Maternal BMI is positively associated with human milk fat: a systematic review and meta-regression analysis.

BACKGROUND: Lack of robust estimates of human-milk nutrient composition and influential maternal factors, such as body composition, are barriers to informing nutrition policies and programs.
OBJECTIVE: The objective was to understand the relation between maternal BMI and human-milk energy, fat, and/or total protein.
METHODS: Four electronic databases (MEDLINE, Embase, CINAHL, and Web of Science) were searched. Outcomes assessed were human-milk energy (kcal/L), fat (g/L), and total protein (g/L) from mothers 1 to 6 mo postpartum. Studies with data on maternal BMI or weight and height that quantified human-milk energy, fat, or protein between 1 and 6 mo postpartum were eligible. Random-effects meta-regression weighted by the inverse of the study-level SE was completed for each of the 3 outcomes. The certainty of evidence for each outcome was assessed using the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) approach.
RESULTS: A total of 11,373 titles and abstracts were identified, and after full-text screening, 69 articles of 66 studies were included. Meta-regression results showed a positive association between maternal BMI and human-milk fat (β: 0.56 g/L; 95% CI: 0.034, 1.1; P = 0.04; I2 = 93.7%, n = 63 datapoints). There was no significant association between maternal BMI and human-milk energy (β: 3.9 kcal/L; 95% CI: -1.6, 9.5; P = 0.16, I2 = 93.3%, n = 40 datapoints) or total protein (β: 0.13 g/L; 95% CI: -0.16, 0.41; P = 0.37, I2 = 99.1%, n = 40 datapoints). The certainty of evidence for human-milk energy was low and the certainty of evidence for fat and total protein was very low.
CONCLUSIONS: Meta-regression analysis of available literature suggested an association between maternal BMI and human-milk fat between 1 and 6 mo postpartum. Future studies are needed to confirm the relation between maternal BMI; variation in human-milk energy, fat, and protein content; and the implications for child growth and development. This review is registered with International Prospective Register of Systematic Reviews (PROSPERO 2018 CRD42018098808) at https://www.crd.york.ac.uk/prospero/.

See corresponding editorial on page 772.

Introduction

There are strongly supported benefits of breastfeeding, both for mothers and for their infants, and thus exclusive breastfeeding is recommended for the first 6 mo of life (1, 2). For infants, benefits include lower risk of morbidity and mortality as well as improved neurodevelopment (1). Human milk is believed to provide all required nutrients for healthy term infants up to 6 mo of age, with the exception of vitamin D (2, 3). However, the concentration of individual nutrients may vary due to a number of maternal factors including body composition (4, 5).

Bravi et al. (4) performed a systematic review and qualitative assessment of the association between maternal diet and a range of human-milk macronutrients and, from 32 studies (1977–2014), concluded that evidence on this topic is both limited and conflicting. Only studies of well-nourished women were included in their review and associations between maternal body composition and human-milk composition were not assessed. A recent meta-analysis by Leghi et al. (6) found that mothers with overweight or obesity have higher human milk fat compared with normal-weight mothers (n = 6 studies). No significant associations were observed with human-milk protein (n = 5 studies). Human-milk energy was not investigated in this review, and underweight women and low- and middle-income countries were not represented.

To our knowledge, there has never been a contemporary systematic review that has quantitively evaluated all available data in the published literature on human-milk energy and macronutrient composition in relation to maternal body size from underweight to obese using BMI as a continuous variable. We reasoned that meta-regression of datapoints from all available studies would be a novel approach to accomplishing this goal. The lack of robust data on the composition of human milk and factors that influence it has recently been highlighted by several international agencies and researchers (7). Data on milk composition and associated health outcomes are required to facilitate evidence-based nutrition and health programs and policies. Given that a significant number of women of reproductive age globally are underweight and the prevalence of overweight and obesity continues to rise (8, 9), clarification of the degree to which maternal body composition is related to human-milk energy and macronutrient composition is required.

Therefore, the main question that this systematic review aimed to answer was, “In mothers of infants 1 to 6 months of age, is maternal body composition, as assessed by BMI, associated with human-milk energy, fat, and/or total protein?” A secondary aim was to examine the influence of quantification techniques for the determination of human-milk energy, fat, and total protein on these estimates. An additional aim was to explore sources of heterogeneity in human-milk composition related to breastfeeding practices, technique of human-milk collection, prepregnancy or postpartum BMI, and income level of the country where research was conducted.

Methods

A protocol for this systematic review was prospectively registered on the International Prospective Register of Systematic Reviews (PROSPERO 2018 CRD42018098808). The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed in the completion of this systematic review (10).

Literature search and eligibility

The electronic databases MEDLINE, Embase, CINAHL, and Web of Science were searched by a research librarian at the Hospital for Sick Children on 4 July 2019 and the search was updated on 19 May 2020 ( and ). Search terms related to maternal BMI and human-milk energy and macronutrients. There were no restrictions (e.g., language) used for the search apart from the restriction to human research studies.

Studies that included mothers of infants between 1 and 6 mo of age were eligible for inclusion. This age range was selected to ensure that the milk-composition analyses were conducted on mature milk samples and not colostrum or transitional milk. All studies where mothers provided any amount of human milk were considered; in other words, mothers were not always exclusively breastfeeding. Articles including preterm infants born before 37 wk of gestation were to be considered in the review but quantitatively analyzed separately from term-born infants. Studies were excluded if authors did not provide data on maternal BMI or weight and height, did not collect human milk, or did not directly assess (i.e., quantify) human-milk energy, fat, or protein concentration.

Study types that were eligible for inclusion were observational studies and intervention trials; reviews, commentaries, and abstracts of conference proceedings were excluded. Research performed in low-, middle-, and high-income countries was considered. Upon completion of the initial screening process, references of eligible articles were examined to identify additional publications that met the inclusion criteria for this systematic review.

Maternal BMI and main outcomes

Maternal BMI (kg/m2) prepregnancy or during lactation was examined as a continuous variable. BMI is a simple and inexpensive proxy indicator of body composition, with a BMI <18.5 representing underweight, a BMI of at least 25.0 representing overweight, and a BMI of at least 30.0 representing obesity (11). The 3 primary outcomes of interest were human-milk energy (kcal/L), fat (g/L), and total protein (g/L) from mothers of infants between 1 and 6 mo postpartum. Based on reviewer feedback during the peer review process, human-milk lactose (g/L) was also evaluated in post hoc analyses. Total protein reflects the total amount of nitrogen in a sample and includes both protein and nonprotein nitrogen components in human milk. If human-milk samples were collected at multiple time points in a study, measures closest to 3 mo postpartum were used. We considered all studies, regardless of the techniques (e.g., methods) used to quantitate energy, fat, and total protein; the 3 main outcomes were determined a priori for this systematic review (). Unlike with the dairy industry, there is no national or international body that recommends a specific technique for human-milk analysis. However, the following methods are described by experts in the field as the “gold standard” or “preferred” methodology: bomb calorimetry (human-milk energy), gravimetric method (human-milk fat), and nitrogen or amino acid analysis (human-milk total protein) (12–15). As described below, in our secondary analyses, we examined associations between maternal BMI and human-milk energy, fat, and total protein using these preferred methods alone or in combination with other commonly utilized techniques.

Study selection and data extraction

Using Covidence online software, articles were independently screened in duplicate to determine eligibility by 2 of 3 reviewers involved in this systematic review (AID, SS, and SI) (16). This first step involved reviewing titles and abstracts followed by full-text review. The third reviewer was involved in the case of any discrepancies. Reasons for excluding full-text articles were documented. Results from this screening process are summarized in  following the PRISMA guidelines (10).

FIGURE 1

Study flow diagram of included and excluded studies in this systematic review.

A data-extraction sheet in Research Electronic Data Capture (REDCap) was used separately by 2 of the reviewers and data were subsequently combined (17). If any studies had missing data on maternal BMI or human-milk energy, fat, or protein, the authors were contacted for this information, although this was required for only 1 study in the review (18). Data extracted were publication year; country and income classification (low-, middle-, or high-income) according to the World Bank Group (19); study design (experimental or observational); maternal age; maternal prepregnancy and/or postpartum BMI; human-milk collection methods (exclusive breastfeeding, complete breast expression, frequency of collection within a 24-h study period); human-milk energy, fat, and total protein concentration; and the respective quantification techniques used (20).

Data were pooled during the data-extraction process for studies with multiple experimental groups or cohorts. However, for all studies that examined human-milk composition separated by BMI categories, these data were disaggregated and included in the analysis as discrete datapoints. Units were converted to kilocalories per liter for energy and to grams per liter for fat and protein upon data extraction. Protein data that were presented as nitrogen content were multiplied by a protein conversion factor of 6.25 (21). All quantitative data, including ranges, IQRs, variance, SEs, and CIs, were converted to SDs at this step of the review by 2 reviewers independently for each article.

Certainty of the evidence across studies

The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach was used to assess the certainty or quality of the evidence across studies, rather than for individual studies (20). This assessment was done for each of the 3 outcomes examined by meta-regression—specifically, human-milk energy, fat, and total protein, respectively. Evidence started at moderate quality because of the observational nature of this systematic review, and was downgraded based on the following 4 GRADE criteria: 1) risk of bias of individual studies included in the meta-regression (study design or execution problems that could impact the outcome), 2) inconsistency (unexplained statistical heterogeneity in meta-regression results), 3) indirectness (whether included studies assessed mothers across the BMI continuum or human-milk quantification collection methods and techniques employed, many of which were proxy measures such as calculated energy for human-milk energy), and 4) imprecision (wide CIs around the estimate of effect for the meta-regression). An overall score for each outcome (high, moderate, low, or very low) was determined by considering certainty across all GRADE criteria.

Statistical analysis

All analyses were completed in Stata version 14 (StataCorp LP) after data were exported from REDCap version 9.3.8 (Vanderbilt University, Nashville, TN) (22). Random-effects meta-regression using the metareg macro was performed to examine the association between maternal BMI as a continuous predictor of human-milk energy, fat, and total protein, respectively.

Again, individual studies were usually included as single datapoints in the meta-regression. For studies that also disaggregated groups by BMI and reported on human-milk composition across groups, results were included as separate datapoints in the meta-regression. Means were weighted by the inverse of the study-level SE for each of human-milk energy, fat, total protein, and lactose. The meta-regression analyses were fitted using restricted maximum likelihood. The residual variation due to between-study variation (I2 residual) was computed (23). This type of analysis assumes that the relations between variables in the meta-regression are linear or follow a normal distribution. The assumption of normality for each outcome was confirmed using Q–Q plots ().

To address our second aim, additional meta-regression analyses were performed with inclusion of only those studies that used higher-quality human-milk quantification techniques described by experts in the field as “gold standard” or “preferred” methodology for human-milk energy (bomb calorimetry), fat (gravimetric method), and total protein (nitrogen or amino acid analysis). We also completed meta-regression with these techniques in combination with other commonly used techniques to quantitate human-milk fat and total protein. Further subgroup analyses were conducted to explore potential reasons for heterogeneity around human-milk energy, fat, and total protein. These subgroup analyses included exclusive breastfeeding (yes/no), collection of a complete breast expression (yes/no), human milk collected over 24 h (yes/no), BMI measured prepregnancy or postpartum, income level of the country from which milk samples were collected (low- and middle-income vs. high-income), and observational or experimental study design. Separate meta-regression analyses were completed for each respective subgroup comparison and I2 values were assessed.

Results

The search of the 4 electronic databases and reference lists of included studies identified 11,373 abstracts after duplicates were removed (Figure 1). After title and abstract screening, there were 277 articles that were potentially eligible for inclusion. There were 11 titles and abstracts for which the full text could not be located, potentially because they were published in print only and we were unable to secure a hard copy (). Available articles were examined in full and 69 articles reporting on 66 studies were included in the systematic review () (18, 24–91). These articles included a total of 4674 mothers with human-milk samples. Sixty-two studies included term-born infants, while just 6 studies included a total of 216 preterm infants (26, 35, 48–50, 63), which was deemed insufficient for meta-regression based on the Cochrane Handbook (section 9.6.4), which specifies that meta-regression analysis should be considered only when there are a minimum of 10 studies () (92). Therefore, only studies with term-born infants were included in the statistical analyses.

TABLE 1

Characteristics of studies of term-born infants included in this systematic review[1]

First author, year (ref)	Country (classification)	Study design	Sample size,[2]n	Maternal age,[3] y	Maternal BMI,[3,4] kg/m2	Human-milk energy quantification	Human-milk fat quantification	Human-milk protein quantification
Abdul Basir, 2019 (86)	Malaysia (LMIC)	Observational	32	30.6 ± 4.4	23.5 ± 4.8 (prepregnancy)	Calculated energy	Creamatocrit	Colorimetry
Agostoni, 2003 (24)	Italy (HIC)	Observational	50	30.3 ± 5.4	20.7 ± 3.4 (postpartum)	—	Gravimetric method	—
Alvarez de Acosta, 2013 (36)	Venezuela (LMIC)	Observational	52	20.8 ± 5.0	23.9 ± 0.8 (postpartum)	Calculated energy	Gerber method	Nitrogen analysis
Anastácio, 2004 (46)	Brazil (LMIC)	Observational	38	27.7 ± 6.5	23.8 ± 4.3 (postpartum)	—	Creamatocrit	Colorimetry
Antonakou, 2011 (57)/Antonakou, 2013 (68)	Greece (HIC)	Observational	39	32.5 ± 3.1	22.2 ± 4.1 (prepregnancy)	—	Creamatocrit	—
Ayah, 2007 (79)	Kenya (LMIC)	Experimental	422	24.7 ± 6.6	22.2 ± 5.6 (prepregnancy)	—	Creamatocrit	—
Barbosa, 1997 (89)	Mexico (LMIC)	Observational	21 normal-weight, 19 overweight	21.3 ± 4.3 normal-weight, 22.4 ± 3.0 overweight	21.4 ± 0.9 normal-weight (postpartum), 25.7 ± 1.9 overweight (postpartum)	Bomb calorimetry	Jeejeebhoy method	Nitrogen analysis[5]
Be'er, 2020 (88)	Israel (HIC)	Observational	31	33.5 ± 3.3	22.8 ± 3.2 (postpartum)	Calculated energy	IR spectrophotometry	IR spectrophotometry
Boniglia, 2003 (90)	Italy (HIC)	Observational	117	31.2 ± 4.5	24.2 ± 3.3 (postpartum)	—	—	Nitrogen analysis
Brown, 1986 (91)	Bangladesh (LMIC)	Observational	33	19.6 ± 6.3	17.8 ± 1.6 (postpartum)	Calculated energy	Gravimetric method	Nitrogen analysis
Bumrungpert, 2018 (25)	Thailand (LMIC)	Experimental	50	24.9 ± 4.7	22.5 ± 3.1 (postpartum)	Bomb calorimetry	Gravimetric method	Nitrogen analysis
Butte, 1984 (27)	United States (HIC)	Observational	13	26.6 ± 5.0	20.8 ± 2.6 (prepregnancy)	Bomb calorimetry	Gravimetric method	Nitrogen analysis
Butte, 1984 (26)	United States (HIC)	Observational	37	28.0 ± 3.1	22.7 ± 3.9 (postpartum)	Bomb calorimetry	Gravimetric method	Nitrogen analysis
Butts, 2018 (28)	New Zealand (HIC)	Observational	78	30.8 ± 0.9	27.2 ± 0.8 (postpartum)	—	Gravimetric method	Nitrogen analysis
Bzikowska-Jura, 2018 (29)	Poland (HIC)	Observational	22	31.1 ± 4.4	23.3 ± 4.0 (postpartum)	Calculated energy	IR spectrophotometry	IR spectrophotometry
Cannon, 2015 (30)	Australia (HIC)	Observational	19	32 ± 3	25 ± 4 (postpartum)	—	Creamatocrit	Colorimetry
Carratù, 2003 (30)	Italy (HIC)	Observational	195	31 ± 4	23.7 ± 3.2 (postpartum)	—	—	Nitrogen analysis
Chang, 2015 (32)	South Korea (HIC)	Observational	230	32.0 ± 3.3	22.9 ± 3.0 (postpartum)	Calculated energy	IR spectrophotometry	IR spectrophotometry
Cissé, 2002 (33)	Senegal (LMIC)	Experimental	133	27 ± 6	22.6 ± 3.7 (postpartum)	—	Other	Nitrogen analysis
Dagnelie, 1992 (34)	The Netherlands (HIC)	Observational	19	31.5 ± 4.7	21.1 ± 2.0 (postpartum)	Calculated energy	Gravimetric method	Nitrogen analysis
De Luca, 2016 (38)	France (HIC)	Observational	50 normal-weight, 47 obese	30.6 ± 4.5 normal-weight, 30.8 ± 4.9 obese	24.0 ± 3.4 normal-weight (postpartum), 35.3 ± 3.1 obese (postpartum)	Calculated energy	IR spectrophotometry	IR spectrophotometry
de Pee, 1997 (39)	Indonesia (LMIC)	Experimental	64	—	21.0 ± 2.6 (postpartum)	—	Creamatocrit	—
Del Prado, 2001 (37)	Mexico (LMIC)	Experimental	10	21.4 ± 3.1	22.8 ± 2.4 (postpartum)	—	Gravimetric method	—
Dewey, 1994 (41)	United States (HIC)	Experimental	33	30.5 ± 5.1	24.3 ± 3.6 (postpartum)	Calculated energy	Gravimetric method	Nitrogen analysis[5]
Du, 2017 (42)	Australia (HIC)	Observational	40	31.8 ± 4.3	26.3 ± 5.7 (postpartum)	—	Creamatocrit	—
Dusdieker, 1994 (43)	United States (HIC)	Experimental	33	31.7 ± 3.7	25.2 ± 2.9 (postpartum)	—	Creamatocrit	Nitrogen analysis
Fornes, 1995 (44)	Brazil (LMIC)	Observational	14	24.9 ± 5.8	22.8 ± 2.8 (postpartum)	—	Creamatocrit	—
Gridneva, 2018 (45)	Australia (HIC)	Observational	15	33.3 ± 4.7	27.2 ± 5.5 (postpartum)	—	—	Colorimetry
Hampel, 2017 (47)	Bangladesh (LMIC)	Experimental	18	20 ± 3	22 ± 4 (postpartum)	—	Creamatocrit	—
Hsu, 2014 (50)	Taiwan (HIC)	Observational	15	32.7 ± 5.2	24.2 ± 2.4 (postpartum)	Calculated energy	IR spectrophotometry	IR spectrophotometry
Jans, 2018 (51)	Belgium (HIC)	Observational	16 normal-weight, 24 overweight, 9 obese, 6 mixed group	29 ± 6 normal-weight, 29 ± 5 overweight, 29 ± 5 obese, 30 ± 4 mixed group	22 ± 2 normal-weight (prepregnancy), 27 ± 2 overweight (prepregnancy), 35 ± 6 obese (prepregnancy), 28 ± 4 mixed group (prepregnancy)	Calculated energy	IR spectrophotometry	IR spectrophotometry[6]
Kugananthan, 2017 (52)	Australia (HIC)	Observational	23	33.4 ± 4.2	27.0 ± 7.3 (postpartum)	—	—	Colorimetry
Kurniati, 2016 (53)	Indonesia (LMIC)	Observational	48	27.3 ± 4.4	25.9 ± 5.9 (postpartum)	—	Creamatocrit	—
Ley, 2012 (54)	Canada (HIC)	Observational	116	34.8 ± 4.3	24.8 ± 5.1 (prepregnancy)	—	Creamatocrit	Colorimetry
Liyanage, 2008 (55)	Sri Lanka (LMIC)	Observational	66	27.1 ± 6.1	20.8 ± 3.4 (postpartum)	—	Creamatocrit	—
Lovelady, 1990 (56)	United States (HIC)	Observational	16	28.7 ± 4.4	20.9 ± 1.7 (postpartum)	Bomb calorimetry	Gravimetric method	Colorimetry
Mandel, 2005 (58)	Israel (HIC)	Observational	27	30.7 ± 2.9	24.5 ± 3.9 (postpartum)	Bomb calorimetry	Creamatocrit	—
Marangoni, 2000 (59)	Italy (HIC)	Observational	7	33.5 ± 1.8	25.5 ± 2.5 (postpartum)	—	Gravimetric method	—
McCrory, 1999 (60)	United States (HIC)	Experimental	67	31.5 ± 5.3	25.2 ± 4.2 (postpartum)	Calculated energy	Gravimetric method	Nitrogen analysis[5]
Michaelsen, 1994 (61)	Denmark (HIC)	Observational	88	29 ± 5	21.3 ± 3.1 (postpartum)	Calculated energy	IR spectrophotometry	IR spectrophotometry[5]
Minato, 2019 (87)	Japan (HIC)	Observational	56	—	21.3 ± 2.8 (postpartum)	Calculated energy	IR spectrophotometry	IR spectrophotometry
Mohammad, 2011 (62)	United States (HIC)	Experimental	7	26.6 ± 1.3	34.5 ± 1.5 (postpartum)	Calculated energy	Other	Colorimetry
Motil, 1995 (64)	United States (HIC)	Experimental	16	29.0 ± 2.6	23.6 ± 3.3 (postpartum)	—	—	Nitrogen analysis
Nikniaz, 2009 (65)	Iran (LMIC)	Observational	182	26.5 ± 6	26.8 ± 4 (postpartum)	—	Gerber method	—
Nommsen, 1991 (66)/Dewey, 1991 (40)	United States (HIC)	Observational	58	30.4 ± 4.6	23.7 ± 2.7 (postpartum)	Calculated energy	Gravimetric method	Colorimetry
Pérez-Escamilla, 1995 (67)	Honduras (LMIC)	Experimental	141	20.2 ± 3.5	22.7 ± 3.3 (postpartum)	Calculated energy	Gravimetric method	Nitrogen analysis[5]
Perrella, 2016 (69)	Australia (HIC)	Observational	1	28	29.7 (postpartum)	Calculated energy	Creamatocrit	Colorimetry
Pines, 2016 (70)	Israel (HIC)	Observational	45	31.7 ± 3.8	22.2 ± 2.7 (prepregnancy)	Calculated energy	IR spectrophotometry	IR spectrophotometry
Prentice, 2016 (71)	England (HIC)	Observational	614	33.9 ± 4.0	22.8 ± 3.2 (prepregnancy)	Calculated energy	Chromatography	Nitrogen analysis
Rakicioğlu, 2006 (72)	Turkey (LMIC)	Observational	21	27.3 ± 5.4	25.6 ± 4.8 (postpartum)	—	Gravimetric method	Nitrogen analysis[5]
Rice, 2000 (73)	Bangladesh (LMIC)	Experimental	147	26.6 ± 5.7	18.9 ± 1.7 (postpartum)	—	Creamatocrit	—
Rocquelin, 1998 (74)	Congo (LMIC)	Observational	102	27.0 ± 6.8	22.3 ± 4.5 (postpartum)	—	Gravimetric method	—
Ruel, 1997 (75)	Guatemala (LMIC)	Observational	52	25 ± 6	24 ± 3 (postpartum)	—	Other	—
Sadurskis, 1988 (76)	Sweden (HIC)	Observational	23	28.6 ± 2.8	22.2 ± 3.3 (prepregnancy)	Bomb calorimetry	—	—
Shehadeh, 2006 (77)	Israel (HIC)	Observational	41	30.8 ± 5.0	25.9 ± 3.9 (postpartum)	—	Gerber method	Nitrogen analysis
Sims, 2020 (18)[7]	United States (HIC)	Observational	63 normal-weight, 68 overweight	30.4 ± 0.4 normal-weight, 30.5 ± 0.4 overweight	22.3 ± 0.2 normal-weight (prepregnancy), 29.0 ± 0.3 overweight (prepregnancy)	Calculated energy	IR spectrophotometry	IR spectrophotometry
Thakkar, 2013 (80)/Thakkar, 2016 (78)	Singapore (HIC)	Observational	50	31 ± 4	21 ± 6 (postpartum)	Calculated energy	IR spectrophotometry	Colorimetry
Urteaga, 2018 (81)	Bolivia (LMIC)	Observational	18	28.8 ± 5.9	25.6 ± 4.5 (postpartum)	Calculated energy	Creamatocrit	—
van Steenbergen, 1983 (82)	Kenya (LMIC)	Observational	75	27.5 ± 7.5	21.3 ± 3.3 (postpartum)	—	Gravimetric method	Nitrogen analysis
Villalpando, 1992 (83)	Mexico (LMIC)	Observational	30	26.0 ± 6.1	23.8 ± 3.2 (postpartum)	Bomb calorimetry	Jeejeebhoy method	Nitrogen analysis
Yang, 2014 (84)	China (HIC)	Observational	90	27.2 ± 3.9	23.3 ± 3.1 (postpartum)	Calculated energy	IR spectrophotometry	IR spectrophotometry
Young, 2017 (85)	United States (HIC)	Observational	33 normal-weight, 23 overweight	31.1 ± 4.4 normal-weight, 32.5 ± 4.6 overweight	22.1 ± 1.4 normal-weight (postpartum), 27.8 ± 3.0 overweight (postpartum)	Calculated energy	Creamatocrit	Colorimetry

HIC, high-income country; IR, infrared; LMIC, low- and middle-income country; Ref, reference; —, indicates that human-milk energy, fat, or protein were not quantified.

Human-milk samples at time point closest to 3 mo postpartum.

Data are presented as means ± SDs for maternal age and BMI.

BMI categories are underweight (kg/m2; <18.5), normal weight (between 18.5 and 24.9), overweight (between 25.0 and 29.9), and obese (≥30.0).

Only human-milk true protein, as opposed to total protein, was presented. These data were not included in the meta-regression.

Unknown whether human-milk total protein or true protein was assessed.

Data for the Sims (2020) paper were provided directly from the authors (not from the published paper).

The mean BMI of the included studies of term-born infants ranged from 17.8 to 35.3. Of the studies that included term-born infants, there was 1 datapoint (1.4%) with a mean maternal BMI within the underweight category, 46 datapoints (66.7%) within the normal-weight category, 19 datapoints (27.5%) within the overweight category, and 3 datapoints (4.3%) within the obese category. For the studies that evaluated preterm infants, the mean maternal BMI was between 20.7 and 26.4, with 4 studies having a mean BMI in the normal-weight category (66.7%), 2 in the overweight category (33.3%), and none with a BMI in the underweight or obese categories.

Forty datapoints (33 studies) were included in the meta-regression of maternal BMI and human-milk energy, 63 datapoints (56 studies) in the analysis of maternal BMI and human-milk fat, 40 datapoints (37 studies) in the analysis of maternal BMI and human-milk total protein, and 20 datapoints (17 studies) in the analysis of maternal BMI and human-milk lactose (Figure 1). Five studies were included as multiple datapoints for the meta-regression analyses as they included data disaggregated by maternal BMI category (18, 38, 51, 85, 89). Four of these studies separated mothers into 2 BMI categories, whereas 1 included 4 different categories that were each discrete datapoints (51).

Associations between maternal BMI and human-milk energy, fat, and total protein: meta-regression results

Human-milk energy

The meta-regression of the relation of maternal BMI and human-milk energy from 40 datapoints (33 studies) showed no significant association between these variables (β: 3.9 kcal/L; 95% CI: −1.6, 9.5; P = 0.16, I2 = 93.3%) (). These data met the assumptions for normality based on the Q–Q plot (Supplemental Figure 1). Subgroup analyses to ascertain whether quantification techniques (bomb calorimetry or calculated), exclusive breastfeeding, complete breast expression, 24-h collection, BMI measured prepregnancy or postpartum, study design, and country income level did not provide insight into the source of the large heterogeneity ().

FIGURE 2

Meta-regression of the relation between maternal BMI and human-milk energy. The bubble sizes in this meta-regression are proportional to the inverse of the study-level SE for human-milk energy. The solid line represents the linear prediction for the means of human-milk energy as a function of the mean BMI observed at the study level (β: 3.9 kcal/L; 95% CI: −1.6, 9.5; P = 0.16, I2 = 93.3%, n = 40 datapoints).

The subgroup meta-regression of the 9 datapoints (8 studies) using bomb calorimetry showed no significant relation (β: 7.6 kcal/L; 95% CI: −21.3, 36.5; P = 0.55, I2 = 91.6%) nor did the meta-regression of the 31 datapoints (25 studies) using calculated energy (β: 4.1 kcal/L; 95% CI: −2.1, 10.2; P = 0.19, I2 = 93.9%).

Human-milk fat

The meta-regression of maternal BMI and human-milk fat from 63 datapoints (56 studies) showed that for each 1-kg/m2 higher BMI, human-milk fat was 0.56 g/L higher (95% CI: 0.034, 1.1; P = 0.04, I2 = 93.7%). Visual examination of the meta-regression and the Q–Q plot suggests that the relation was linear and met assumptions of normality, respectively ( and Supplemental Figure 2).

FIGURE 3

Meta-regression of the relation between maternal BMI and human-milk fat. The bubble sizes in this meta-regression are proportional to the inverse of the study-level SE for human-milk fat. The solid line represents the linear prediction for the means of human-milk fat as a function of the mean BMI observed at the study level (β: 0.56 g/L; 95% CI: 0.034, 1.1; P = 0.04, I2 = 93.7%, n = 63 datapoints).

When including only the gravimetric method to quantitate fat and excluding other techniques, there were 17 datapoints (16 studies) in this meta-regression. These showed a similar trend between maternal BMI and human-milk fat, but the relation was not significant (β: 0.84 g/L; 95% CI: −0.45, 2.1; P = 0.19, I2 = 93.6%). The subgroup analysis of 46 datapoints (40 studies) using all other techniques also did not show a significant relation between maternal BMI and human-milk fat (β: 0.44 g/L; 95% CI: −0.17, 1.1; P = 0.15, I2 = 93.6%). There was a significant positive relation between maternal BMI and human-milk fat for the 34 datapoints (30 studies) that used gravimetric or infrared spectrophotometry (β: 0.58 g/L; 95% CI: 0.016, 1.2; P = 0.04, I2 = 90.0%). Other commonly used human-milk-fat techniques were examined in meta-regression subgroup analyses and showed similar trends to the primary meta-regression (Supplemental Table 6). Additional subgroup analyses did not reveal any specific source for the high I2 values in the meta-regression analyses.

Human-milk total protein

There was no significant association between maternal BMI and human-milk total protein based on the meta-regression of 40 datapoints (37 studies) (β: 0.13 g/L; 95% CI: −0.16, 0.41; P = 0.37, I2 = 99.1%) (). Examination of the Q–Q plot suggests that the data met the assumptions of normality (Supplemental Figure 3).

FIGURE 4

Meta-regression of the relation between maternal BMI and human-milk total protein. The bubble sizes in this meta-regression are proportional to the inverse of the study-level SE for human-milk total protein. The solid line represents the linear prediction for the means of human-milk total protein as a function of the mean BMI observed at the study level (β: 0.13 g/L; 95% CI: −0.16, 0.41; P = 0.37, I2 = 99.1%, n = 40 datapoints).

In subgroup analyses, including studies with only nitrogen or amino acid analysis used to quantitate protein in the meta-regression (16 datapoints, 15 studies), there was again no significant association (β: 0.21 g/L; 95% CI: −0.31, 0.73; P = 0.40, I2 = 97.7%). The meta-regression of 24 datapoints (22 studies) using other techniques also showed no significant association (β: 0.20 g/L; 95% CI: −0.18, 0.58; P = 0.28, I2 = 99.0%). Colorimetry and infrared spectrophotometry, 2 commonly used techniques to quantitate human-milk total protein, were assessed in meta-regression subgroup analyses one at a time in combination with nitrogen and amino acid analysis (Supplemental Table 6). The association between maternal BMI and total protein was not statistically significant in either of these meta-regression analyses, although there was a trend towards a positive relation when examining the 29 datapoints that used nitrogen and amino acid analysis and colorimetry (β: 0.28 g/L; 95% CI: −0.041, 0.61; P = 0.084, I2 = 97.5%).

As with the subgroup analyses for maternal BMI and human-milk energy and fat, respectively, exploration of subgroups did not give insight into the sources of the high statistical heterogeneity based on the I2 values.

Human-milk lactose

An additional meta-regression of human-milk lactose, which included 20 datapoints (17 studies), did not show a significant relation between maternal BMI and lactose (β: −0.45 g/L; 95% CI: −1.8, 0.88; P = 0.49, I2 = 99.6%) ().

Certainty of the evidence

The GRADE assessment indicated that the certainty of the evidence for human-milk energy was low, whereas the certainty of evidence for human-milk fat and total protein were very low (). The risk of bias was low for all 3 of the outcomes examined. However, the quality of the evidence scores for all 3 outcomes was downgraded as there were limited studies including mothers with BMI in the underweight and obese categories. Second, the evidence quality score for all outcomes was downgraded due to high unexplained heterogeneity in the meta-regression analyses for all 3 outcomes. Finally, the certainty of the 3 outcomes was downgraded due to differences in exclusivity of breastfeeding across studies, the range of strategies used to collect human-milk samples, including whether or not a complete breast expression was performed, and different techniques used to quantitate human-milk fat and total protein. Specifically, 36 studies out of 62 studies (58.1%) including term-born infants stated that mothers were exclusively breastfeeding. Thirty-four studies (54.8%) specified that they collected fully expressed human milk. Sixteen studies (25.8%) collected human milk over 24 h, which is the ideal method for collecting human milk, particularly for human-milk energy and fat analysis. However, subgroup analyses did not show differences in statistical heterogeneity in relation to these factors.

Discussion

Results from this systematic review showed a positive association between maternal BMI and human-milk fat in term-born infants between 1 and 6 mo postpartum. Specifically, the meta-regression analysis of maternal BMI and human-milk fat in the present systematic review (63 datapoints) suggests that every 1-kg/m2 increment in maternal BMI is associated with a 0.56-g/L higher milk-fat concentration. Across a modest BMI difference of 5 kg/m2 this translates to ∼7.2% higher human-milk fat. No significant association was observed between maternal BMI and human-milk energy or total protein.

Evidence for the significant association between maternal BMI and human-milk fat was of very low quality, as assessed using the GRADE approach, meaning that the certainty in these results is highly limited. The certainty of the evidence for human-milk energy and total protein was low and very low, respectively, using the GRADE assessment (20). One of the reasons was that several human-milk quantification techniques were used, particularly for fat and total protein. However, subgroup analyses for each of the 3 outcomes did not indicate that these methods explained the considerable statistical heterogeneity.

Several years ago, Prentice et al. (93) in their summative review examined the relation between average reported maternal BMI and human-milk volume, energy, and fat composition in the literature between 1980 and 1992 from countries of all income levels. They reported that a number of original studies in the literature found positive associations between maternal BMI and human-milk fat but not energy, albeit acknowledged that other studies reported no significant association or a negative association between maternal BMI and human-milk fat. This review did not consider the relation between maternal BMI and human-milk protein. It is noteworthy that the highest mean BMI of studies in the Prentice et al. review was below the BMI category cutoff for overweight. In their recent systematic review, Leghi et al. (6) examined original studies that tested the association between maternal BMI and human-milk macronutrient content and reported that mothers with overweight or obesity have higher human-milk fat than normal-weight mothers based on a meta-analysis of 6 studies; no difference was found in the association between maternal adiposity and human-milk protein (n = 5 studies).

There was no significant relation between maternal BMI and human-milk energy based on the meta-regression, similar to the earlier findings by Prentice et al. Based on the positive relation between maternal BMI and human-milk fat, it could be hypothesized that energy would also be greater from mothers with overweight or obesity since fat is the most energy dense of the macronutrients and contributes ∼50% of milk energy content (94–96). Additional original research studies are needed to elucidate the simultaneous relations between maternal BMI and human-milk energy and macronutrients.

Under normal circumstances, infants modify their consumption to some degree based on the energy and macronutrient composition of their enteral feeds (97). Regulation of volume intake was first supported by the early controlled study by Fomon et al (98). where term-born infants exhibited an early capacity to regulate intake in response to the energy density of formula provided. The association between macronutrient concentration and volume intake has been reported by numerous others in both formula-fed and breastfed infants subsequently (97, 99–101). Whether infants are able to adjust the volume of intake over the range of fat concentrations observed in the present systematic review is unknown and is worthy of future research. Understanding these associations in mothers across the BMI spectrum is important as mothers with obesity often have delayed lactogenesis and reduced exclusivity and duration of breastfeeding (97, 102). Beyond infancy, data pooled from 37 cohorts have shown that higher maternal prepregnancy BMI is associated with overweight and obesity in children (103), although it is unclear how human-milk energy and macronutrients might mediate this relation. Other milk components, including human-milk oligosaccharides and other bioactives, may also be related to growth and require further exploration in relation to maternal factors (104).

Limitations of this systematic review include the fact that BMI does not differentiate between lean body mass and fat mass and may not represent metabolic status precisely. BMI was considered in mothers either before pregnancy or during lactation, depending on available data, which may reduce the precision of results. Only 6 studies reporting on human-milk macronutrient composition from mothers of preterm infants met the inclusion criteria and none of these had a mean maternal BMI in the underweight or obese categories. We are therefore unable to make conclusions about the associations between maternal BMI and human-milk energy, fat, and/or total protein in this population. Even among mothers of term-born infants, there was a limited number of studies with women across the spectrum of BMI, with just 1 study with a mean BMI in the underweight category and 3 studies with a mean BMI in the obese category. Furthermore, only the human-milk collection time point closest to 3 mo was considered for each of the 21 studies that collected human milk at multiple times. While there is variability in human-milk composition for some nutrients over the first 6 mo postpartum, and this conceivably may be impacted by maternal BMI, a previous systematic review reported comparable human-milk energy, protein, and fat between 1 and 6 mo postpartum (105).

With regard to the statistical analysis, the variance in maternal BMI as the predictor variable was not reflected in the meta-regression. To account for this, for studies that categorized mothers by >1 BMI category, these were included as separate datapoints in the analyses; only 5 studies disaggregated the data by BMI category. Further, given the limited number of studies at the low and high end of the maternal BMI distribution, we cannot be certain the relations between maternal BMI and human-milk fat and total protein are linear; however, visual inspection of the meta-regression suggests this is the case. That said, further studies are needed to confirm the pattern of association between maternal BMI and human-milk composition.

In conclusion, this systematic review showed a statistically significant positive association between maternal BMI and human-milk fat in term-born infants between 1 and 6 mo. No significant associations were found between maternal BMI and human-milk energy or total protein. This systematic review further found the certainty of evidence for these findings was low to very low. Future studies investigating the relation between maternal nutritional status, variation in human-milk fat content, and the implications for child growth and development are required.

Click here for additional data file.