Breastfeeding and maternal health outcomes: a systematic review and meta-analysis

Aim To evaluate the effect of breastfeeding on long-term (breast carcinoma, ovarian carcinoma, osteoporosis and type 2 diabetes mellitus) and short-term (lactational amenorrhoea, postpartum depression, postpartum weight change) maternal health outcomes. Methods A systematic literature search was conducted in PubMed, Cochrane Library and CABI databases. Outcome estimates of odds ratios or relative risks or standardised mean differences were pooled. In cases of heterogeneity, subgroup analysis and meta-regression were explored. Results Breastfeeding >12 months was associated with reduced risk of breast and ovarian carcinoma by 26% and 37%, respectively. No conclusive evidence of an association between breastfeeding and bone mineral density was found. Breastfeeding was associated with 32% lower risk of type 2 diabetes. Exclusive breastfeeding and predominant breastfeeding were associated with longer duration of amenorrhoea. Shorter duration of breastfeeding was associated with higher risk of postpartum depression. Evidence suggesting an association of breastfeeding with postpartum weight change was lacking. Conclusion This review supports the hypothesis that breastfeeding is protective against breast and ovarian carcinoma, and exclusive breastfeeding and predominant breastfeeding increase the duration of lactational amenorrhoea. There is evidence that breastfeeding reduces the risk of type 2 diabetes. However, an association between breastfeeding and bone mineral density or maternal depression or postpartum weight change was not evident.


INTRODUCTION
Breast milk is the natural first food for newborns. It provides all the energy and nutrients that an infant needs for the first six months of life, up to half or more during the second half of infancy and up to one-third during the second year of life (1,2). For mothers, breastfeeding has been reported to confer lower risk of breast and ovarian carcinoma (3,4), greater postpartum weight loss (5) and decreased blood pressure (6) compared with no breastfeeding. The World Health Organization (WHO) recommends exclusive breastfeeding in the first six months and continuation of breastfeeding for 2 years and beyond (1).
The association between breastfeeding and breast carcinoma in mothers has received increased scrutiny in recent years. A number of studies have suggested that breastfeeding, particularly for an extended period of time, may be associated with a decreased risk of breast carcinoma, even after adjustment for potential confounders (7). It is difficult, however, to estimate the magnitude of association between breastfeeding duration and breast carcinoma if any, because of the different methodologies used in breastfeeding his-tories. Parity is also a protective factor against breast carcinoma (8), and there may be an interaction between parity and breastfeeding duration interplay in protecting women from breast carcinoma.
Ovarian cancer is one of the most common cancers in female (9,10). Reproductive factors have been identified as markers of risk for ovarian cancer. These reproductive factors mainly include total number of pregnancies, parity, age at menarche and menopause, as well as breastfeeding (11). Evidence from previous analyses indicates an inverse association between breastfeeding and the risk of ovarian carcinoma (4,12).
Calcium metabolism and bone metabolism are substantially altered with increased calcium demands during pregnancy and lactation. Bone densities can decrease by between 3 and 10 per cent in the span of a few months in a healthy mother (13). Confounders commonly considered in the studies of the relationship between fracture risk and breastfeeding are age, hormone replacement therapy, parity and BMI (4).
Available literature suggests that breastfeeding reduces the risk of maternal type 2 diabetes in some cohort studies, but the evidence from published studies has differed with regard to the strength of the association (14,15).
The literature suggests that exclusive breastfeeding protects against pregnancy (16,17). Some studies, however, show that exclusive breastfeeding is not always associated with inhibition of ovulation (18,19).
The incidence of postpartum depression (PPD) is high (10-15%) (20), and depression during pregnancy usually continues into the postpartum period (21). Postpartum depression has an immediate impact on mothers. It carries long-term risks for their mental health (22) and may also have significant negative effects on the cognitive, social and physical development of their children (23). The evidence for an association between breastfeeding and PPD is, however, unclear (23,24).
Postpartum weight retention is a predictor for future overweight and obesity (25) and is associated with obesityrelated illnesses, such as type 2 diabetes mellitus and cardiovascular disease (26). Breastfeeding may promote weight loss due to lactation (27), but there is a lack of strong evidence to support this hypothesis (28).
We conducted this review to summarise the literature and explore the relationship of breastfeeding and its duration with long-term (breast carcinoma, ovarian carcinoma, osteoporosis and type 2 diabetes mellitus) and shortterm (lactational amenorrhoea, postpartum depression, postpartum weight change) maternal health outcomes. Outcomes for review were selected during an expert meeting at the World Health Organization (October 2014) that was reviewing the impact of breastfeeding on maternal and child health.

METHODS
A search strategy (Box 1) was developed and reviewed by all authors. Medical Subject Heading (29) terms and keywords were used in various combinations. We searched published literature from PubMed, Cochrane Library and CABI databases to identify studies examining the effect of type and duration of breastfeeding on maternal health outcomes. We conducted the search in February 2015. No language or date restrictions were employed in the electronic search.
Two review authors (RC and BS) screened the titles and abstracts independently to identify potentially relevant citations. These review authors retrieved the full texts of all potentially relevant articles and independently assessed the eligibility of the studies using predefined inclusion criteria. We extracted data from all articles found to be relevant by both authors. Any disagreements or discrepancies between reviewers were resolved by discussion and if necessary by consulting a third author (JSM). In addition to the electronic search, we searched reference lists of the articles identified. We used Web-based citation index for citing manuscripts of these identified articles.
We identified four recent systematic reviews addressing the following outcomes: ovarian carcinoma (30), type 2 diabetes mellitus (31), postpartum depression (32) and postpartum weight change (33). We planned to update these reviews and provide new quantitative estimates of breastfeeding on these health outcomes. For other maternal health outcomes, that is breast carcinoma, osteoporosis and lactational amenorrhoea, we planned for new reviews.

Inclusion criteria
We selected all observational studies (prospective/retrospective cohort and case-control), randomised controlled trials (RCTs), including cluster randomised trials, and quasiexperimental trials which examined the impact of duration and type of breastfeeding on maternal health outcomes. For articles not written in English, we attempted to get an English abstract. If it was not available, the article was excluded.
Abstraction, summary measure, breastfeeding categories and analysis We abstracted data using a modified Cochrane data abstraction form. If a study provided separate estimates for hospital-and community-based populations, then the outcome estimates were pooled separately. We used odds ratios (ORs), both adjusted and unadjusted, as our outcome estimate for breast and ovarian carcinoma. Relative risk (RR) was used as the outcome estimate for lactational amenorrhoea. To examine the effect on breast and ovarian carcinoma, breastfeeding was categorised into ever breastfed vs. never breastfed and also by breastfeeding duration, that is breastfed less than six months vs. not breastfed; breastfed 6 to 12 months vs. not breastfed; and breastfed >12 months vs. not breastfed. For lactational amenorrhoea, we used exclusive, predominant, partial, any and no breastfeeding as the categories (Table A1). Standardised mean differences in bone mineral density between highest and lowest breastfeeding duration categories were used for osteoporosis outcome. A narrative approach was used to summarise the studies for postpartum weight change as the studies were very heterogeneous.
We performed meta-analysis with Stata 11.2 software (StataCorp, College Station, TX, USA). We calculated the pooled estimates of the outcome measures from the odds ratios (ORs), relative risks (RRs), standardised mean differences (SMDs) and 95% confidence intervals (CIs) of the individual studies by inverse variance or DerSimonian and Laird method in Stata (34). High heterogeneity was defined by either a low p-value (<0.10) or I 2 value greater than 60%. In cases of high heterogeneity, the random-effects model was used and causes were explored by conducting subgroup analysis and meta-regression. Subgroup analyses were carried out based on breastfeeding categories (ever vs never, less than six months vs never, 6-12 months vs never, >12 months vs never). Among the ever vs never breastfeeding category, subgroup analyses were carried out based on sample size (<500, 500-1499, ≥1500), individual study setting (i.e. highincome country (HIC) or low-and middle-income country (LMIC) (35)), study design (cohort, case-control), mean age of diagnosis (≤49 years, >49 years), adjustment for parity (fine adjustment, i.e. adjustment according to each parity number measured as 0, 1, 2, 3, 4+; crude adjustment, i.e. groupwise adjustment measured as 0, 1-3, 4+ children; and no adjustment), control for confounding (thorough, i.e. controlled for all potential socio-demographic and reproductive factors such as age, income, ethnicity, parity, contraceptive use, family history of carcinoma, menopausal status and smoking; partial, i.e. only partially controlled for potential socio-demographic and reproductive factors; and none) and quality of study (adequate, i.e. study had none or one among selection bias, measurement bias, attrition (20%) and confounding bias; inadequate) (36). We also evaluated the presence of publication bias in the extracted data for the primary outcome using Begg's test or Egger's test or funnel plots (37).

RESULTS
We screened the 12 071 titles identified. Of these, after reviewing abstracts of 1501 articles, we selected 341 for fulltext review. We identified 163 articles for inclusion in our final database (Fig. 1). Among these, 100 studies examined the impact of breastfeeding on breast carcinoma, 40 studies on ovarian carcinoma, 12 studies on lactational amenorrhoea, five studies on postpartum weight change and six studies on osteoporosis. We did fresh meta-analysis for breast carcinoma, ovarian carcinoma, osteoporosis and lactational amenorrhoea and updated the review on postpartum weight change. No new studies subsequent to the existing reviews on type 2 diabetes mellitus and postpartum depression (31,32) were found to be eligible for inclusion.
Effects of breastfeeding on long-term maternal health outcomes Breast carcinoma We identified 98 estimates  of the association between ever breastfeeding and breast carcinoma risk (Tables 1 and A2). Ever breastfeeding was associated with 22% (OR 0.78, 95% CI 0.74-0.82) (Fig. 2) reduction of breast carcinoma risk compared with never breastfeeding. Compared with no breastfeeding, breastfeeding for less than six months (39 estimates) and breastfeeding for 6-12 months (36 estimates) were associated with 7% (OR 0.93, 95% CI 0.88-0.99) and 9% (OR 0.91, 95% CI 0.87-0.96) risk reduction of breast carcinoma, respectively. We found that mothers who breastfed for >12 months compared with those who did not breastfeed had a 26% lower risk of developing breast carcinoma (50 studies; OR 0.74, 95% CI 0.69-0.79), and when restricted to high-quality studies, only (41 studies) breastfeeding >12 months was associated with 23% lower risk of developing breast carcinoma (OR 0.77, 95% CI 0.72-0.83) (not shown in Table 1). There was, however, an indication of publication bias. Asymmetry was observed in funnel plot when inspected visually. Both Egger's test (p bias <0.001) and Begg's test (p bias <0.001) showed statistically significant findings.
Subgroup analysis of the effects of ever breastfeeding on risk of breast carcinoma among studies conducted in highincome countries, with large sample sizes (>1500), of cohort design, with thorough control of confounding factors and adequate quality showed a smaller breast carcinoma risk reduction. Studies where fine adjustment for parity was made showed a smaller effect of breastfeeding on breast carcinoma risk reduction (OR 0.92, 95% CI 0.88-0.96) compared with studies where crude adjustment or no adjustment was made. A restricted analysis including parous women in the fine adjustment subgroup showed a risk reduction of 7% for breast carcinoma (OR 0.93, 95% CI 0.89-0.97; 14 estimates) (not shown in Table 1).

Ovarian carcinoma
Pooled results from 41 estimates (65,69, showed that mothers who ever breastfed their children had a 30% reduction in the risk of ovarian carcinoma, when compared with those who never breastfed (OR 0.70, 95% CI 0.64-0.77) (Tables 2 and A3; Fig. 3). The risk of ovarian carcinoma was 17% lower among women who had breastfed for less than six months when compared with those who did not breastfeed (OR 0.83, 95% CI 0.78-0.89). The risk of ovarian carcinoma among mothers who breastfed for 6-12 months was 28% lower (OR 0.72, 95% CI 0.66-0.78; 19 In subgroup analysis, studies with sample sizes of more than 1500 showed a significant protection of 24% from     Table 2).

Osteoporosis
A total of six studies (174)(175)(176)(177)(178)(179) were identified (Table 3). Two studies were from LMICs (174,178) and four studies from HICs (175)(176)(177)179). Bone mineral density (BMD) was generally measured at two sites, that is femoral neck and distal radius. For femoral neck, four studies (175,(177)(178)(179) were identified with small sample size (total 489 women  Effects of breastfeeding on short-term maternal health outcomes Lactational amenorrhoea We identified 12 studies (173,(180)(181)(182)(183)(184)(185)(186)(187)(188)(189)(190) that examined the association between breastfeeding and lactational amenorrhoea (Table 4). Four studies (180,182,185,188) did not provide either RR or OR. They reported that exclusive compared to mixed feeding, or longer duration of any breastfeeding, was associated with an increased mean or median duration of lactational amenorrhoea. The remaining studies provided data from which the following estimates were derived: the probability of continued   (Table 5) (191)(192)(193)(194)(195). In the review by Neville et al., the majority of identified studies reported little or no association between breastfeeding and weight change. Of those five studies, three studies were performed in low-and middle-income countries, one was performed in high-income country, and one was multicentre study (Brazil, Ghana, India, Norway, Oman, USA). In studies performed in low-and middle-income countries, we have not found any potential differential effect for breastfeeding and postpartum weight loss response as a function of countries being low to middle and high income. Two of the five additionally identified studies (194,195) reported a significant reduction in postpartum weight with breastfeeding. Sarkar and Taylor (191) in a cross-sectional study in Bangladesh revealed that body weight of mothers was negatively correlated with 1-12 and 13-24 months of lactation after controlling for height, education and food consumption. Stuebe et al. (192) showed that women who exclusively breastfed for greater than six months had the lowest BMI at 3 years postpartum as well as the lowest postpartum weight retention at 3 years compared with women who never exclusively breastfed. A multicentre study showed that lactation intensity and duration explained little variation in weight change patterns (193)(194)(195). Overall, the role of breastfeeding on postpartum weight change remains unclear.

DISCUSSION
The aim of this review was to systematically examine the effect of breastfeeding on important maternal health outcomes. The risk of developing breast carcinoma was reduced by 26% among women who cumulatively breastfed for more than 12 months, compared with women who did not breast feed.
Previous reviews suggested that breastfeeding was not strongly related to risk of breast carcinoma (196,197) or found a small but statistically significant protective association (198)(199)(200). Our meta-analysis findings are comparable with but suggest a higher level of protection than that found by the Collaborative Group on Hormonal Factors in Breast Carcinoma (201). In this pooled analysis of approximately 50 000 carcinoma cases from 47 studies in 30 countries around the world and after adjustment for confounders including parity and exclusion of nulliparous women, the authors estimated that the risk of invasive breast carcinoma decreased by 4.3% for every 12 months of breastfeeding (201). However, one of the challenges of comparing studies on cumulative breastfeeding duration and determining the effect on breast carcinoma risk is the lack of a standard protocol for grouping the lifetime number of months of breastfeeding for analysis and the adjustment of parity. Lifetime duration of breastfeeding is related to the number of children breastfed, that is parity and the duration of breastfeeding for each child. Our  results showed that when controlled for parity, breastfeeding independently contributed to a modest but significant risk reduction for breast carcinoma. The risk reduction for breast carcinoma was 8% among ever breastfed mothers when finely adjusted for parity, while it was 22% when all studies were pooled together. Even when our analysis was restricted to only parous women, finely adjusted for parity, ever breastfeeding was associated with a 7% reduction in risk of breast carcinoma compared with never breastfeeding. Longer duration of breastfeeding (>12 months) was associated with more protection of breast carcinoma than shorter duration of breastfeeding (breastfeeding <6 and 6-12 months) when compared to never breastfeeding. Even when our analysis was restricted to studies with adequate quality, breastfeeding >12 months showed more protection against breast carcinoma. Possible biological mechanisms include that protection may occur through parity-specific changes in levels of circulating hormones such as estradiol, prolactin and growth hormone, as each of these has been associated with breast cancer risk (202), or that the parous mammary gland may contain epithelial cells with a more differentiated and less proliferative character which are less susceptible to transformation (203). Breastfeeding by women for more than 12 months was also associated with a 35% reduction in ovarian cancer, compared with women who had not breastfed. The protective effect was less in women who had only ever breastfed (for any duration) ranging from 30% in an unadjusted analysis to 18% when the analysis was restricted to ever breastfeeding parous women (finely adjusted for parity). A number of physiological mechanisms may account for the protective effect of breastfeeding against ovarian cancer through modulating ovarian cycle length (204), and therefore, parity is an important confounder. Longer duration of breastfeeding suppresses ovulation longer and causes suppression of gonadotropins, resulting in depressed production of plasma estradiol, considered to be a potential causal mechanism of ovarian cancer when present at high levels (205). However, breastfeeding must also have an independent effect to explain the estimated reduction in ovarian cancer when parity is adjusted for. There did not appear to be a significant effect of breastfeeding on the risk of osteoporosis. Calcium metabolism and bone metabolism are substantially altered during pregnancy and lactation, and high calcium demand during lactation makes women more prone to bone resorption and subsequent osteoporosis. There was no evidence of such risk, and it has been suggested that during lactation, oestrogen imposes minor inhibitory effect on periosteal bone formation and permits periosteal expansion which increases bone size after weaning (206).
Available review suggests that longer duration of breastfeeding reduces risk of development of type 2 diabetes mellitus by 32%, and in linear dose-response analyses, there was a 9% reduction in relative risk for each 12-month increase in lifetime duration of breastfeeding. Our review shows that exclusive or predominant breastfeeding during the first six months postpartum was associated with longer periods of amenorrhoea. Less intensive breastfeeding, captured under 'any or partial breastfeeding', offers less clear benefit. This finding is biologically plausible. Breastfeeding suppresses the resumption of ovarian activity after childbirth and is thus associated with a period of infertility. Exclusive breastfeeding and predominant breastfeeding are associated with a higher frequency of suckling than other patterns of breastfeeding. Frequent suckling inhibits gonadotropin-releasing hormone and decreases the release of luteinising hormone and follicle-stimulating hormone (207), thus preventing early return of menses.
The association between breastfeeding and postpartum weight change remains uncertain. Factors such as age, gestational weight gain and prepregnancy weight confound such analyses (208,209). As prepregnancy weight and gestational weight gain were found to be strong determinant factors of postpartum weight change, future research should include the preconception period with continued monitoring into the postpartum period to capture the true trajectory of weight change. Even though BF may not lead to postpartum weight loss under 'natural' conditions, it remains unknown whether women who wish to lose weight intentionally in the postpartum period are more likely to be successful at doing so if they are vs. if they are not breastfeeding.
Although our original review plans included exploring the associations between breastfeeding and the risk of maternal postpartum depression and type 2 diabetes, we were unable to identify new studies following the reviews published in 2015 (31) and 2013 (32). The evidence suggests that the relationship between breastfeeding and postpartum depression is lacking.
The range of the maternal outcomes examined and the various categories of breastfeeding exposures that we considered are important strengths of this review. Despite the expanded scope of review, other important maternal health outcomes such as maternal hypertension and cardiovascular disease were not addressed and should be considered in future research and reviews. Also important was the attempt to look for dose-response relationships and the evaluation of heterogeneity and publication biases. However, some limitations should be acknowledged. We have pooled data from many observational studies that are prone to be affected by biases such as in recall or due to selection. Some studies did not control for or collect information on potential confounders that could have affected the association between breastfeeding and the outcome of interest. For postpartum weight change, we were constrained to take a narrative approach to present the outcomes because of the heterogeneous nature of the studies. In cases of significant heterogeneity in study results, we have performed post hoc subgroup analysis and meta-regression and have used the random-effects model. But in some cases even within subgroups, there was significant heterogeneity which suggests some other unidentified factors causing such heterogeneity. Although the meta-regression seemed to explain around 80% of the heterogeneity for breast and ovarian carcinoma, we need to acknowledge the limitation of post hoc subgroup analysis.

CONCLUSION
Our meta-analysis shows that women who had ever breastfed and who breastfed for longer duration have a lower risk of breast and ovarian carcinoma and also type 2 diabetes mellitus. Exclusive or predominant breastfeeding during the first six months postpartum prolongs lactational amenorrhoea. We found no evidence of a clear association between breastfeeding and bone mineral density, maternal depression or postpartum weight change. port extended by Ms. Nicola Mendes during manuscript preparation.

CONFLICT OF INTEREST
The authors have no financial relationships or other conflict of interests to disclose.

DISCLAIMER
The authors alone are responsible for the views expressed in this article and they do not necessarily represent the views, decisions or policies of the institutions with which they are affiliated. Exclusive breastfeeding Breast milk from mother or wet nurse or expressed breast milk No other liquids or solids except vitamin drops or syrups, mineral supplements, or prescribed medicines Predominant breastfeeding Breast milk from mother or wet nurse or expressed breast milk Water and water-based drinks No food-based fluid with the exception of fruit juice and sugar water Vitamin drops or syrups, mineral supplements, or prescribed medicines Partial breastfeeding Breast milk from mother or wet nurse or expressed breast milk Any other liquids or nonliquids, including both milk and nonmilk products

APPENDIX
Any breastfeeding Breast milk from mother or wet nurse or expressed breast milk Includes children exclusively, predominantly, fully and partially breastfed No breastfeeding Formula and/or animal's milk No breast milk