An initial search across six databases identified 301 records (172 after duplicates removed). Title and abstract analysis excluded 140 studies, leaving 32 potential studies - narrowed to nine studies after reading full text. Most excluded studies did not focus specifically on LIW, used a teleintervention to facilitate in-person visits or provided teleinterventions to both the control and intervention group (for further details see Additional File Table 2). Study selection process is outlined in Fig. 1.

PRISMA flowchart of study selection process

All the nine studies included in this review were conducted in the US between 2002 and 2020. Together they included 3522 mothers [27, 27, 31,32,33,34,35,36,37,38]. Most studies focused on ethnic minorities (Hispanic or African American), either by design [34, 38] or by virtue of area demographics [27, 31, 32, 39]. Palacios et al. covered mainly White and Hispanic mothers [35] and only one study had a majority of White mothers [33]. In all studies the mean maternal age was between 20 and 30 years, lower than that in general population in many HIC (30–34 years) [40].

Most studies did not collect data on factors known to affect breastfeeding (parity, delivery mode and previously breastfeeding), limiting results’ comparability [18]. Six studies did not record delivery method [27, 32, 33, 35, 37, 38], three did not include parity [27, 31, 33] and just three recorded previous breastfeeding experience [37,38,39]. Only one study recorded all the three factors [39]. Study characteristics are outlined in Table 1.

Most teleinterventions were phone based; six studies delivered breastfeeding support through telephone calls [27, 31, 33, 34, 38, 39] and two utilised text messages [35, 37]. Only Fik et al. assessed a complex web-based support group with online sessions, regular posts, and psychological support [32]. Of the nine included trials, four were delivered postpartum [27, 33, 34, 39], and the other five were conducted during both the antenatal and postpartum periods. The definition of ‘exclusive breastfeeding’ varied between studies and was not reported by two papers who instead recorded ‘any’ or ‘predominant’ breastfeeding [34, 39].

All studies included in this review had a high risk of bias, represented in Fig. 2A and B (full rationale presented in Additional File Table 4). Three studies provided insufficient information on the allocation sequence generation and implementation, raising ‘slight concerns’ of selection bias [31, 33, 37]. Two studies were judged as having ‘serious concerns’ of performance bias as they did not specify if the data collector was an external agent (not the peer support worker) [27, 34]. Trial protocols for three studies were unavailable and no protocols included a full analysis plan [27, 33, 38]. The consequential lack of a pre-publication analysis plan raises concerns of reporting bias in all trials. Additionally, three studies had ‘serious concerns’ of selective reporting due to protocol deviations [31, 38] or insufficient analysis information [27]. Funnel plots were not used to assess publication bias, as they have a low predictive power with < 10 studies.

‘Risk of Bias‘ summary

The number of mothers who exclusively breastfed for six months was only measured by five of the included studies [27, 31,32,33, 38]. Three studies recorded EBF at one month, all of which assessed supportive phone calls [31, 33, 38]. Two provided postpartum peer support and one delivered ante- and postnatal phone education from lactation educators [38]. The average effect from the pooled results indicates a modest breastfeeding increase, with borderline statistical significance (Fig. 3).

Effect of teleinterventions on exclusive breastfeeding

Four studies recorded EBF at 3–4 months, and their pooled results indicate teleinterventions may marginally increase EBF (Fig. 3). This effect was slightly attenuated following a sensitivity analysis which excluded Efrat et al.’s study due to the high risk of attrition bias (RR 1.10, 95% CI 0.97–1.24) [38].

Pooled results (Fig. 3) at 6 months show a beneficial but non-statistically significant effect on EBF, which almost disappeared in a sensitivity analysis excluding studies with particularly high attrition (RR 1.01, 95% CI 0.85–1.2) [33, 38].

Definitions of ‘partial breastfeeding’ varied between studies [34, 38]. To standardise pooled results this meta-analysis used the subcategory that included all breastfeeding mothers (exclusive, any or partial) from each study.

Five studies (all providing supportive phone calls) reported breastfeeding at one month [27, 31, 33, 34, 38]. Pooled results indicate these significantly increased breastfeeding (Fig. 4). A sensitivity analysis including only the three studies without a high risk of attrition bias enhanced this effect (RR 1.16, 95% CI [1.09,1.24], P < 0.0001) with minimal heterogeneity (I2 = 0%, P = 0.8).

Effect of teleinterventions on breastfeeding

Seven studies reported breastfeeding between 3 and 4 months [27, 31, 34, 35, 37,38,39]. Of these, one assessed passive educational text messages [37], one two-way motivational texts [35], two evaluated nurse phone calls [34, 38], and three provided telephone peer support [27, 31, 39]. On average, they did not increase breastfeeding at 3–4 months postpartum (Fig. 4). Heterogeneity was high and therefore the studies were divided into subgroups based on the main intervention component (education or peer support)(Fig. 4) [27, 31, 39].

Four studies utilised educational teleinterventions [34, 35, 37, 38]. These included uni- (where the mother could not respond) and bi-directional (where responses from the mother were answered) text messages [35, 37] and phone calls from lactation educators [38] or nurses [34]. One study provided just 2 weeks of postpartum nurse calls [34] whilst the other three were started antenatally and continued for > 4 m [35,