Figure 7, C; P-h < 0.01, I2 = 97.25). There was more consistency for breastfeeding continuation with most studies showing a positive relationship between continuation of breastfeeding and multiparity (Table XII, Figure 8, C; P-h < 0.01, I2 = 68.69). Mother-infant dyad separation generally focused on 2 outcomes: early skin-to-skin contact or rooming-in during the birth hospitalization. Taken together as a single metric, positive dyad connections (skin-to-skin vs not or rooming-in vs not) were associated with increased initiation (Table XII, Figure 7, D; P-h < 0.001, I2 = 95.54) and continuation (Table XII, Figure 8, D; P-h < 0.01, I2 = 76.59) of breastfeeding. Sensitivity analyses of studies examining skin-to-skin and rooming-in separately similarly found that the association was stronger for initiation. For skin-to-skin alone, the summary RR was 1.79 (95% CI 1.30-3.11; P-h < .001; I2 = 93.1) for initiation (7 studies) and 1.14 (95% CI 1.07-1.86; P-h = 0.001; I2 = 75.62) for continuation (6 studies). Models limited to rooming-in had a summary RR for initiation of 2.20 (95% CI 1.44-3.36; P-h < 0.001; I 2 = 87.05) in 7 studies and 1.04 (95% CI 0.93-1.16; P-h 0.93; I 2 < 0.01) for continuation in 3 studies. Measurement and categorization of maternal education was variable and often dependent on the country in which the study was conducted. For analysis, the highest education level vs the lowest was used. Although the magnitude of effect varied across studies, the direction was consistent with nearly all individual studies showing a higher likelihood of breastfeeding initiation and continuation among women with higher vs lower levels of education. The summary RRs for breastfeeding initiation (Table XII, Figure 7, E; P-h < 0.01, I2 = 92.27) and continuation (Table XII, Figure 8, E; P-h < 0.001, I2 = 97.06) were higher for women with the highest level of education vs the lowest. We grouped interventions directed specifically at mothers or parents to increase knowledge and confidence around breastfeeding into the factor “breastfeeding education.” These included attendance at prenatal breastfeeding classes, peer counseling on breastfeeding, and lactation consultation before or after delivery (full listing of measurements shown in Table XI). Interventions at the clinic or hospital-level or those directed at clinicians were excluded. Receiving some type of breastfeeding education or support was positively associated with increased likelihood of either breastfeeding initiation (Table XII, Figure 7, F; P-h < 0.01, I2 = 80.45) or continuation (Table XII, Figure 8, F; P-h < 0.001, I2 = 69.67). A set of sensitivity analyses were conducted for breastfeeding continuation that limited the follow-up time to less than 4 months to reduce potential biases related to the introduction of foods other than breastmilk or formula that may begin after 4 months. The summary RRs for breastfeeding continuation in studies limited to 4 months or less of follow-up are summarized in Table XII. The patterns were similar as in the analysis of all time periods for continuation as reported by the individual studies. Discussion In this systematic review and meta-analysis, 6 factors were examined in association with breastfeeding initiation and continuation: maternal smoking, vaginal delivery, multiparity, dyad separation and connection, maternal education level, and breastfeeding education/support. Smoking was one of the strongest and most consistent factors associated with early breastfeeding. Approximately 11% of women in the US smoke during pregnancy and the numbers are even higher in Europe,25,26 indicating an efficient potential target for increased breastfeeding interventions. In addition to women who smoke throughout pregnancy, 50%-80% of women who quit during pregnancy will relapse to smoking within the first 6 months after birth,27,28 and smoking among breastfeeding women is associated with both shorter duration and reduced milk production.29-34 Smoking may be serving, at least in part, as a surrogate measure for SES (and the associated challenges related to breastfeeding), but nonetheless, its strong association indicates that it may be a useful characteristic in identifying women who would benefit from additional support in establishing and maintaining breastfeeding. A large volume of literature was available to examine delivery mode in relation to breastfeeding. Despite some heterogeneity in the categorization of delivery types, vaginal delivery was consistently associated with significant increases in both breastfeeding initiation and continuation. Maternal body size is an important confounder between delivery mode and breastfeeding, but out of 34 studies of breastfeeding initiation, only 7 controlled for maternal body mass index, and in 19 studies of breastfeeding continuation, only 3 controlled for maternal body mass index. Thus, a portion of the effect size of the relationship between delivery mode and breastfeeding outcomes is likely attributable to underlying confounding by maternal body size. Cesarean deliveries might also lead to less breastfeeding initiation because of a disruption of the infant/ mother dyad. Lactogenesis may be disrupted in women who have cesarean deliveries as a result of decreased oxytocin secretion or maternal stress, which may result in decreased milk December 2018 ORIGINAL ARTICLES Factors Associated with Breastfeeding Initiation and