Study population and data collection

The study population and data collection used in this observational cohort study have been described earlier [20]. In short, the study population included mothers and their children born 1999–2000 that were followed at child health care centres in a southwestern Swedish municipality with approximately 38 000 inhabitants in 1999. The municipality consisted of a small urban area surrounded by rural areas. Individual data were collected from primary care child health care centres using the unique personal identity number assigned to every Swedish resident [21]. The data obtained from the child health care centres were retrieved with start in 2000 and included sex, gestational age at birth, body weight and height at birth, body weight and height from 0–6 years and occurrence of breastfeeding of the child, which was recorded at each visit. Occurrence of breastfeeding included both exclusive and partial breastfeeding. Similar, the mothers’ body weight and height, BMI, age, parity and smoking habits were retrieved from the antenatal clinics. Data on weight and height at 7–16 years of age were retrieved between 25 March 2019 and 6 February 2020 using school health care records on routine measurements of body weight and height at approximately 7, 10, 13 and 16 years of age. If a child had multiple adjacent visits, data were chosen from the visit closest to 4 (mean age 4.1 ±  standard deviation 0.19 years), 7 (6.9 ± 0.53), 10 (10.4 ± 0.54), 13 (13.2 ± 0.76) and 16 (16.4 ± 0.58) years of age. As the BMI for children varies with age and sex, ISO-BMI was assessed with cutoffs for underweight, overweight and obesity according to Cole’s classification [22,23] at the exact age at the time of the examination.

Statistical analysis

We used descriptive statistics for the baseline data. The breastfeeding duration of the children was categorised into two groups: < 6 months and ≥ 6 months, based on the World Health Organization’s recommendation of 6 months of exclusive breastfeeding [24]. Boxplots were used to describe the BMI of the girls and boys according to age and breastfeeding duration. To preserve individual confidentiality, potentially identifying outliers were not graphically presented but were included in the analyses. Unadjusted linear regression models were used to estimate the association between breastfeeding duration and the mother’s BMI and parity. Simple autoregressive linear regression models were used to estimate the association between breastfeeding duration in months and the children’s BMI as a continuous variable. In these models, adjustment was made for BMI at the age immediately preceding ages 7, 10, 13 and 16 years. To adjust for potential confounders, multiple autoregressive linear regression models were additionally adjusted for maternal age, parity, smoking and BMI, gestational age at birth, weight at birth and the child’s sex. In addition, we used linear mixed-effects regression models with repeated measurements at the exact ages at the examinations to estimate the association between breastfeeding duration (in months) and children’s BMI as a continuous variable. To address the correlation of repeated BMI measurements within individuals, the models included a random intercept for each child. Both unadjusted and multiple adjusted models were fitted. The adjusted models also accounted for potential confounders, including maternal age, parity, smoking and BMI, gestational age at birth, weight at birth and the child’s sex. We also fitted unadjusted and multiple adjusted linear mixed-effects models, including a three-way interaction term between breastfeeding duration, sex, and age, to predict and visualize BMI with 95% confidence intervals for boys and girls across different breastfeeding durations at specific ages.

The proportion of girls and boys with underweight or normal weight compared to overweight or obesity, based on ISO-BMI at approximately 4, 7, 10, 13 and 16 years of age, were presented graphically according to breastfeeding duration < 6 months and ≥ 6 months. Binary logistic autoregressive regression models were used to estimate the odds ratios (OR) for children to be classified as overweight/obese compared to underweight/normal weight according to ISO-BMI, by duration of breastfeeding (≥6 months versus < 6 months). In these models, we adjusted for the BMI at the age immediately preceding ages 7, 10, 13 and 16 years. In addition, we used logistic mixed-effects models with repeated ISO-BMI classifications at approximately 4, 7, 10, 13 and 16 years of age to estimate the OR for children to be classified as overweight/obese compared to underweight/normal weight according to ISO-BMI, by duration of breastfeeding (≥6 months versus < 6 months). With respect to the limited number of children with overweight/obesity in each age group, no further adjustments were made in the logistic regression models.

Statistical analyses were performed for all the children and separately for the girls and boys. As some mothers had multiple births, clustered standard errors were used in the regression models. All tests were two-tailed and conducted at a 0.05 significance level. The statistical analyses were performed using Stata 18.0 (Stata Corp., College Station, TX, US).

Ethics

The study was approved by the Regional Ethics Review Board in Gothenburg (reference 935-18). The authors collecting data had access to information that could identify individual participants. Thereafter, the data were pseudonymised before analysis. Informed consent was not retrieved from the studied individuals since the results are reported from retrospective data from registers. The need for consent was waived by the Regional Ethics Review Board in Gothenburg.

Findings in relation to other studies

National breastfeeding data from 1999–2000 from the Swedish National Board of Health and Welfare show a breastfeeding rate of 83% for more than four months and 72% for more than six months [25]. This frequency is higher than in our study, in which 62% of the children were breastfed for more than six months, which could be explained by socioeconomic factors that have been shown to influence, among other factors, breastfeeding frequency [26]. Our results are in line with results from some other Scandinavian studies from the same period that found no significant association between breastfeeding and child weight. A Swedish prospective cohort study found that the duration of exclusive breastfeeding was not associated with obesity in early childhood after adjusting for confounding factors [27]. Another study based on 7‐year‐old and 11‐year‐old children born 1997–2003 enrolled in the Danish National Birth Cohort found that a longer duration of exclusive breastfeeding did not seem to protect 5-month-old infants with high weight for age from overweight in childhood [28]. In addition, the Belarusian cluster-randomised Promotion of Breastfeeding Intervention Trial (PROBIT), which allocated maternity hospitals into intervention or control arms in 1996–1997, showed no significant effect of breastfeeding on childhood obesity and even a negative association regarding overweight [29]. In contrast, a recent systematic review and meta-analysis that included 159 studies of children born from 1934 to 2017 showed that breastfeeding was associated with a significantly reduced risk of overweight and obesity [14]. Another meta-analysis showed a dose-response protective effect of breastfeeding duration on childhood obesity [30].

Even though our study did not show an association between breastfeeding duration and overweight and obesity, breastfeeding remains important for its numerous health benefits for both mother and child [11,12]. It may be important to recognize the possibility that there might be no direct or strong link between breastfeeding duration and a reduced risk of obesity in children as breastfeeding is not an option for all women due to health issues or personal reasons. Health care interventions should be grounded in a broader understanding of the many contributing factors to childhood obesity and not just one single factor, such as breastfeeding. Obesity is influenced by a multitude of factors, including genetics, diet, physical activity, socioeconomic status, and environmental factors [1].

Another possible consideration related to the null finding of our study is the composition of modern infant formula, which has become increasingly similar to breast milk in terms of nutritional content [31,32]. This similarity may reduce any potential protective effect that breastfeeding could have against obesity, as infants who are formula-fed may receive comparable levels of essential nutrients. Therefore, differences in BMI or obesity risk between breastfed and formula-fed children may be diluted, particularly when other factors influencing weight gain are accounted for.

Strengths and limitations

In Sweden, children are offered free health care at child health care centres in primary health care. About 99% of children in Sweden aged 0–6 years participate in the programme. Thereafter, school health care with specialised nurses offers regular health visits up to the 12th grade [33]. Accordingly, one strength of our study was that most of the children born 1999–2000 were included in this study and were continuously monitored in a real-world setting over a long-term period. Also, data on breastfeeding were prospectively recorded, especially frequently during the first 6 months, reducing the risk of recall bias, which could be the case in retrospectively reported data. The data on height, weight and smoking were retrieved from records from antenatal clinics and child health care centres in primary health care and from school health care data that are also used in national registers [25]. However, this study also had several limitations, including incomplete data on the addition or composition of milk formula. Further, we had no data on frequency of exclusive or partial breastfeeding which could possibly influence the results. The study also had a small sample size with risk of underpowered analyses and no data on socioeconomic status, which might be important for the frequency and duration of breastfeeding. Likewise, data on factors that can influence BMI development, such as lifestyle, psychological and physical illnesses in the children, were not available. Even though we found no association between breastfeeding duration and BMI when using continuous data, the dichotomization of breastfeeding duration into < 6 months and ≥ 6 months poses a limitation. Specifically, the < 6 months category is heterogeneous, including mothers who breastfed for only a short period alongside those who breastfed for nearly six months, which may obscure nuanced effects on BMI. In addition, although the study is based on a real setting in child health care centres and many other countries have similar conditions with regular health check-ups for children, the frequency and duration of breastfeeding may differ among populations and our study might not be relevant in countries with different breastfeeding patterns.