggesting prematurity (<2,500 g), in any neonatal unit setting were eligible. We considered reviews involving term and preterm infants, provided they reported separately on preterm infants. We considered all types of interventions in NICUs designed to increase breastfeeding, standalone or in bundles, and comparators as reported.

Primary outcomes were:

• Breastfeeding (exclusive or any) at discharge home

• Breastfeeding (exclusive or any) at any time from discharge up to 6 months post-discharge

Breastfeeding, unless otherwise specified, refers to breastmilk feeds in preterm infants directly at the breast (direct breastfeeding) or by another method (tube, bottle, syringe, specified when available, referred to as indirect breastfeeding). Breastmilk was defined as mother’s own breastmilk.

Secondary outcomes were:

• Adverse events

  • o Mortality

  • o Episodes of “infection,” including sepsis and NEC

• Health service use

  • o Length of hospital stay

  • o Rehospitalization

• Infant growth

  • o Weight

  • o Length

  • o Head circumference

• Infant feeding

  • o Breastmilk expression: quantity of milk expressed

  • o Nutrient quality of milk expressed (e.g., fat, sodium, calories)

  • o Use of dummy/pacifier

• Parental satisfaction (e.g., mother’s view, any scale, as reported)

We systematically searched electronic databases for relevant reviews with English language the only restriction, on October 31, 2022, and searched references of included reviews (see online suppl. material S1 for databases and search strategy; for all online suppl. material, see https://doi.org/10.1159/000536660). Two authors independently assessed records retrieved and extracted data from included reviews. Disagreements were resolved by consensus or discussion with a third author. These authors also independently assessed the methodological quality of the included reviews using the Risk of Bias in Systematic Reviews (ROBIS) tool [29] (see online supplementary file S2 for details on characteristics and statistical summaries extracted).

Using the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) approach, we assigned an evidence quality rating (high, moderate, low, or very low) to intervention effects on primary outcomes [30‒32]. When available, we extracted and reported the assessments of the authors of the included reviews. When reviewer assessments were incomplete or unavailable, two authors, working independently, added to, or conducted the assessments, following GRADE principles. Subsequently, we summarised these outcomes in “Summary of findings (SOF)” tables. We did not reassess risk of bias of studies included within reviews. We collected this information during the data extraction process.

We summarised effects in narrative and tables organised by comparison and outcome. Following the framework adopted in a recent overview [33], we assigned graphic icons to effects shown for primary outcomes, to communicate direction of effect estimates and our confidence in available data (see Fig. 1). For clear benefit or harm, the CI associated with effect size did not cross the line of no effect. For clear evidence of no effect, we considered a CI approximating the relative risk range 0.75–1.25 as sufficiently narrow to indicate a minimal effect relative to the comparator; thresholds recommended by GRADE [33]. A p value was <0.005 was considered significant. We did not aim to examine indirect comparisons or report network meta-analyses.

From 982 records retrieved, we included seven reviews [19‒25]. See PRISMA flow chart (online suppl. Fig. S1) for details about search results and study selection.

Overall, 86 studies (70 trials, 16 QI studies), conducted in a wide range of countries, most high or middle income, were reported in the seven reviews. The number of preterm and/or low birthweight participants in the reviews ranged from 335 [25] to 33,172 [31]. There was some overlap between reviews in included studies. Specifically, one trial [35] was included in three [19, 22, 23]. Additionally, the five RCTs in Flint 2016 [22] were included in Allen 2021 [19]. Another two trials [36, 37] were reported in two reviews [23, 24].

Gestational age (GA) and birthweight of the included infants varied between reviews. In Allen 2021 [19], mean GA was 32 weeks. Flint 2016 [22] included studies enrolling infants with mean GA between 29 and 35 weeks. Foster 2016 [23] included 12 studies enrolling infants with mean GA ranging from 28 to <34 weeks and birthweight ranging from 1,076 to 1,676 g. Conde 2016 [20] included infants with mean birthweight 968–2,076 g and did not report mean GA of included infants. We assumed, given the mean birthweight, that most included infants were preterm. In Fang 2021 [21], six included studies enrolled infants with GA between 22 and 37 weeks, and 11 included infants with birthweight <1,500 g; mean GA and birthweight were not reported. Nasuf 2018 [25] included infants with GA ranging from 25 to 32 weeks’ gestation, and birthweight 410–2,500 g. Greene 2016 [24] included infants <37 weeks’ gestation and did not report mean GA or birthweight of infants in included trials.

The reviews reported breastfeeding outcomes for seven standalone interventions: Avoidance of bottles during transition to direct breastfeeds in mothers intending to breastfeed (by utilising alternative feeding devices to supplement breastfeeds, including gavage tube, cup, spoon, dropper, finger feed, paladai, and other); KMC (any timing); early onset KMC (commencing <24 h post-birth); cup feeding; non-nutritive sucking (NNS) involving use of a pacifier or other method (e.g., gloved finger); oral stimulation (by finger stimulation); and oropharyngeal colostrum (OPC) administration to the buccal mucosa in the early neonatal period. Additionally, a bundle of quality improvement (QI) interventions was assessed. Interventions in the bundles included: multidisciplinary expert teams; development of evidence-based interventions; education of hospital staff; parental education (prenatal consultations and postnatal education); increased availability of pumps; initiating early breastmilk expression; OPC administration; lactation consultant tracking of visits or phone calls; skin-to-skin care; NNS; human milk management; standardised enteral feeding guidelines; preparation for discharge (family-integrated care, transition to direct breastfeeding); post-discharge lactation support and follow-up care.

Whilst for three of the interventions (avoidance of bottles, cup feeding and oral stimulation) pooled effect estimates on direct breastfeeding were reported, for two (NNS and QI bundle), the effect estimates for our primary outcomes relate to indirect or direct breastfeeding. For the remaining interventions (KMC, early KMC, OPC), it was not specified whether the pooled effects reported for our primary outcomes relate to direct breastfeeding only, or also indirect.

All reviews were assessed at low overall risk of bias (considering the assessments across all ROBIS domains), and five at low risk of bias in all domains [19, 20, 22, 23, 25]. Fang 2021 [21] was downgraded from low to unclear risk of bias in two domains due to lack of detail about adjustment in the included pre-post QI studies. Greene 2016 [24] was downgraded to unclear risk of bias in one domain, due to lack of detail on the rationale for the evidence certainty rating assigned to the reported breastfeeding outcome.

Table S1 in the online supplementary material provides further details on the review characteristics. Review risk of bias assessments are shown in online suppl. Table S2, and online suppl. Table S3 provides the original study bias assessments.

Breastfeeding at Discharge (data shown in Table 1 and GRADE Assessment in online suppl. Table S4) Avoidance of bottles (utilising alternative feeding devices to supplement breastfeeds) versus breastfeeds complemented with bottles during transition to breastfeeds: For exclusive breastfeeding, low-quality evidence indicated possible benefit (the pooled effect estimate CIs did not cross the line of no effect and favoured the intervention, see Fig. 1) of avoiding bottles to supplement breastfeeds when compared with breastfeeds complemented with bottles during transition to breastfeeds. For any breastfeeding, there was moderate-quality evidence suggesting clear benefit when these interventions were compared. Subgroup analysis by intervention type was not statistically significantly different for any breastfeeding (χ² = 5.09, df = 2 [p = 0.08], I² = 60.7%) or exclusive breastfeeding (χ² = 3.52, df = 2 [p = 0.17], I² = 43.1%).

KMC versus conventional neonatal care: Moderate-quality evidence indicated a possible benefit of KMC on rates of any breastfeeding and exclusive breastfeeding.

QI bundle versus no QI bundle: Low-quality evidence showed a possible benefit of QI bundles aimed at increasing breastfeeding, compared to standard care on rates of any breastfeeding. For exclusive breastfeeding, there was very low-quality evidence showing unknown benefit when these interventions were compared. Heterogeneity between studies was substantial, I² >50%, thought to reflect different content of QI bundles between studies.

Cup feeding versus other forms of supplemental enteral feeding: Low-quality evidence showed possible benefit of cup feeding in preterm infants unable to breastfeed fully compared to utilisation of other types of supplemental feeds in rates of not exclusively breastfeeding. For rates of not any breastfeeding, very low-quality evidence showed unknown benefit or harm when these interventions were compared.

NNS involving the use of a pacifier or other method versus standard care: Very low-quality evidence showed unknown benefit or harm of NNS in preterm infants compared to standard care in exclusive breastfeeding rates.

Oral stimulation:

• Oral stimulation versus standard care or no intervention: Very low-quality evidence showed unknown benefit or harm of oral stimulation in preterm infants compared to standard care in rates of exclusive breastfeeding and any breastfeeding.

• Oral stimulation versus non-oral stimulation: Very low-quality evidence showed unknown benefit or harm with oral stimulation in preterm infants compared to another non-oral intervention in exclusive breastfeeding rates.

OPC versus control (water, saline, or no intervention): Very low-quality evidence showed unknown benefit or harm of OPC in preterm infants compared to a control in rates of any breastfeeding.

Breastfeeding Post-Discharge (data shown in Table 2 and GRADE Assessment in online suppl. Table S5) Avoidance of bottles (utilising alternative feeding devices to supplement breastfeeds) versus breastfeeds complemented with bottles during transition to breastfeeds: Avoidance of bottles showed a possible benefit in rates of full breastfeeding at 3 months and 6 months post-discharge and of any breastfeeding 3 months post-discharge and 6 months post-discharge. Subgroup analysis by intervention type was not statistically significant, for full breastfeeding 3 months’ post-discharge (χ² = 2.35, df = 2 [p = 0.31], I² = 14.7%), full breastfeeding 6 months’ post-discharge (χ² = 3.45, df = 1 [p = 0.06], I² = 71.0%), any breastfeeding 3 months’ post-discharge (χ² = 2.16, df = 2 [p = 0.34], I² = 7.3%), and any breastfeeding 6 months’ post-discharge (χ² = 3.46, df = 1 [p = 0.06], I² = 71.1%). However, the effect on breastfeeding outcomes was evident at all time points for tube feeds and for all except any breastfeeding at 3 months post-discharge for cup feeding.

Kangaroo Mother Care:

• KMC versus conventional neonatal care: For this comparison, low-quality evidence showed possible benefit on rates of any breastfeeding at 3 months post-discharge and 1–3 months post-discharge. Very low-quality evidence showed unknown benefit or harm on rates of any breastfeeding at 6 months’ post-discharge and of full breastfeeding at 1–3 months’ post-discharge. Subgroup analyses demonstrated intermittent KMC was associated with a statistically significant increase in any breastfeeding at 6 months’ post-discharge.

• Early-onset KMC versus late-onset KMC: Very low-quality evidence showed unknown benefit or harm on rates of full breastfeeding at 6 months of age.

Cup Feeding versus Other Forms of Supplemental Enteral Feeding: Low-quality evidence showed possible benefit of cup feeding in preterm infants unable to fully breastfeed compared to other types of supplemental feeds in rates of not any breastfeeding at 3 months. For rates of not any breastfeeding at 6 months, very low-quality evidence showed unknown benefit or harm for this comparison.

Adverse events: KMC compared to conventional care showed reduced mortality and sepsis, but early-onset compared to late-onset KMC no difference in either. No difference was seen in episodes of NEC between OPC and a control, or episodes of infection between avoidance of bottles-feeds and breastfeeds with bottle.

Health service use: Avoidance of bottle-feeds compared to breastfeeds supplemented with bottle, KMC compared to conventional care, early-onset compared to late-onset KMC, cup feeds compared to other forms of supplemental enteral feeds, oral stimulation compared to standard care, and OPC administration compared to a control showed no evidence of difference in length of hospital stay. NNS compared to standard care showed reduced length of hospital stay. KMC compared to conventional care showed reduced rehospitalisation rates, but early-onset compared to late-onset KMC no difference in this outcome.

Infant growth: Avoidance of bottle-feeds compared to breastfeeds with bottle-feeds, early-onset compared to late-onset KMC, NNS compared to standard care, oral stimulation compared to standard care, and OPC administration compared to a control showed no difference in growth outcomes. KMC compared to conventional care showed an increase in weight gain, length, and head circumference at latest follow-up.

Parental satisfaction: When compared to conventional care, KMC showed an increase in satisfaction scores.

This is the first overview using established methodology to identify, assess for quality, and interpret meta-analysis of effects of neonatal unit interventions targeted at enabling and sustaining breastfeeding in preterm infants. The evidence showed a clear or possible benefit, defined as the ability to increase breastfeeding (exclusive or any) at discharge or up to 6 months after, for four interventions: avoidance of bottles during breastfeeding establishment (i.e., use of alternate feeding devices, e.g., gavage tube, cup, spoon, dropper, finger feed, paladai, and other), KMC (any time started following the birth), a bundle of interventions implemented as part of QI, cup feeding, and early-onset KMC. For the remaining interventions, NNS, OPC, and oral stimulation, the evidence was insufficient to assign a level of effectiveness. Importantly, and not surprisingly, we did not identify any evidence of harm of the interventions for either of our primary outcomes (i.e., a clear or possible reduction in breastfeeding rates at or after hospital discharge).

We were able to include seven systematic reviews with pooled analysis, of which six included trials in their pooled analysis for our primary outcomes (randomised or quasi), and one included pre-post intervention QI studies. There was high variation between intervention comparisons in the number of studies and participants.

This overview suggests that if neonatal units implemented at least one of the interventions with clear or potential benefit, there could be an improvement in breastfeeding (direct or indirect) at discharge in the order of 11–64%. However, the meta-analysis assessing the bundle of breastfeeding interventions could not discern which interventions within QI bundles are key for beneficial impact on breastfeeding rates or are likely to have the greatest ability to enable and sustain breastfeeding in preterm infants.

Additionally, none of the meta-analyses stratified by GA or birthweight groups. Thus, whether treatment effects of the beneficial breastfeeding support interventions vary by GA or birthweight is unknown.

Available meta-analyses in seven high-quality systematic reviews, suggest that the standalone interventions: avoidance of bottles during breastfeeding establishment, KMC (any), early-onset KMC, and cup feeding, and additionally, a QI bundle designed to support breastfeeding, can increase breastfeeding rates in preterm infants. It is unclear which are most effective, or what bundles are likely to work best (and in which setting or preterm infant populations). There is insufficient pooled evidence to establish impacts on breastfeeding rates among preterm infants of NNS, oral stimulation, and OPC. However, this does not mean these interventions are not of benefit.

Updates of the meta-analyses assessing the effects of NNS, OPC, and oral stimulation on breastfeeding rates in preterm infants could be helpful. Systematic reviews with meta-analysis are required to determine the effectiveness of other well-endorsed strategies or promising interventions for which we did not identify any pooled effects, such as medications aimed at increasing milk supply, donor human milk, lactation consultant support, staff education, maternal education, early initiation of breastmilk expression, and use of pacifiers/dummies.

Future research on this topic should aim to stratify effects by GA and birthweight subgroups. Investigating the effects of standalone interventions to enable and sustain breastmilk feeds in preterm infants admitted to neonatal units and different types of intervention combinations (bundles) in different settings is important, as knowledge is required on key features of effective bundles and which interventions will work best in what settings.

We were aware of risks of introducing bias at all stages of the overview process and took several steps to minimise this, using best practice overview methodology. A potential source of bias relates to authors of this review being authors of one of the included reviews¹⁹. Data extraction and quality assessment for all reviews were carried out by two overview authors who were not authors of any included review.

Our evidence selection criteria were broader than is typical in an overview of evidence of intervention effects, as we did not limit review eligibility on the type of original study design considered. This is a strength, as it allowed inclusion of all the summary data on our topic.

However, we chose to limit review eligibility to those with pooled analysis, with the intention of offering a summary of high-quality evidence (including precise effect estimates) to inform strategy and research prioritisation. Thus, there are systematic reviews with narrative summaries of effects on breastfeeding rates among preterm infants of relevant interventions in neonatal units that we did not assess in this overview, and which may be promising.

While we summarised the limited pooled effects of neonatal unit interventions on several secondary outcomes, including adverse events, health service use and infant growth, the secondary outcome effects reported must be interpreted cautiously. This is because we chose to select reviews for inclusion based on whether they reported a pooled effect for at least one of our primary outcomes. We did not evaluate certainty of evidence for the secondary outcomes. As is common in all systematic reviews, there is a risk of publication bias due to the possibility that there may be relevant data not included due to relevant publications since the last date of our search.

Avoidance of bottles during breastfeeding establishment (use of alternate feeding devices), KMC (any), early-onset KMC, cup feeding during the establishment of breastfeeding in preterm infants, implemented as standalone interventions in neonatal units, as well as QI bundles focused on enabling breastmilk use in preterm infants, can increase rates of breastfeeding in preterm infants at discharge and in the following 6 months. Further research is required to determine key features of effective QI bundles and standalone interventions (such as NNS, OPC, and oral stimulation) not shown to be of possible or clear benefit. Research should be designed to explore potential modification by GA at birth and birthweight subgroups.

The authors would like to thank Natalie Dempster for contributing to the development of the search strategy for this systematic review and conducting the search of bibliographic databases.

The authors have no ethical conflicts to disclose. An ethics statement is not applicable because this study is based exclusively on published literature.

The authors have no conflict of interest to declare.

A/Professor Keir is in receipt of National Health and Medical Research Council (NHMRC) Fellowships (APP1161379). Dr Alice Rumbold is in receipt of NHMRC Funding: Centre of Research Excellence (2024589) and NHMRC Clinical Trials and Cohort Funding (2024018). The views expressed in this article are solely the responsibility of the authors and do not reflect the views of the NHMRC. Dr Hilditch receives support through an Australian Government Research Training Program Scholarship. The funders had no role in the design, data collection, data analysis, and reporting of this study.

C.H. conceptualised the protocol, designed the data collection forms, selected studies for inclusion, extracted data, assessed risk of bias and certainty of evidence, drafted the initial manuscript, and reviewed and revised the manuscript. A.R. conceptualised the protocol, supervised risk of bias, certainty of evidence assessments, and data analysis interpretation, and reviewed, and revised the manuscript. A.K. and P.M. conceptualised the protocol and reviewed and revised the manuscript. J.G. conceptualised the protocol, supervised study selection, data extraction, risk of bias and certainty of evidence assessments, and data analyses, and drafted, reviewed, and revised the manuscript. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

All data generated or analysed during this study are included in this article and its supplementary material files. Further enquiries can be directed to the corresponding author.