Relationship Between Breastfeeding and Early Childhood Obesity: Results of a Prospective Longitudinal Study from Birth to 4 Years

Abstract

Aim:

To study a potential link between breastfeeding in infancy and obesity at age 4.

Materials and Methods:

A total of 30,508 infants born during 2002–2007 from the databases of the Preventive Child Health Services in two Swedish counties and from national registers were studied. The outcome variable was obesity at age 4. Analyses were conducted by logistic regression models using the methodology of generalized estimating equations. Analyses were adjusted for child sex and maternal anthropometric and sociodemographic variables.

Results:

In unadjusted analyses, any breastfeeding up to 9 months was linked to successively decreasing odds ratios (ORs) for obesity at age 4 (ORs 0.78–0.33), however, not significantly for 1 week and 2 months of breastfeeding. In adjusted analyses, the same pattern remained statistically significant for breastfeeding for 4 (OR 0.51), 6 (OR 0.55), and 9 (OR 0.47) months. Child sex, maternal education, maternal body mass index, and maternal smoking additionally influenced child obesity.

Conclusion:

Breastfeeding duration for at least 4 months may contribute independently to a reduced risk for childhood obesity at 4 years.

Introduction

Studies of breastfeeding as protecting against child overweight and obesity have come to conflicting results.^1–4 Grummer-Strawn and Rollins emphasize the difficulty of drawing conclusions from breastfeeding studies.⁵ Several studies suffer from limitations because of small populations, lack of measures of feeding behavior,⁶ protein content in infant formulae,⁷ and confounders causing methodological difficulties.^4,8–11 Few studies have applied a longitudinal prospective design. Warnings against the risk of bias and confounders have been raised by the WHO¹² and Martin et al.¹³ A randomized design has been considered optimal, but would not be ethically acceptable. Smithers et al. suggested triangulation, that is, to use different methods, an approach that does not share the same bias.⁴

Exclusive breastfeeding for 6 months is recommended by the WHO,¹⁴ but this goal is difficult to reach.^15,16 Recommendations are discordant when the WHO and the American Academy of Pediatrics, on the one hand, recommend exclusive breastfeeding until 6 months, and Guidelines from the European Society of Pediatric Gastroenterology, Hepatology, and Nutrition, on the other, state that complementary foods may be introduced between the 17th and the 26th week of life.¹⁷

Sociodemographic factors are important determinants of breastfeeding duration and could be targeted by preventive initiatives. We know that young mothers, smoking mothers, and mothers with low education initiate breastfeeding to a lower extent than others.^18,19 According to an attitude survey by Chezem, mothers who intended to breastfeed exclusively had a higher education and were older than mothers planning mixed feeding.²⁰

Owing to the availability of a large prospective database, we have had a unique opportunity to study, with a longitudinal design, possible connections between any breastfeeding of varying duration in infancy and obesity at 4 years of age, adjusting for biological and socioeconomic confounders.

Aim

To study breastfeeding in infancy as a potential predictor for childhood obesity at age 4 in the presence of childhood and maternal biological and socioeconomic variables.

Materials and Methods

The total population consisted of 38,118 four-year-old children born during 2002–2007 in the Swedish counties of Uppsala and Orebro and registered as residents in the respective county by the 31st of December, the year when the child was 4 years old. The study population was limited to children with growth information at 4 years measured within ±2 months from birth date and with height and weight within ±3 SD.²¹ Children lacking growth information, with growth information out of limits (n = 5,221, 13.7%), or lacking information on breastfeeding from their first year of life (n = 2,389, 6.3%) were excluded. The study population thus consisted of 30,508 children.

Study variables

Data on weight and height of the child at age 4, breastfeeding status at child's age 1 week, 2, 4, 6, 9, and 12 months, the child's sex, first child of the mother, and maternal smoking at child age 4 weeks were collected from the databases of statistics of the Preventive Child Health Care Services in Uppsala and Orebro County Councils. The databases consist of core variables from child health records reported by the nurses in the child health centers (CHCs). Data in the databases were collected with parental consent and linked to the personal identification number of the child. Educational data were collected from the LISA database²² at Statistics Sweden. Maternal age at delivery and maternal weight and height (measured at the first visit to the prenatal healthcare) were collected from the Medical Birth Registry²³ at the Swedish National Board of Health and Welfare.

Outcome variable

Based on information on weight and height at age 4, the child's body mass index (BMI) was calculated (weight in kilograms/height in meters²) and transformed into four sex- and age-specific ISO-BMI groups, i.e., with cutoffs particularly adapted for children, (underweight, normal weight, overweight, and obesity) according to Cole et al.^24,25 The outcome variable was thereafter dichotomized as obese/not obese.

Independent variables

Breastfeeding status at 1 week, 2, 4, 6, 9, and 12 months was used as a basis for calculating the main independent variable, breastfeeding length. Breastfeeding was defined as the child being breastfed exclusively or partially at each specific time point. Based on those data, the outcome variable was defined as “never” if the child was registered as not breastfed at 1 week of age and the following time points. If the child was registered as breastfed up to one specific time point but not on the following time point, the value of the outcome variable was set to the former time point, for example, breastfed at 1 week, 2, and 4 months but not at 6 months resulted in the value of breastfeeding length equal to 4 months. Mother's first child was defined as the child being the mother's first biological or adopted child. Information on maternal smoking habits, defined as daily smoking, was dichotomized as smoking/not smoking. The educational level of the mother was divided into three categories: low (elementary), medium (upper secondary school or college ≤2 years), and high (college ≥3 years or graduate). Maternal age was divided into six categories (≤18, 19–24, 25–29, 30–34, 35–39, and ≥40 years). Based on data on maternal weight and height, the mother's BMI was calculated (in weight in kilograms/height in meters²) and divided into four BMI groups (underweight, normal weight, overweight, and obesity). Maternal data from Statistics Sweden and the National Board of Health and Welfare were linked to child data through the Statistics Sweden Multigenerational Register.

Statistical methods

Logistic regression models were used to analyze the effect of the independent variables on child obesity. As data for this study were collected over six calendar years and many children had siblings in the material, analysis was conducted by logistic regression models using the methodology of generalized estimating equations (GEEs), in which an exchangeable correlation structure was used to control for the dependence of infants on the same mother.

The associations of the independent variables with child obesity at age 4 were analyzed by applying two models. In Model 1, the individual association of each independent variable with the outcome variable was investigated. In Model 2, the impact of breastfeeding length on child obesity at age 4 was adjusted for all independent variables simultaneously (for variable categories used as references, see Table 3). Results are expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Interaction analyses were applied for each category of the independent variables and each level of breastfeeding duration. Statistical analyses were performed using the SPSS^®22 for Macintosh software.²⁶ The study was conducted within the SHIP study (Social differences in Health among Infants and Preschool children in Uppsala and Orebro Counties) and approved by the local medical ethics committee at the Medical Faculty of Uppsala University, registration number 2008/367.

Results

The study population (n = 30,508) consisted of children of 24,168 mothers. Of these mothers, 75% (n = 18,227) had only one child, 23% (n = 5,542) had two children, and 2% (n = 399) had three to five children. The dropout rate in the independent variables was low, between 0.8% and 2.0%, except for maternal BMI, in which 17.1% of the study population had missing data.

Background characteristics of the study population are presented in Table 1 and outcomes by independent variables are in Table 2. Mothers of 98.1% of all children initiated breastfeeding, 81.1% of the children were breastfed at 4 months of age, and mothers of 70.4% of the children were still breastfeeding at 6 months (Table 1). At age 4, 11.2% of the children were overweight and 1.9% obese. Of children who were never breastfed, 3.6% were obese at age 4 compared with 1.3% of children breastfed for at least 9 months (Table 2).

Table 1.

Characteristics of Study Population

	Frequency	Percent
Body mass index, child
Underweight	351	1.2
Normal weight	26,165	85.8
Overweight	3,421	11.2
Obesity	571	1.9
Any breastfeeding
Never breastfed	578	1.9
At least to
1 week	2,234	7.3
2 months	2,918	9.6
4 months	3,307	10.8
6 months	10,438	34.2
9 months	5,470	17.9
12 months	5,563	18.2
Year of birth of child
2002	2,554	8.4
2003	5,312	17.4
2004	5,493	18.0
2005	5,313	17.4
2006	5,725	18.8
2007	6,111	20.0
Child sex
Boy	15,624	51.2
Girl	14,884	48.8
Mother's first child
No	16,894	55.4
Yes	12,992	42.6
Missing	622	2.0
Educational level of mother
Low	2,847	9.3
Medium	16,661	54.6
High	10,771	35.3
Missing	229	0.8
Maternal age at birth (years)
≤18	233	0.8
19–24	3,739	12.3
25–29	9,223	30.2
30–34	10,900	35.7
35–39	4,865	15.9
≥40	997	3.3
Missing	551	1.8
Body mass index, mother
Underweight	563	1.8
Normal weight	15,390	50.4
Overweight	6,249	20.5
Obesity	3,096	10.1
Missing	5,210	17.1
Maternal smoking at 4 weeks (child age)
Not smoking	28,560	93.6
Smoking	1,603	5.3
Missing	345	1.1

Percentage distribution, n = 30,508.

Table 2.

Child Obesity at 4 Years of Age by Independent Variables

	Frequency	Percent obese
Any breastfeeding
Never breastfed	21	3.6
At least to
1 week	69	3.1
2 months	78	2.7
4 months	56	1.7
6 months	170	1.6
9 months	73	1.3
12 months	104	1.9
Child sex
Boy	204	1.3
Girl	367	2.5
Mother's first child
No	331	2.0
Yes	230	1.8
Educational level, mother
Low	85	3.0
Medium	341	2.1
High	141	1.3
Maternal age at birth (years)
≤18	4	1.7
19–24	76	2.0
25–29	159	1.7
30–34	191	1.8
35–39	109	2.2
≥40	23	2.3
Body mass index, mother
Underweight	3	0.5
Normal weight	162	1.1
Overweight	149	2.4
Obesity	162	5.2
Maternal smoking at 4 weeks (child age)
Not smoking	498	1.7
Smoking	67	4.2

According to the unadjusted analyses, obesity rates at age 4 decreased gradually with the length of breastfeeding up to 9 months, however, not significantly for breastfeeding for 1 week and 2 months. Girls were at significantly higher risk of obesity at 4 years than boys. Mother's first child and maternal age at birth were not significantly correlated with child obesity at 4 years (except for age 35–39 years). Maternal education above low level was significantly associated with lower rates of child obesity. Maternal overweight/obesity and maternal smoking were highly correlated with child obesity. See Table 3, Model 1.

Table 3.

Logistic Regression Models of Potential Determinants for Child Obesity at 4 Years of Age

	Model 1 ^a			Model 2 ^b
		95% CI			95% CI
	OR	Lower	Upper	OR	Lower	Upper
Any breastfeeding
Never breastfed	1.00			1.00
At least to
1 week	0.78	0.475	1.275	0.74	0.441	1.232
2 months	0.68	0.415	1.104	0.73	0.435	1.208
4 months	0.42	0.257	0.642	0.51	0.302	0.870
6 months	0.41	0.204	0.542	0.55	0.341	0.900
9 months	0.33	0.291	0.751	0.47	0.278	0.786
12 months	0.47	0.291	0.751	0.64	0.384	1.049
Child sex
Boy	1.00			1.00
Girl	1.91	1.607	2.267	1.91	1.600	2.277
Mother's first child
No	1.00			1.00
Yes	0.91	0.768	1.073	1.09	0.915	1.308
Educational level, mother
Low	1.00			1.00
Medium	0.68	0.531	0.866	0.76	0.576	0.991
High	0.43	0.326	0.568	0.56	0.404	0.769
Maternal age at birth (years)
≤18	0.97	0.358	2.651	0.64	0.221	1.828
19–24	1.16	0.886	1.524	0.92	0.685	1.228
25–29	0.98	0.792	1.207	0.96	0.769	1.192
30–34	1.00			1.00
35–39	1.28	1.006	1.621	1.21	0.949	1.555
≥40	1.31	0.845	2.028	1.13	0.724	1.758
Body mass index, mother
Underweight	0.43	0.011	0.014	0.42	0.134	1.286
Normal weight	1.00			1.00
Overweight	1.92	1.566	2.356	1.80	1.452	2.224
Obesity	4.32	3.522	5.292	3.78	3.029	4.704
Maternal smoking at 4 weeks (child age)
Not smoking	1.00			1.00
Smoking	2.43	1.872	3.169	1.68	1.252	2.264

ORs and 95% CIs.

Independent variables tested individually on the outcome (obesity at 4 years of age).

Breastfeeding adjusted for all independent variables.

CI, confidence interval; ORs, odds ratios.

Logistic regression analyses adjusted for all independent variables essentially confirmed the previous significant associations between breastfeeding and child obesity for breastfeeding length 4 months (OR 0.51, 95% CIs 0.302–0.870), 6 months (OR 0.55, 95% CIs 0.341–0.900), and 9 months (OR 0.47, 95% CIs 0.278–0.786). However, the OR for 6 months was slightly higher than the OR for 4 months; Table 3, Model 2. Child sex, educational level, maternal overweight/obesity, and maternal smoking were independently associated with child obesity regardless of breastfeeding, although the ORs for maternal BMI and smoking weakened when adjusted for other independent variables. Mother's first child and maternal age at birth were not significantly associated with obesity at 4 years. See Table 3, Model 2.

Interaction analyses resulted in a few significant associations, most of which, however, did not seem theoretically meaningful. Thus there were significant interactions for maternal obesity at 1 week of breastfeeding (p = 0.047), maternal overweight at 4 months of breastfeeding (p = 0.022), and maternal age ≤18 years at 12 months of breastfeeding (p = 0.019). There was also a significant interaction (p = 0.019) between child sex at 1 week of breastfeeding. Boys breastfed for 1 week (up to <2 months) were at higher risk of obesity at 4 years than girls breastfed for 1 week.

Discussion

Main findings

Our study found a mainly decreasing rate of obesity for children at 4 years of age in relation to breastfeeding duration from 1 week to 9 months. The association was significant for 4, 6, and 9 months of breastfeeding. However, the OR for 6 months was slightly higher (0.55) than the OR for 4 months (0.51). Furthermore, female sex of the child, low maternal education, maternal overweight/obesity, and maternal smoking were significantly associated with obesity at 4 years regardless of breastfeeding. These associations were particularly strong for child sex and maternal education. There was a significant interaction effect for child sex at 1 week of breastfeeding (up to <2 months).

Limitations

The attrition in the study variables was generally very low. However, the study population was restricted because of a comparatively large missing rate for growth data, 13.7%. The group lacking growth data comprised significantly more smoking, more low-educated, and more obese mothers. All of these variables were associated with obesity in children at 4 years of age. If we had had access to growth data for all these missing cases, the link between breastfeeding and obesity would probably have been strengthened rather than weakened. A similar interpretation can be made concerning the loss of data on maternal BMI, 17.1%.

Information in the databases of statistics in Uppsala and Orebro was reported by nurses at the CHCs. To avoid problems of validity with different breastfeeding categories, we chose to use any breastfeeding rather than breastfeeding of varying intensity as our independent variable. The validity of this kind of data has been analyzed in a special study from Uppsala County and was found adequate for research purposes.²⁷

The study population included adoptees whose breastfeeding patterns are likely to deviate from those of biological children. However, the adopted children constituted a very low proportion of all children in the study, 0.1% (n = 38). Thirty-seven of the adopted children were not breastfed, and one child was breastfed for 1 week. This is not expected to have affected the outcomes.

As data in this study were collected over six calendar years, many of the children had siblings in the material. For this reason, we applied an exchangeable correlation structure in the GEEs analysis to control for the dependence of infants on the same mother. We assumed that there was a correlation between children of the same mother and that this correlation was equal among all the children, that is, the same between children 1 and 2 and between children 1 and 3. However, as this is only an assumption, although reasonable, it cannot necessarily be assumed to apply in all cases. The ability to estimate the possible impact of a miscalculation on the results, however, is very limited.

Despite different attempts to adjust for potential confounders, many factors are difficult to control, for example, heredity. Others may vary with time, such as life styles or societal changes. Food policies of preschools, for instance, are subject to change outside of parental control. Differences may be because of behavioral patterns associated with feeding rather than with breast milk as such,²⁸ or with the fact that bottle-fed children may not regulate their intake in the same way as do breastfed children.²⁹ According to many studies, mother's educational level is the most important single factor in overweight/obesity prevention. Education, however, is a proxy for many factors, such as food and eating habits, physical exercise, and control of children's screen time.

A methodological fallacy lies in the fact that the composition of infant formulae has changed in later years. It is well known that formulae contain less protein today than 10 years ago. This is of course favorable for children but has to be taken into account in longitudinal studies, for instance, when evaluating studies with 10–15 years of follow-up. Changes in formula composition may or may not have affected the results of this study. For future research, studies of feeding behavior controlling for human milk versus formula and mode of administration will be of interest. More longitudinal studies will be needed.

The results of this study were based on a population collected from 2 of Sweden's 21 counties, Uppsala and Orebro. According to available information from Statistics Sweden,³⁰ there is only a marginal difference between an average of Uppsala/Orebro and the rest of the country in terms of educational background, average income, age distribution, and proportion of foreign-born children. It is, therefore, likely that the findings of this study could be generalized to the country as a whole.

Interpretation of findings

This study was fortunate to have access to a large prospective population, which we believe to have strengthened the validity of our findings. Our most important finding was a significant association between duration of breastfeeding (4, 6, and 9 months) and lower rates of child obesity at 4 years. For those with 12 months of breastfeeding, the rate of obesity at 4 years was somewhat higher than for those with 9 months of breastfeeding. We can only speculate about the reasons for this finding. Perhaps the advantages of breastfeeding had begun to wane at about 1 year of age, a possible subject for future research. We feel justified to draw the tentative conclusion that breastfeeding of a certain duration might have a protective effect against obesity, at least during early childhood.

The significant interaction for child sex at 1 week of breastfeeding is not easy to interpret. Possibly certain differences may be associated with child sensitivity and sex when it comes to early discontinuation of breastfeeding (before 2 months' duration, as in our present case).

Conclusion

Our study indicates that breastfeeding for at least 4 months may contribute independently to a reduced risk of obesity at 4 years of age. As child obesity was significantly associated with factors such as maternal education, overweight/obesity, and smoking, we recommend that the preventive child health services maintain and strengthen the overall breastfeeding supportive measures, including smoking prevention and extended support for low-educated mothers.

Footnotes

Acknowledgment

The Swedish National Board of Health and Welfare generously contributed financially to the creation of the database.

Disclosure Statement

No competing financial interests exist.

References

Horta

, Loret de Mola

, Victora

. Long-term consequences of breastfeeding on cholesterol, obesity, systolic blood pressure and type 2 diabetes: A systematic review and meta-analysis. Acta Paediatr Suppl, 2015; 104:30–37.

Grube

, von der Lippe

, Schlaud

, et al. Does breastfeeding help to reduce the risk of childhood overweight and obesity? A propensity score analysis of data from the KiGGS study. PLoS One 2015;10:e0122534. Erratum in: PLoS One, 2015; 10:e0126675.

Almquist-Tangen

, Dahlgren

, Roswall

, et al. Milk cereal drink increases BMI risk at 12 and 18 months, but formula does not. Acta Paediatr, 2013; 102:1174–1179.

Smithers

, Kramer

, Lynch

. Effects of breastfeeding on obesity and intelligence: Causal insights from different study designs. JAMA Pediatr, 2015; 169:707–708.

Grummer-Strawn

, Rollins

. Summarising the health effects of breastfeeding. Acta Paediatr Suppl, 2015; 104:1–2.

Reyes

, Hoyos

, Martinez

, et al. Satiety responsiveness and eating behavior among Chilean adolescents and the role of breastfeeding. Int J Obes (Lond), 2014; 38:552–557.

Michaelsen

, Larnkjær

, Mølgaard

. Amount and quality of dietary proteins during the first two years of life in relation to NCD risk in adulthood. Nutr Metab Cardiovasc Dis, 2012; 22:781–786.

Marchi

, Berg

, Dencker

, et al. Risks associated with obesity in pregnancy, for the mother and baby: A systematic review of reviews. Obes Rev, 2015; 16:621–638.

Emmett

, Jones

. Diet and growth in infancy: Relationship to socioeconomic background and to health and development in the Avon Longitudinal Study of Parents and Children. Nutr Rev, 2014; 72:483–506.

10.

Geoffroy

, Power

, Touchette

, et al. Childcare and overweight or obesity over 10 years of follow-up. J Pediatr, 2013; 162:753–758.e1.

11.

Durmuş

, Heppe

, Gishti

, et al. General and abdominal fat outcomes in school-age children associated with infant breastfeeding patterns. Am J Clin Nutr, 2014; 99:1351–1358.

12.

WHO. Evidence on the Long-Term Effects of Breastfeeding: Systematic Review and Meta-Analyses. Geneva: World Health Organization, 2007.

13.

Martin

, Patel

, Kramer

, et al. Effects of promoting longer-term and exclusive breastfeeding on adiposity and insulin-like growth factor-I at age 11.5 years: A randomized trial. JAMA, 2013; 309:1005–1013.

14.

WHO. Global Strategy for Infant Feeding & Young Child Feeding. Geneva: World Health Organization, 2003.

15.

Forsyth

. Policy and pragmatism in breast feeding. Arch Dis Child, 2011; 96:909–910.

16.

Fewtrell

, Wilson

, Booth

, et al. Six months of exclusive breast feeding: How good is the evidence?. BMJ, 2010; 342:c5955.

17.

ESPGHAN Committee on Nutrition, Agostoni

, Braegger

, Decsi

, et al. Breast-feeding: A commentary by the ESPGHAN Committee on Nutrition. J Pediatr Gastroenterol Nutr, 2009; 49:112–125.

18.

Turner

, Layte

. Obesity levels in a national cohort of women 9 months after delivery. Am J Obstet Gynecol, 2013; 209:124.e1–e7.

19.

Wallby

, Hjern

. Region of birth, income and breastfeeding in a Swedish county. Acta Paediatr, 2009; 98:1799–1804.

20.

Chezem

. Breastfeeding attitudes among couples planning exclusive breastfeeding or mixed feeding. Breastfeed Med, 2012; 7:155–162.

21.

Inera

. Tables NEXT. Available at www.inera.se/TJÄNSTER&PROJEKT/Tillvaxtinformationförbarnochungdomar/Dokument/ (In Swedish). (Accessed December 12, 2016).

22.

Statistics Sweden. Longitudinal integration database for health insurance and labour market studies. Statistics Sweden. Available at www.scb.se/lisa (accessed July 12, 2016).

23.

Swedish National Board of Health and Welfare. Evaluation of the Swedish Medical Birth Registry. In: Forskningsrapport från

EpC

, Källén

, ed. Stockholm: Socialstyrelsen, Epidemiologiskt Centrum, 2002. (In Swedish.)

24.

Cole

, Bellizzi

, Flegal

, et al. Establishing a standard definition for child overweight and obesity worldwide: International survey. BMJ, 2000; 320:1240–1243.

25.

Cole

, Flegal

, Nicholls

, et al. Body mass index cut offs to define thinness in children and adolescents: International survey. BMJ, 2007; 335:194.

26.

IBM SPSS Statistics for Macintosh, Version 22.0. Armonk, NY: IBM. Corp., 2013.

27.

Wallby

, Lagerberg

, Magnusson

, et al. Basta—Validity in a Regional Child Health Care Register. Uppsala: Barnhälsovården i Uppsala län, 2011. (In Swedish.)

28.

DiSantis

, Collins

, Fisher

, et al. Do infants fed directly from the breast have improved appetite regulation and slower growth during early childhood compared with infants fed from a bottle?. Int J Behav Nutr Phys Act, 2011; 8:89.

29.

, Magadia

, Fein

, et al. Risk of bottle-feeding for rapid weight gain during the first year of life. Arch Pediatr Adolesc Med, 2012; 166:431–436.

30.

Statistics Sweden. Available at www.scb.se/sv_/Hitta-statistik (In Swedish.) (Accessed December 10, 2016).