Maternal Body Mass Index and Breastfeeding Non-Initiation and Cessation: A Quantitative Review of the Literature

This study aims to investigate which maternal body mass index (BMI) categories are associated with the non-initiation or cessation of breastfeeding (BF) based on a quantitative review of the literature. We searched Ovid MEDLINE and EBSCO CINAHL for peer-reviewed articles published between 1946 (MEDLINE) or 1981 (CINAHL), and 2019. Selected studies were either cross-sectional or cohort studies, of healthy mothers and infants, that reported nutrition method (exclusive/full or any) and period (initiation/duration/cessation) of breastfeeding according to maternal BMI levels. Pairwise meta-analyses of 57 studies demonstrated that the pooled odds risks (OR) of not initiating BF among overweight and obese mothers compared to normal weight mothers were significant across 29 (OR 1.33, 95% confidence interval (CI), 1.15–1.54, I2 = 98%) and 26 studies (OR 1.61, 95% CI, 1.33–1.95, I2 = 99%), respectively; the pooled risks for BF cessation were inconsistent in overweight and obese mothers with substantial heterogeneity. However, we found that overweight mothers (n = 10, hazard ratio (HR) 1.16, 95% CI, 1.07–1.25; I2 = 23%) and obese mothers (n = 7, HR 1.45, 95% CI: 1.27–1.65; I2 = 44%) were both associated with an increased risk of not continuing any BF and exclusive BF, respectively. Overweight and obese mothers may be at increased risk of not initiating or the cessation of breastfeeding.


Introduction
Breastfeeding has overwhelming positive evidence for both infants and mothers. Breast milk consists of bioactive factors that play a role in immunological strengthening [1], carcinogenic development [2], neural and psychological benefits [3], and a possible protective factor for obesity in childhood [4]. In addition, recent literature demonstrates that children who are exclusively breastfed have lower infectious morbidity and mortality than those who are not breastfed, or partially breastfed [5]. Breastfeeding has a positive effect on women's health in that it can prevent breast cancer [6], diabetes [7], and ovarian cancer [8]. According to a particular article published in the Lancet in 2016, the scaling up of breastfeeding can prevent an estimated 823,000 child deaths and 20,000 breast cancer deaths every year [9].

Outcome of Breastfeeding
The outcome of interest in our meta-analyses was any or exclusive breastfeeding (ABF and EBF, respectively) in relation to non-initiation or cessation. According to the definition proposed by the WHO [31], children were considered to be exclusively breastfed if they were given no other food, drink, added nonhuman milk or infant formula, except breast milk. ABF was considered if the child received liquids like water-based drinks, fruit juice, and ritual fluids, other than breast milk [31], or if the study did not clearly describe exclusive or any. We collected either ORs or HRs of BF non-initiation or cessation and, if the study provided risks of BF initiation or continuation, we recalculated the risk of BF non-initiation or cessation by taking the inverse of the risk of the BF status to align the direction of the risk. To estimate the pooled effect size of the risk measures, we used risk ratios in multivariable analyses rather than unadjusted models.

Quality Assessment
We rated the methodological quality of the studies according to a modified version of the criteria provided by the Newcastle-Ottawa scale for cohort studies [32] because our included studies were prospective, retrospective cohort, or cross-sectional studies. The Newcastle-Ottawa Scale uses a star system (with a maximum of 9 stars) to evaluate a study in 3 domains: selection of participants (4 items; Representativeness of the Exposed Cohort, Selection of the Non-Exposed Cohort, Ascertainment of Exposure, Pre-Pregnancy BMI Measurement), comparability of study groups (2 items; Control for Confounders), and the ascertainment of outcomes of interest (3 items; Assessment of Breastfeeding, Was Follow-Up Long Enough for Outcomes to Occur?, Adequacy of Follow-Up of Cohorts).
Representativeness of the Exposed Cohort earns one star if the representativeness of exposed individuals in the particular community is described or if the sample size is larger than 1000. Selection of the Non-Exposed Cohort earns one star if only the normal weight group was defined as a reference group. Ascertainment of Exposure earns one star if maternal weight and height were measured and not self-reported to calculate maternal BMI. Pre-Pregnancy BMI Measurement earns one star if the study clearly stated a BMI measurement in the earlier timing of pregnancy. This indicates that the study is prospective and breastfeeding was not present at the start of the study. A maximum of 2 points can be given for comparability. If a study controls at least 2 variables, including maternal age, gestational week and delivery mode to calculate a risk of BF non-initiation or cessation, it will receive one point. If the study also controls for other confounding factors (i.e., socio-economic factors, psychological factors, and social support), it will receive another point.
The assessment of breastfeeding earns one star if initiation or continuation of the particular type of breastfeeding (i.e., exclusive, full, dominant, or any breastfeeding) is clearly defined. "Was Follow-Up Long Enough for Outcomes to Occur?" earns one star if the study calculated the risk estimates at a minimum of one month for initiation and 6 months for continuation, because the WHO recommends 6 months of breastfeeding. Adequacy of Follow-Up of Cohorts earns one star if the study was a cohort study and 70% of the initially enrolled sample was followed up, and N/A if the study was a cross-sectional study.
With a maximum of 9 stars, we judged studies that received a score of 9 stars to be at low risk of bias, studies that received 7 (for cross-sectional studies) or 8 stars (for cohort studies) to be at medium risk, and those that received 6 (for cross-sectional studies) or 7 stars (for cohort studies) (or fewer) to be at high risk of bias.

Data Analysis
Pairwise meta-analyses were performed to compare the estimate of BF non-initiation, and ABF or EBF cessation between each pre-pregnancy BMI category (i.e., underweight, normal weight as reference, overweight, and obesity). Pooled ORs and HRs and 95% confidence intervals (CIs) of being underweight, overweight, or obese on an effect of (i.e., above 1) breastfeeding non-initiation/cessation and were estimated using random-effects models with the Sidik-Jonkman estimator [33]. Subgroup analyses were conducted according to BF (ABF) or EBF, the timing of the risk assessment, ethnicity, and quality score (≤5 vs. ≥6). For a study reporting only subgroup results, we first estimated the combined ORs or HRs within the study by using random-effects models, and then combined these with other findings from the selected study [34]. Meta-regression analyses were also performed to examine the impact of categorical variables on study effect size using regression-based techniques. Heterogeneity across individual studies by Cochran's Q (a measure of weighted squared deviations), the I 2 statistics (the ratio of true heterogeneity to total observed variation), and Tau 2 (the variance of the true effect sizes) were estimated. Tau 2 can be viewed as a point estimate of the among-study variance of true effects, while I 2 would be a measure of inconsistency and the proportion of variability in the point estimates [35]. For Cochran's Q test, the P value indicating significant heterogeneity was set at less than 0.10 [36]. For I 2 , values less than 25%, from 25% to 50%, and more than 50% indicated modest, moderate and substantial heterogeneity, respectively. To assess publication bias, we constructed funnel plots and tested the asymmetry using Egger's and Begg's test. If Begg's test was significant, we further constructed a trim and fill funnel plot by imputing missing studies to adjust for publication bias.
All the analyses were conducted in the statistical software package STATA version 16. p < 0.05 was considered statistically significant, unless otherwise specified.

Discussion
This study investigated which maternal BMI categories are associated with BF non-initiation or cessation using a quantitative review. The effects of being underweight on BF non-initiation and

Discussion
This study investigated which maternal BMI categories are associated with BF non-initiation or cessation using a quantitative review. The effects of being underweight on BF non-initiation and cessation were all insignificant irrespective of the types of BF practice (i.e., ABF or EBF), or risk measurement (i.e., HR or OR). The pooled risks of non-initiation among those who were overweight and obese were both significant, although there was substantial heterogeneity observed. The pooled risk for ABF and EBF cessation in overweight and obese mothers was inconsistent with substantial heterogeneity. However, the hazard risks of ABF cessation in overweight mothers and EBF cessation in obese mothers were both significant with less heterogeneity (i.e., I 2 = 23% and 44%, respectively). This quantitative review suggests that overweight and obese mothers may have an associated increased risk of BF non-initiation or cessation.
There are two previous meta-analysis studies [14,120] that investigated an association between maternal weight status and BF non-initiation and cessation which agree that obesity may be associated with an increased risk of BF non-initiation or cessation compared to normal weight mothers. However, their study hypotheses differ from ours; one study [120] not only included maternal weight status, but diet and supplement use as determinants of breastfeeding and complementary feeding; the other [14] investigated BF cessation associated with maternal BMI and gestational weight gain. The study conducted by Huang et al. [14] is similar to ours, but does not include cross-sectional studies, and thus includes only 30 studies. The exclusion of these studies may overlook the true association of interest because most studies that focused on breastfeeding initiation are cross-sectional studies. Hence, the advantages of our study are that, first, the numbers of studies included in our meta-analyses are much larger in our study than in Huang Table S2. The description of studies excluded from the meta-analyses is important to reproduce a systematic review for a third person. Second, by performing pooled analyses according to odds or hazard ratios, the heterogeneity of some meta-analyses is less observed, which makes it clearer to present the true association of interest. The discrepancy is that Huang et al. reported underweight mothers to be less likely to initiate BF compared to normal weight mothers (n = 14, RR 1.28, 95% CI:1.11-1.48). However, we did not observe such a negative impact among underweight women; both studies observed substantial heterogeneity across the studies included in the meta-analyses. In this regard, the effect of the mother being underweight on breastfeeding still requires further evidence.
There are several reasons for the substantial heterogeneity across the studies. First, there were only two studies awarded the maximum number of seven stars for quality, and the other studies were awarded six stars or fewer, which indicates the majority of studies included in these meta-analyses had, at least, a moderate risk of bias. The reason for the lower quality scores across studies includes the nondifferential misclassification or recall bias by self-reporting of maternal weight. In addition, the majority of the reviewed studies failed to provide follow-up rates that were long enough to assess breastfeeding duration. Indeed, few studies in which mothers continued breastfeeding up to six months (as recommended by the WHO and UNICEF) remained for analyses. Second, we excluded 10 studies [21][22][23][24][25][26][27][28][29][30] that provided continuous risk estimates because risk with a one unit increase in BMI is exponential, and thus difficult to interpret by combining the different sources of populations. Indeed, the 10 studies [21][22][23][24][25][26][27][28][29][30] vary across countries, BF types, and the length of the follow-up period. Nine studies [21][22][23][24][25][26][27][28][29] were reported from Western countries and three investigated initiation [23,25,27]. Among the seven studies [21][22][23][24]26,29,30] that assessed the risk of timing at six months after delivery, only two [27,30] reported that the BF significantly decreased by a unit increase of BMI. Furthermore, we also attempted to collect mean BMI according to BF status in order to perform a quantitative assessment, but only two studies [25,29] provided such data. These two studies also reported the risks at different points in time (i.e., 18 months [25] and six months [29]) and thus we did not include these. Third, although we were able to assess the impact of being underweight on breastfeeding practice, because the studies in this area of research focus on obesity, the numbers of underweight studies included in the meta-analyses were few, especially for EBF cessation. Fourth, the WHO encourages EBF, especially in developing countries where maternal biological characteristics and socio-economic factors are significantly different from Western countries. In this regard, although there was no publication bias observed, the number of studies published from developing countries were still few, and thus the generalizability of the present meta-analyses may be limited.
Previously, several potential mechanisms have been reported. First, overweight or obese mothers have large heavy breasts, which may physiologically interfere with infant latching or adversely affect lactation [88]. Rasmussen and Kjolhed [121] demonstrated that excess adiposity in obese women contributes to dysregulation of the hypothalamic-pituitary-gonadal axis. In those women, the prolactin response to the baby's suckle is low with a delayed onset of milk production [122]. Some evidence from research in dairy cows and mice suggests that obesity in early life may negatively influence breast glandular development [123]. Obese mice also show a reduction in the milk proteins β-casein, whey acidic protein, and α-lactalbumin, which are essential for milk production [124]. The breastmilk of normal weight women contains a higher concentration of medium-chain fatty acids; on the contrary, the milk of obese women is characterized by the presence of longer-chain fatty acids. These are more difficult to digest, especially in newborns with immature gastrointestinal systems, thus favoring supplementation with formula and/or solid food [125]. Another study showed that overweight women have a negative physiological pattern [126] influencing "maternal-fetal attachment". Overweight and obese women who are less likely to intend to breastfeed [127], had less confidence in sufficient milk supply and lower body shape satisfaction [128]. They often reported postpartum depression [129], and therefore bottle-feeding is more precocious. However, this association may be confounded by biological and social factors [130]. Finally, breastfeeding practice may be influenced by sociocultural factors. As our Begg's test indicates that publication bias may exist in our subgroup analyses, the majority of the reviewed studies were reported from developed countries that would be a great source of heterogeneity.

Conclusions
In summary, despite substantial heterogeneity across the reviewed studies, our quantitative review suggests that overweight and obese mothers are associated with breastfeeding non-initiation and cessation. Breastfeeding determinants are multifaceted, but among the relevant factors weight control is important to contribute to breastfeeding practice. This should be kept in mind for all relevant health care workers who are responsible for catering to women of reproductive age.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/12/9/2684/s1, Table S1. Search Strategy, Table S2. Characteristics of excluded studies that investigated BMI as a continuous variable, Table S3. Newcastle-Ottawa Scale for the assessment of the methodological quality of included 57 studies.
Author Contributions: K.N. (Kyoko Nomura) and N.K. conceptualized and designed the study, ran individual literature searches, and critically reviewed and collected references for meta-analyses. K.N. (Kyoko Nomura) mainly carried out the analyses which was supervised by K.N. (Kengo Nagashima) for important intellectual content. K.N. (Kyoko Nomura) drafted the manuscript which was critically reviewed by all authors. S.M. and M.O. supervised data collection and 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 authors have read and agreed to the published version of the manuscript.
Funding: All phrases of this study were supported by grant from the Ministry of Sports, Culture, and Science (the Grants-in-Aid for Scientific-Research Program No. 16H052621).