Infant Feeding and Weight Gain: Separating Breast Milk From Breastfeeding and Formula From Food

Pregnant women were enrolled in the Canadian Healthy Infant Longitudinal Development (CHILD) birth cohort (www.childstudy.ca) between 2009 and 2012.¹³ Infants born before 38 + 0 weeks or missing gestational age or anthropometric data were excluded from the current analysis (Supplemental Fig 3). This study was approved by the Human Research Ethics Boards at the University of London, McMaster University, and the Universities of Manitoba, Alberta, Toronto, and British Columbia.

Feeding exposures are summarized in Table 1. Feeding in hospital was recorded by nursing staff and validated by chart review for Manitoba participants with available hospital records (N = 847), revealing strong agreement for breastfeeding (98.3% agreement) and formula supplementation (87.6%). Subsequent feeding was reported by mothers at 3, 6, 12, 18, and 24 months postpartum, including the following: breastfeeding initiation and cessation, feeding of expressed breast milk, use of formula, other fluids, and foods. The end of exclusive breastfeeding was defined as the introduction of any formula, food, juice, or nonhuman milk. Water and vitamin supplements were not considered.

The primary outcome was BMI z score. In large cohort studies, researchers have demonstrated its equivalence or superiority as a predictor of childhood obesity, compared with weight-for-length.⁴ Secondary outcomes included overweight, weight gain velocity, and rapid weight gain. Weight and length were recorded at birth by nurses and at 12 months (mean: 12.5 ± 1.5 months) by CHILD staff, following a standardized protocol. Sex-specific weight-for-age (WFA) and BMI-for-age z scores were calculated according to the World Health Organization (WHO) Child Growth reference.¹⁵ BMI z score was dichotomized to define overweight (z score >2).¹⁶ Weight gain velocity was calculated as the change in WFA z score from birth to 12 months³ and dichotomized to define rapid weight gain (weight gain velocity >0.67).³ 

Potential confounders based on existing literature¹⁷ were documented from hospital records (maternal age, diabetes, mode of delivery, parity, infant sex, gestational age, and birth weight) or self-reported (maternal ethnicity, education, smoking during pregnancy, and diet quality¹⁸) (Supplemental Table 6). Maternal prepregnancy BMI was determined from measured height and self-reported prepregnancy weight (N = 1751) or estimated from measured weight 12 months postpartum (N = 837). Validation against health records in a subset (N = 224) revealed strong agreement for both measures (mean difference: −1.0 kg [95% confidence interval (CI): −1.5 to −0.4] and +1.3 kg [95% CI: 0.5 to 2.2], respectively).

Potential confounders were screened in bivariate analyses (t tests, analysis of variance [ANOVA], χ²) to identify associations with feeding exposures and anthropometric outcomes. Multivariable regression models were used to estimate the independent effects of feeding exposures on anthropometric outcomes, with adjustment for infant sex, birth weight, established obesity risk factors (maternal smoking, prepregnancy BMI), and potential confounders identified through bivariate screening (study site, maternal age, parity, ethnicity, education, marital status, and infant gestational age). We also applied a propensity score approach, in which the same covariates were used to predict feeding exposures in multinomial regression models, and outcome models were weighted for the inverse predicted probability of feeding exposures.

Next, because different feeding exposures may be interrelated, models were mutually adjusted to evaluate the independent effects of breastfeeding exclusivity, solid food introduction, and breastfeeding duration. For example, mothers who cease breastfeeding earlier may also introduce foods earlier to their infant; these exposures were evaluated together in the same model to determine if one explained the other, or if they were independently associated with infant BMI.

Interaction terms were tested in regression models to evaluate the a priori hypothesis that breastfeeding effects may be modified by maternal obesity or infant sex.

Results are presented as crude and adjusted odds ratios (aORs) and adjusted β estimates (aβ) with 95% CIs. Multivariable regression analyses were conducted for mother-infant dyads with available feeding and covariate data and confirmed in the full eligible subset (N = 2553) following multiple imputation of missing covariate data. Multiple imputation (20 imputed data sets) was performed with fully conditional specification (chained equations) by using the covariates listed above. All analyses were performed by using SAS version 9.4 (SAS Institute, Inc, Cary, NC).

Of 2870 eligible term infants, 2553 (89.0%) were assessed at 12 months (Supplemental Fig 3, Table 2). The mean maternal age was 32.4 ± 4.6 years, and the mean prepregnancy BMI was 24.7 ± 5.4. The majority of mothers were white (75.1%), had a postsecondary (57.0%) or postgraduate (19.8%) degree, and delivered vaginally (75.7%). The mean infant BMI z score at 12 months was +0.21 ± 1.07. Nearly one-fourth of infants (N = 566, 22.3%) experienced rapid weight gain, and 126 (4.9%) were overweight at 12 months. Infants with missing outcome data were more likely to be of nonwhite ethnicity, have mothers who smoked, and be breastfed for a shorter duration.

Nearly all women in the CHILD cohort (96.6%) initiated breastfeeding (Supplemental Table 7, Fig 1). The majority (73.8%) continued beyond 6 months, and 43.5% were still breastfeeding at 12 months. The median duration of exclusive breastfeeding was 4.0 months (interquartile range [IQR]: 0.5–5.0), and the median duration of any breastfeeding was 10.0 months (IQR: 5.0–14.0). Most infants (73.9%) received solid foods before 6 months, although only 17.9% received them before 4 months.

At 3 months, 1686 infants (60.8%) were exclusively breastfed according to maternal report; however, 460 of these infants (27.3% of exclusively breastfed infants, 16.6% overall) briefly received formula supplementation in hospital. Over half of exclusively breastfed infants had received some expressed breast milk (N = 879, 54.6% of exclusively breastfed infants; 33.0% of all infants). By 6 months, only 18.0% of infants were exclusively breastfed. An additional 59.4% were partially breastfed, including 33.6% receiving solid foods, 5.4% receiving formula, and 20.4% receiving both formula and solids.

Breastfeeding duration and exclusivity were positively associated with maternal age and education and negatively associated with maternal obesity, smoking, cesarean delivery, and single parenthood (Table 3). Breastfeeding rates also differed by study site and were highest in Vancouver. Formula supplementation in hospital and feeding of expressed breast milk after discharge were more common in first-time mothers. Many of these factors were also associated with infant BMI and/or weight gain velocity (Table 3), and all were balanced effectively in the propensity score analyses (Supplemental Tables 8 and 9).

Infants who achieved the WHO recommendation¹⁴ of 6 months of exclusive breastfeeding had a mean BMI z score of −0.04 ± 1.06, closely matching the WHO infant growth standard (Fig 2A, Table 4). BMI z scores were half a SD higher among infants who were not breastfed for at least 6 months (mean: +0.51 ± 1.07 above the WHO standard; crude β = +.54; 95% CI: .41 to .67). This association remained significant after adjustment for study site, maternal age, prepregnancy BMI, ethnicity, education, marital status, smoking, mode of delivery, parity, infant sex, gestational age, and birth weight (aβ = +0.44; 95% CI: 0.30 to 0.57). Partial breastfeeding with formula supplementation appeared to have an intermediate effect (mean: +0.30 ± 1.08; aβ = 0.25; 95% CI: 0.13 to 0.38), whereas partial breastfeeding without formula (ie, with solid foods only) was not significantly associated with infant BMI (mean: +0.07 ± 1.03; aβ: +0.07; 95% CI: −0.05 to 0.19).

Breastfeeding duration was inversely associated with BMI z score in a dose-dependent manner (aβ = +0.48 for <3 months, +0.29 for 3 to 6 months, +0.19 for 6 to 12 months, compared with breastfeeding beyond 12 months; P for trend <.0001) (Fig 2A, Table 4). These associations were relatively unchanged after further adjustment for the timing of introduction to solid foods (eg, mutually adjusted β = +0.48, +0.32, +0.18 for <3 months) (Table 4).

Introducing solid foods before 5 months was associated with a higher BMI z score (aβ = +0.17; 95% CI: 0.04 to 0.29) compared with introduction after 6 months. This association was attenuated after adjusting for breastfeeding duration (mutually adjusted β = +0.12; 95% CI: −0.01 to 0.25). By contrast, introducing solid foods between 5 and 6 months was not significantly associated with BMI z score (mutually adjusted β = +0.05; 95% CI: −0.06 to 0.16). Thus, consistent with the above results, shorter breastfeeding duration and introduction of solid foods before 5 months (but not between 5 and 6 months) was associated with a higher BMI z score at 12 months.

Mutual adjustment for breastfeeding exclusivity and duration revealed that these factors were independently associated with infant BMI (Table 4). Effect estimates for partial breastfeeding with formula were attenuated but remained significant after adjustment for breastfeeding duration (eg, aβ = +0.24; 95% CI: 0.14 to 0.35 for formula supplementation before 3 months; attenuated to aβ = +0.14; 95% CI: 0.03 to 0.26 after adjustment for breastfeeding duration).

Exclusively breastfed infants receiving some expressed breast milk had higher BMI z scores than those receiving only direct breast milk (mean: +0.14 ± 1.00 vs −0.02 ± 1.06; aβ +0.12; 95% CI: 0.01 to 0.23) (Fig 2A, Table 4); however, they remained leaner than infants who were partially breastfed (aβ +0.28; 95% CI: 0.16 to 0.39) or not breastfed (aβ +0.45; 95% CI: 0.30 to 0.59).

Among infants who were “exclusively breastfed” for at least 3 months according to maternal report, those who briefly received formula supplementation as neonates had slightly higher BMI z scores at 12 months (mean: +0.11 ± 1.04 vs +0.04 ± 1.04); however, this difference was not significant in crude or adjusted models (aβ = +0.05; 95% CI: −0.07 to 0.17) (Fig 2A, Table 4).

Consistent with BMI z score results, there was an inverse dose-dependent association between breastfeeding duration and the risk of overweight at 12 months (P < .0001) (Table 4). The risk of overweight was threefold higher among infants who were not breastfed at 6 months compared with those who were exclusively breastfed (8.3% vs 2.4%, aOR = 3.20, 95% CI: 1.58 to 6.51). There was an intermediate association for partial breastfeeding with formula (6.1%, aOR = 2.03, 95% CI: 1.00 to 4.11) and no association for partial breastfeeding without formula (3.2%, aOR = 0.93, 95% CI: 0.44 to 1.97). Brief formula supplementation in hospital was not associated with overweight risk. Similar patterns of association were observed for weight gain velocity and risk of rapid weight gain (Fig 2B, Table 5).

Results were confirmed after inverse probability weighting for propensity scores and essentially unchanged with multiple imputation of missing data (Supplemental Table 10). There was no evidence of effect modification by infant sex or maternal obesity (not shown).

In the prospective CHILD birth cohort, we found that breastfeeding was inversely associated with weight gain velocity, BMI, and overweight risk in the first year of life. These associations were dose dependent (stronger with longer and more exclusive breastfeeding) and independent of maternal BMI and socioeconomic status, using multiple approaches to address confounding bias. After controlling for these and other potential confounders, breastfeeding cessation before 6 months was associated with a twofold increased risk of rapid weight gain, a +0.44 SD increase in BMI by 12 months of age, and a threefold increased risk of overweight, compared with exclusive breastfeeding. These effects are substantial, surpassing the estimated effect of maternal obesity (+0.20, in the same multivariable model).

Further analysis of specific feeding practices revealed stronger associations for direct breastfeeding versus expressed breast milk and stronger attenuation from supplementation with formula versus solid foods. Finally, brief formula supplementation during the neonatal period did not measurably alter these associations, so long as exclusive breastfeeding was established and sustained for at least 3 months after hospital discharge.

A major strength of this study is the prospective collection of infant feeding data, including many details that are rarely captured in other studies. Still, these exposure measurements were limited by a lack of quantitative information about the amount of expressed milk or formula consumed. Also, because of the high breastfeeding initiation rates in this cohort, we had limited power to examine the impact of never breastfeeding, which is an important concern in other settings. Another limitation is that we did not capture information about mothers’ intentions to breastfeed, feeding styles (eg, paced bottle-feeding, infant-led weaning, level of maternal control), or reasons for supplementing, pumping, and weaning.

We used multiple statistical approaches to consider many potential confounders, including sociodemographic factors (maternal age, ethnicity, education, and marital status) to address the social patterning of breastfeeding that is reported in other populations¹⁷ and confirmed in our cohort. However, residual confounding remains possible in this observational study. Finally, although breastfeeding rates in the CHILD cohort are nationally representative,¹⁹ our findings may not apply in other settings with different breastfeeding practices and policies or in populations challenged with undernutrition and stunting rather than overnutrition and obesity.

Our findings are consistent with previous observational studies, demonstrating that breastfeeding is inversely associated with infant growth velocity,²⁰ BMI,²¹ and overweight.^7,17,21 In contrast, a randomized trial of breastfeeding promotion did not affect infant WFA²²; however, it did not include a nonbreastfed control group, and through observational analyses, it was confirmed that infants exclusively breastfed for at least 6 months had lower WFA compared with those weaned or supplemented earlier.²² 

We provide new evidence for the independent and dose-dependent effects of breastfeeding exclusivity and duration, which are rarely examined simultaneously. In our study, these effect estimates were attenuated but remained significant in mutually adjusted models. The attenuation is expected because formula supplementation can decrease milk supply or may reflect breastfeeding difficulties, leading to shorter breastfeeding duration²³; however, our results show that sustained breastfeeding is beneficial even if formula supplementation occurs.

We uniquely examined the mode of breast milk feeding, distinguishing bottled breast milk from direct breastfeeding at the breast. Few studies make this distinction, despite increasing trends in milk expression.⁹ In our study, over half (55%) of exclusively breastfed infants received some breast milk in a bottle; these infants had “intermediate” BMI z scores and weight gain velocities that were higher than infants fed exclusively at the breast, but lower than infants receiving formula. This finding is consistent with evidence that exclusively bottle-fed infants gained more weight than infants fed at the breast, regardless of the milk type (breast milk or formula) in the bottle.¹¹ 

Together, these studies suggest that direct breastfeeding confers the strongest protection against rapid weight gain and overweight, whereas bottled breast milk provides intermediate protection. This has important policy implications because the primary reason for feeding bottled breast milk is returning to work, which is strongly impacted by national and institutional maternity leave policies.

Another novel aspect of our study is the distinction between partial breastfeeding supplemented with formula versus solid food. At 6 months, almost half (43%) of partially breastfed infants were receiving formula, whereas the remainder were being supplemented with solids only. Compared with exclusive breastfeeding, partial breastfeeding with formula was associated with faster weight gain (aβ: +0.29), higher BMI (aβ: +0.25), and a twofold increased risk of overweight by 12 months, whereas partial breastfeeding without formula (ie, with solids only) was not significantly associated with these outcomes. This is a noteworthy finding because recent evidence indicates that introducing certain “allergenic” foods before 6 months may be beneficial for allergy prevention.²⁴ The optimal timing for introducing these foods remains to be determined, but our study suggests that introducing solids between 5-6 months does not adversely affect obesity-related outcomes, with a recent review in which researchers concluded there is little evidence of adverse outcomes associated with introducing solids before 6 months in developed countries.²⁵ 

We uniquely investigated the potential consequences of formula supplementation in hospital. Few researchers have addressed this question, yet many neonates briefly receive formula, often without medical indication.²⁶ In our cohort, a striking 27% of exclusively breastfed infants (according to maternal report) had briefly received formula during the neonatal period. These infants were not significantly different from their exclusively breastfed counterparts in terms of weight gain or BMI at 12 months, suggesting that brief formula supplementation at birth does not adversely impact weight gain, so long as the infant is exclusively breastfed after hospital discharge. This is an important caveat because supplementation can discourage or delay breastfeeding initiation²⁷ and reduce a new mother’s confidence in her ability to breastfeed, which may lead to early cessation or sustained supplementation. Further research is needed to determine if brief supplementation influences other processes such as inflammation, immunity, or the gut microbiome.

Our results suggest that shorter breastfeeding duration, feeding bottled breast milk, and formula supplementation all independently influence infant weight gain, BMI, and overweight risk. Potential mechanisms for these effects include biological differences in the macronutrient profiles of formula versus breast milk²⁸ or differential effects on the gut microbiota, which are profoundly impacted by formula²⁹ and contribute to energy absorption and weight gain.¹² In addition, many bioactive components of human milk are absent from formula and may be altered during expression and storage¹⁰; these include hormones that regulate satiety, microbiota that seed the infant gut, and oligosaccharides that support microbiota development.³⁰ 

Our results also suggest that feeding method is important, which is consistent with evidence that the “baby-led” nature of breastfeeding promotes satiety responsiveness later in childhood.³¹ It is thought that differential “programming” of satiety and self-regulation results from bottle-feeding versus breastfeeding because infants fed at the breast actively suckle and self-regulate, whereas those fed from a bottle (regardless of its contents) are more passive and may not learn to appropriately balance energy intake.¹¹ 

The results of this study highlight the importance of documenting detailed information about infant feeding practices and may help explain the inconsistencies observed across previous studies in which these details were not precisely captured. Further work by researchers incorporating these measures (eg, duration, exclusivity, and mode of breastfeeding; type and timing of complementary foods) is required to replicate our observations and determine their generalizability to other settings and populations. Extended analysis of weight gain and body composition is also required to determine if these associations persist and influence obesity later in life.

Additional research is warranted to investigate the potentially differential effects of specific formulas (eg, containing prebiotics or probiotics) and first foods, to establish exposure thresholds (eg, differentiate between “any” and “predominantly” expressed milk feeding), to explore the role of different feeding styles, and to understand and address mothers’ motivations and challenges related to infant feeding. Finally, it is important to study and define the causal mechanisms that mediate the observed associations between infant feeding practices, weight gain, and overweight risk. These future directions are essential to inform and optimize infant feeding guidelines and develop effective early-life interventions for obesity prevention, including initiatives to support breastfeeding and alternative solutions for those who cannot be breastfed.

This study confirms that sustained and exclusive breastfeeding is associated with favorable anthropometric outcomes during infancy, and contributes novel evidence regarding common feeding practices that are rarely addressed in other studies. Firstly, formula supplementation of breastfed infants significantly attenuated the observed associations, whereas complementary feeding of solid foods introduced between 5 and 6 months had no impact. Secondly, feeding expressed breast milk appeared to have a weaker beneficial effect compared with direct feeding at the breast, although expressed milk remained beneficial compared with formula. Finally, brief formula supplementation during the neonatal period did not measurably alter these associations so long as exclusive breastfeeding was established and sustained for at least 3 months. Altogether, this study provides new evidence to inform feeding recommendations and guide further research about infant feeding practices and how they influence the development and prevention of childhood obesity.

aβ

adjusted β estimate

ANOVA

analysis of variance

aOR

adjusted odds ratio

BMI

body mass index

CHILD

Canadian Healthy Infant Longitudinal Development

CI

confidence interval

IQR

interquartile range

WFA

weight-for-age

WHO

World Health Organization

We thank all the families who took part in this study and the whole CHILD team, which includes interviewers, nurses, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, and receptionists. We also thank Dr Atul Sharma and Ms Allison Feely (George and Fay Yee Centre for Healthcare Innovation) for their statistical expertise.