Selection and study characteristics
A search of five databases yielded 4,817 studies out of which 164 were selected for full text review, with 119 reporting breastfeeding practices as intervention outcomes. Only 16 studies targeted LBW and VLBW infants. Figure 1 illustrates the study selection process.
[Figure 1 Caption: Identification and selection of studies
Please insert Figure 1 here]
The studies, published between 1991 (32) and 2023 (33), were conducted only in two countries of South Asia; Bangladesh (34,35) and India (32,33,36–47). There are 12 RCTs, three quasi-experimental (32,34,46), and one pre-post study (47). All interventions were conducted in urban areas, except one in a semi-urban area (44). Most studies were conducted in hospital/facility setting, one in community setting (44) and in one study, control group was in a facility setting whereas the intervention group was in a community setting (42). Although all participants were LBW infants, only four studies (33,37,38,42) enrolled VLBW infants. Characteristics of the included studies are outlined in Table 2.
Intervention characteristics
Five different types of interventions were found across the included studies: KMC, Non-nutritive sucking, Mother Baby Friendly Initiatives Plus (MBFI+), Nutrition education and Premature Infant Oromotor Intervention (PIOMI). KMC was found as the core intervention component across 12 studies (34,36–39,41–47). However, the overall dose and delivery mode of the KMC differed in terms of practice period, setting (facility vs. community), practice frequency, intervention arms and material supported with, for example, cap, sock, diaper, reclining bed, chair, kangaroo bag etc. (Supplementary file, item 4). A Non-nutritive sucking (emptied breast vs. finger) method was used in three studies (32,33,40), MBFI+ was conducted in a before-after (pre-post) study with no control group (47). Nutrition education focused on exclusive breastfeeding was implemented by one study (35) and Premature Infant Oromotor Intervention (PIOMI) was also implemented by one study (33).
KMC was integrated within the MBFI+ method (47) and non-nutritive sucking was a component of the Premature Infant Oromotor Intervention (PIOMI) program (33). Notably, all intervention was initiated once the infants were clinically stable, which included the inability to suckle at the breasts. The criteria for intervention participants also varied across the interventions, with details provided in Supplementary file (item 4).
Intervention duration and follow-up duration varied across the interventions. For example, two studies used KMC with varying lengths of intervention and follow-up to promote exclusive breastfeeding (36,37). One study (36) implemented a seven-day KMC intervention with a seven-day follow-up (36), while another study continued KMC until infants achieved a specific weight gain rate and followed-up until 40 weeks of gestational age (37). Similarly, the methodologies used to quantify outcomes also vary across different studies. One study (34) used face-to-face interviews and observations to evaluate the initiation of breastfeeding, another study (41) solely relied on observations despite both studies using KMC as an intervention.
Intervention effects
Initiation of breastfeeding
Table 3 outlines the studies, interventions and effects of those interventions on the initiation of breastfeeding. Of the studies that used KMC, three reported reductions in time for initiation of breastfeeding in the intervention groups (34,41,46) and one (45) reported a longer duration to initiate breastfeeding (3.76±4.49 days in the experiment group vs. 2.95±3.85 days in the control). The other interventions, PIOMI, Non-nutritive sucking and Nutrition education reported positive outcomes for the initiation of breastfeeding. When PIOMI was used, the preterm oral feeding readiness assessment scale (PORFAS) score was higher (33), non-nutritive sucking (emptied breast) cut down on the time it took to start breastfeeding (40), and nutrition education helped more mothers (59.8% in intervention vs. 37% in control) initiate breastfeeding (35).
Exclusive breastfeeding
Table 4 provides a summary of the success of various interventions, including KMC, Non-nutritive sucking, Nutrition education, and MBFI+, in encouraging exclusive breastfeeding. Out of the seven studies that used KMC (including early KMC and community initiated KMC), two studies (39,45) found that exclusive breastfeeding was achieved in a shorter amount of time. The remaining five studies (34,36,38,42,44) reported a higher percentage of infants exclusively breastfed at specific time intervals.
Non-nutritive sucking on the infant’s finger was found to promote exclusive breastfeeding at six weeks, three months, and six months for a higher proportion of infants compared to non-nutritive sucking on the emptied breast (40). In contrast, another study (32) on sucking emptied breasts reported a higher proportion of infants exclusively breastfed compared to sucking full breasts. Though the mean duration of exclusive breastfeeding (in a four-month follow-up period) in both emptied and full breast cases was the same (3.7±1.3 months).
Nutrition education effectively promoted exclusive breastfeeding among infants, with 59.8% of recipients achieving this after two months of intervention, compared to 37% who did not (35). The MBFI+ intervention resulted in a 20% increase in exclusive breastfeeding infants, from 60.5% before to 80.7% after the intervention (47).
Other breastfeeding practices
Apart from initiation of breastfeeding and exclusive breastfeeding, other breastfeeding outcomes such as colostrum feeding, breastfeeding rate, breastfeeding frequency, full/partial breastfeeding and total lactation were identified. Their details are summarized in Table 5.
Nutrition education (35) and MBFI+ intervention (47) foster colostrum giving to infants. However, KMCs improve breastfeeding rate and breastfeeding frequency (37,46) and full/partial breastfeeding compared to conventional medical care (CMC) (37) and increases the Bristol Breastfeeding Assessment Tool (BBAT) score, which quantifies overall lactation, particularly attachments and positioning (43). The mean duration of total lactation was higher among recipients of non-nutritive sucking of empty breast compared to nutritive sucking of full breast when both groups received an additional intermittent bolus via the orogastric route (32).
Meta-analysis
The pooled Standard Mean Difference (SMD) for duration of initiation of breastfeeding was lower (-1.08 days, 95% CI: -2.15 days to -0.01 days, p<0.001) for infants receiving KMC interventions compared to those who did not (Figure 2). The high heterogeneity (I2=95.8%, p<0.000), uncertainty (R2=97.39%) and publication bias (Egger p=0.0001) was detected in the meta-analysis (supplementary file, item 3). While exploring the heterogeneity by meta-regression using both univariate and multivariate model, the country of the study setting, and study types (RCT, Quasi-experimental and non-RCT) were found to have significant associations (Supplementary file, item 3).
[Please insert Figure 2 here]
Risk of bias, quality assessment and strength of assessment
The biasness of the included studies was measured by the RoB2 tool for RCTs, and ROBINS-I tool for quasi-experimental studies. Of three non-randomized control trails evaluated, one showed moderate risk of bias, while two other studies demonstrated serious concerns regarding bias. Among 12 RCTs, four studies were identified having some concerns of bias and eight studies were assessed as having high risk of bias. Some domains, such as deviations from the intended interventions and bias in the measurement of outcomes were found susceptible among the included articles. However, other domains such as bias due to missing outcomes, randomization process (RCTs only), selection of reported results (for non-RCTs only) were relatively well described. One study (47), which was assessed by NIH Quality Assessment tool for before and after studies with no control group, was found to have good quality. Details of the quality assessment/risk of bias are presented in the supplementary file (item 5).