shorter time window was used in the analysis. For exposures with multiple categories, the highest value (eg, the highest level of multiple categories of educational attainment) was compared with the lowest. Estimates of relative risk (RR) and CIs from each study were combined in separate meta-analysis models for each combination of the 6 factors and 2 outcomes (initiation and continuation) using Comprehensive Meta-Analysis software (v 3.2.00089; Bio stat, Englewood, New Jersey). Randomeffects meta-analysis models were used to generate summary RR estimates, 95% CIs, and corresponding P values for heterogeneity (P-h). Study weights were equal to the inverse of the variance of each study’s effect estimate.24 Statistical heterogeneity was assessed with Cochran Q, which tests for between-study statistical variation, and I2 , which indicates percentage of variation attributable to between-study Volume 203 • December 2018 191 heterogeneity. Formal tests of publication bias were not conducted. Results Table V shows the number of studies screened for each factor as well as the number that went on to full-text review and abstraction. Most papers were excluded at the abstract level due to lack of relevant exposure or outcome data, inappropriate study design (review or case study), or duplication. Figures 1-6 (available at www.jpeds.com) show flow diagrams for each factor under study and Tables VI-XI (available at www.jpeds.com) list each paper selected for abstraction. For breastfeeding initiation, the time frame was consistent across studies with initiation being evaluated within the birth hospitalization period or up to 1 week after birth. For breastfeeding continuation, the time frame for evaluation varied considerably across studies from 1 month up to 1 year although most studies focused on the first 2-4 months of the postpartum period (Tables VI-XI). For maternal smoking, papers were harmonized according to smoking during pregnancy vs not. Results (Table XII, Figure 7, A, and Figure 8, A [available at www.jpeds.com]) were Table V. Search terms and results for each factor Factors Search terms Articles identified Full text reviewed Selected for Inclusion Added after Identification from other factor reviews Total included papers All searches (((breastfeeding[MeSH Terms]) OR breastfeeding[Title/ Abstract]) AND((duration[Title/Abstract]) OR initiat*[Title/Abstract] OR establish*[Title/Abstract] OR continu*[Title/Abstract])) AND Smoking (“Smoking”[MeSH Terms] OR “Tobacco Products”[MeSH Terms] OR smok*[All fields] OR tobacco[All fields]) 396 131 31 8 39 Mode of delivery (((“Cesarean Section”[MeSH Terms] OR “Extraction, Obstetric”[MeSH Terms] OR “Delivery, Obstetric” OR “Labor Presentation”[MeSH Terms] OR Cesarean[All fields] OR caesarean[All fields] OR “Vaginal Delivery”[All fields] OR “vaginally delivered”[All fields]))))))) AND (“2005/01/05”[PDat] : “2016/02/12”[PDat] AND English[lang]))) 325 128 28 19 47 Parity (Parity[MeSH Terms] OR “Reproductive History”[MeSH Terms] OR parity[All fields] OR “first birth”[All fields] OR nullipar*[All fields] OR primipar*[All fields] OR multipar*[All fields]) 477 177 22 9 31 Dyad separation (“Time to first”[All fields] OR “Separation of dyad” [All fields] OR “Dyad separation”[All fields] OR Skin-to-skin [All fields] OR Rooming-in[All fields] OR Rooming[All fields]) 146 56 9 8 17 Maternal education ((Education[MeSH Terms] OR Schooling[MeSH Terms] OR education[All fields] OR college[All fields] OR high school[All Fields] OR educ*[All fields])) 1859 508 54 8 62 Breastfeeding education Health Knowledge, Attitudes, Practice[MeSH Terms] OR Patient Education as Topic[MeSH Terms] OR Health Education[MeSH Terms] OR “breastfeeding information”[All fields] OR “breastfeeding education”[All fields] OR “lactation counseling”[All fields] OR “lactation consultant”[All Fields] OR “prenatal class”[All fields] OR “childbirth class”[All fields] 702 223 20 12 32 Table XII. Summary RRs and 95% CIs for 6 factors in relation to breastfeeding) initiation and continuation, overall, and limited to studies with follow-up of 4 months or less (continuation only) Breastfeeding initiation Breastfeeding continuation Breastfeeding continuation with follow-up <4 mo Factors Included studies* (RR, 95% CI) Included studies* (RR, 95% CI) Included studies* (RR, 95% CI) Smoking: nonsmoking vs smoking 1Parity: multiparous vs primiparous 20 1.04 (ies identified in the systematic review (Table XII) may not have been included in the meta-analysis if key information such as variance or CIs were unavailable. THE JOURNAL OF PEDIATRICS • www.jpeds.com Volume 203 192 Cohen et al relatively consistent in showing increased breastfeeding initiation (P-h >< 0.01, I2 = 67.37)) and continuation (P-h < 0.001, I 2 = 80.16) for nonsmokers compared with smokers. One study included paternal smoking in relation to continuation; when removed in a sensitivity analysis, results were similar (summary RR 1.96; 95% CI 1.72-2.23; P-h < 0.001, I2 = 79.48). Papers examining delivery mode most often compared vaginal to cesarean delivery, but there was heterogeneity in more subtle groupings within vaginal (spontaneous vs induced, or forceps or vacuum assistance vs no assistance) and cesarean delivery (elective/not in labor vs nonselective/in labor). Whenever available, we used RRs with spontaneous vaginal delivery as the independent variable in the meta-analyses. Summary RRs for breastfeeding initiation (P-h < .01, I2 = 93.11) and continuation (P-h < 0.01, I2 = 55.46) were higher for vaginal compared with cesarean delivery (Table XII, Figure 7, B, and Figure 8, B). For papers examining parity in relation to breastfeeding initiation, there was a great deal of heterogeneity in results between the individual studies, and the summary RR for multiparous woman vs primiparous woman was nonsignificant (Table XII,