The impact of breastfeeding on FTO-related BMI growth trajectories: an application to the Raine pregnancy cohort study

Abstract

Introduction For years, body mass index (BMI) has been used by scientists to track weight problems and obesity in children and adults. Recent studies have implicated the fat mass and obesity gene (FTO) in the increase of BMI in young adults. A longer duration of breastfeeding is known to reduce the risk of being overweight later in life, but its ability to modify the effect because of FTO is not known.

Methods We studied 1096 children from the Western Australian Pregnancy (Raine) cohort who were followed up from birth to 14 years of age. Linear mixed-effects models were used to investigate BMI growth trajectories in boys and girls separately.

Results An association was found between BMI growth and the duration of exclusive breastfeeding (EXBF) among carriers of the risk allele of the FTO SNP rs9939609. In girls, EXBF interacts with the SNP at baseline and can reverse the increase in BMI because of SNP risk allele by age 14 years after 3 months of EXBF. In boys, EXBF reduces BMI both in carriers and non-carriers of the risk allele with an association found after 10 years of age. Six months of EXBF will put the boys’ BMI growth curves back to the normal range.

Conclusions Our study could have major health implications by providing new perspectives for the prevention of growth problems in children carrying risk alleles in the FTO gene.

BMI, obesity, longitudinal study, growth, FTO gene, breastfeeding, children, RAINE, Australia, linear mixed-effects model

Introduction

Body mass index (BMI) has been used by health professionals and scientists to track weight problems and obesity in adults for many years. Recent studies have supported the role of the (FTO) gene in increasing BMI in young adults.^1–5 Genome-wide association studies on BMI and adiposity have successfully identified an association with genetic variants in the FTO locus in both adults and children.^2,3,6–8 In meta-analyses, the addition of each minor (A) allele at the SNP rs9939609 within the first intron of FTO has been shown to be consistently associated with a higher BMI of up to 0.33 kg/m² or ∼0.1 standard deviation (SD).^2,9 The biological mechanisms underlying this association are yet to be determined; however, evidence from population-based and functional studies suggest that this locus is likely involved in the hypothalamic regulation of appetite or energy expenditure and metabolic rate.^4,10–15

What those studies have not yet reported is whether the long-term increase in BMI attributed to FTO risk alleles can be mitigated by environmental factors. Breastfeeding, particularly the duration of breastfeeding, is a good candidate as risk modifier. Some studies have shown that a longer duration of breastfeeding was associated with a decrease of the risk of being overweight later in life.^16–19 A recent study from the Raine cohort in Australia²⁰ showed that early infant feeding can affect BMI trajectories through a possible change in the timing of adiposity rebound (AR). These studies have also pointed out that the definition of breastfeeding can affect the association results, in particular the distinction between exclusive and non-exclusive breastfeeding (BF). It is still unknown in longitudinal studies whether a long duration of exclusive breastfeeding or mixed feeding can prevent the FTO-induced increase in BMI. Dedoussis et al.²¹ showed that a short period of at least 1 month of breastfeeding was associated with reduced obesity in a childhood cohort from Greece; however, the study was cross-sectional at different time points, and there was no distinction between exclusive and non-BF. Also, the results were not replicated in the Avon Longitudinal Study of Parents and Children (ALSPAC) child cohort from the UK,²² even when a longer period of breastfeeding was considered. In our study, we investigate the long-term impact of the duration of breastfeeding, more specifically exclusive breastfeeding (EXBF), on FTO-induced BMI growth curves in children from the Western Australian Pregnancy (Raine) cohort.

Materials and methods

The Western Australian Pregnancy Cohort (Raine)

Recruitment of the Western Australian pregnancy cohort (Raine) has been previously described in detail.^23,24 In brief, between 1989 and 1991, 2900 pregnant women were recruited before 18 weeks of gestation into a randomized controlled trial to evaluate the effects of repeated ultrasound in pregnancy. Recruitment predominantly took place at King Edward Memorial Hospital (Perth, Western Australia). Women were randomized to repeat ultrasound measurements at 18, 24, 28, 34 and 38 weeks of gestation or to a regular ultrasound assessment at 18 weeks. Children have been comprehensively phenotyped from birth to 18 years of age by trained members of the Raine research team. Data up to the age of 14 years only were available at the time of our study, and these data were used in our analyses. DNA was collected at age 14 years in addition to data on dietary intake and duration of exercise undertaken outside school hours. Month and year of first menstrual period for girls was recorded. Weight was measured using a Wedderburn Digital Chair Scale to the nearest 100 g, with each child dressed in running shorts and a singlet top; height was measured to the nearest 0.1 cm with a Holtain Stadiometer and BMI was calculated as weight (kg)/height (m²). A description of the main characteristics of the study population is given in Table 1. The cohort was representative of the population presenting at an antenatal tertiary referral centre in Western Australia.²⁵ The study was conducted with institutional ethics approval, and a signed informed consent letter was obtained from all the mothers.

Table 1

Cohort characteristics, Perth, Western Australia, 1989–2005

	Boys (n = 417)	Girls n = 389)
Variables	Mean (SD)	Mean (SD)
EXBF (months)a	3.07 (1.87)	3.09 (1.98)
BF (months)b	7.69 (6.57)	7.67 (6.58)
Gestational age (weeks)	39.31 (2.06)	39.30 (2.09)
Mother’s BMIc	26.17 (5.56)	26.47 (6.15)
Mother’s education	n	n
High-school/professional degree = 0	92	88
College/University = 1	325	301
SNP rs9939609 genotypes
TT	149	153
TA	208	185
AA	60	51
Number of BMI measurements by Age (years)
0	374	349
1	404	377
2	124	133
3	288	279
6	392	368
7	13	10
8	351	316
9	58	63
10	121	112
11	270	246
14	411	377
Average number of measurements per child	6.76	6.8

^aEXBF = duration of exclusive breastfeeding.

^bBF = duration of breastfeeding (non-exclusive).

^cMother’s pre-pregnancy BMI.

Definition of breastfeeding-related variables

Information pertaining to early infant feeding was collected at age 1, 2 and 3 years. Mothers recorded the age at which breastfeeding was stopped (in months), and the age at which milk other than breast milk was introduced (in months). This information was determined from the mother’s diary of early feeding milestones, as well as from an interview with the study nurse at the early follow-ups and survey questions at later follow-ups. The duration of EXBF was defined for the subset of mothers who breastfed their baby as the time from birth until they started feeding their baby with other milk (non-breast milk) or any solid. The duration of non-BF was derived as the time from birth until breastfeeding was stopped, where breastfeeding could be mixed with any other milk or solid. We also defined two instrumental variables (IVs). An IV is defined as a random variable that is correlated with the true variable but is unobserved or is not an accurate variable, and it is uncorrelated with the model error term(s). In epidemiology, it is an observable variable that is predictive of an outcome, but it has no direct association with the disease and is independent of the unobserved confounders.²⁶ The first instrumental variable IV1 was defined as the time from birth to the time when mother stopped breastfeeding or introduced other milk, whichever came first. We also defined IV2 as the time from birth to the time when mother stopped breastfeeding or introducing solids, whichever came first. Both IV1 and IV2 were highly correlated with EXBF, r = 0.79 and r = 0.89, respectively.

Definition of other variables

Several known confounding variables were added to our longitudinal models including mother’s education (college/university = 1 vs high-school/professional degree = 0), gestational age (in months) and mother’s pre-pregnancy BMI.

Statistical methods

For a typical child, the BMI curve increases from birth until it reaches a peak ∼9 months [adiposity peak (AP)], such that 1-year-old children seem chubby. Then the curve decreases until it reaches a minimum ∼5.5/6 years [i.e the adiposity rebound (AR)]. After this, BMI typically increases again until adulthood. The proportion of fatness decreases after 1 year of age and varies across children, such that the AR can occur at age between 4 and 8 years, with an earlier rebound often associated with greater adiposity in early adulthood.²⁰ To our knowledge, there is not a well-established parametric model that describes BMI growth throughout infancy and childhood. After careful investigation of the Raine data (Figures 1 and 2), it was decided to choose a break-point mixed-effects model²⁷ for the following reasons:

• BMI shows stages of ‘growth’ and ‘decline’ over this period and a point separating these stages (‘break-point’);

• BMI differs between individuals at the origin (intercept); therefore, we need random intercept to explain this behaviour;

• Children exhibit variable rates of change in BMI with age; therefore, we use a random slope to capture this feature.

Figure 1

Population-based estimate of BMI vs age by SNP genotypes in children from birth to age 14 years in the Raine study

Figure 2

BMI vs age (for a sample of 12 individuals) in children from birth to age 14 years in the Raine study

This break-point model emphasizes that there are two distinct time windows, infancy (between birth and age 1.5 years) and childhood until puberty (aged between 1.5 and 14 years) but does not require the data to be split. The break-point was chosen such that it occurs between the times at AP and AR and allows a more precise estimate of these two points (smaller SD). To ensure our results were not dependent on the placement of the break-point, we tested different break-point models with the point aged between 1 and 2 years; our results remained consistent. A random effect was not added to the model for the break-point itself, as it is challenging to estimate. A co-dominant genetic model was used to allow two genotype-associated parameters (TA vs TT and AA vs TT) to be estimated for the SNP of interest. As the BMI growth pattern and the timing of pubertal growth differ between boys and girls, we fit a separate model for each gender. We also used models with no breastfeeding-related variables and other models adjusting for EXBF or BF.

Based on Scott et al.’s previous work²⁷ on break-point linear mixed-effects for modelling changes in lung function in Duchenne’s muscular dystrophy over time, we decided to use the same modelling framework for our longitudinal BMI. If we fix the break-point at age 1.5 years for all subjects, the model can be written as follows:  

(1)

where Y is BMI at a given age, represents time-independent covariates, including the two FTO SNP genotypes, EXBF (BF), mother’s education (MomEdu), gestational age (GA), mother’s pre-pregnancy BMI (MomBMI), the interaction between the SNP and EXBF (or BF), and represents time-dependent covariates, including the interaction between age and BF (or EXBF), SNP genotypes and mother’s education. The coefficients and represent the mean rate of change of BMI during infancy and childhood. The coefficients and represent the mean acceleration/deceleration of BMI during infancy and childhood.

Physiologically, BMI is continuous across all ages. Thus models with enforced continuity at the break-point may fit the physiological process better. We modified model (1) to enforce continuity between the two time windows²⁷:  

(2)

where I is an indicator function. We have written model (2) in short form as  

(3)

In this model, the vector of coefficients measures the variation in BMI associated with a change in one unit of the time-independent covariates at baseline (age 1.5 years), and the vector of coefficients measures the variation in BMI associated with a change in one unit of the time-dependent covariates where time corresponds to a particular child’s age.

Given BMI growth trajectories are not identical across individuals, we can represent the model (3) in linear mixed-effects model framework:  

(4)

where is the response vector for the ith individual, is the vector of fixed effects and is the vector of random effects, and are the fixed effect and random effect regressor matrices, respectively, and is the within subject error vector. The linear mixed-effects model was applied to our data with random intercept and random slope for age. The time dependency was accounted for by a continuous auto-regressive structure, which best fits our data.

The relation between EXBF and BMI growth curves at particular ages was estimated and tested using the general linear hypothesis approach²⁸ (see Supplementary Appendix 1, available as Supplementary data at IJE online). The times to AR and AP were obtained by maximizing the predicted BMI curves in the time intervals around the AR and AP using the Newton–Raphson maximization method. All the models were fitted with the function lme included in the R library nlme version 3.1-102.

Results

Table 1 presents the characteristics of the 1096 children in our analyses. Their distribution according to the SNP genotypes showed that 560 (44.9%) were TT, 523 (42.0%) were AT and 163 (13.1%) were AA. The allele A is considered as the risk allele associated with increased BMI.¹ The mean duration of EXBF was 3.1 months (SD = 1.9 months), and it did not vary with a child’s genotype. After removing children with missing covariate information at all time points, we had 959 children left for our analyses, 498 boys and 461 girls.

Cross-sectional analysis at age 14 years

Figure 3 represents the box-plot distribution of the sample-based estimate of BMI obtained from model (1) at age 14 years with respect to the SNP genotypes and various durations of EXBF in boys and girls. The dashed horizontal line delineates the BMI categories ‘normal weight’ and ‘overweight’ as defined by the Centers for Disease Control and Prevention. This definition is based on expert committees’ recommendations to classify BMI-for-age between the 85th and 95th percentile as at risk of overweight.²⁹ It is clear from the left panel that EXBF impacts on all genotype categories in boys with a substantial decrease in the median BMI as the duration of EXBF increases. For example, with <2 months of EXBF, the median BMI among boys who carry either the AT or AA genotype is in the overweight category. With 2 to 4 months of EXBF, <20% of boys are in the overweight category and no boys are overweight if they had ≥5 months EXBF. In girls (right panel), the parameter estimate for the duration of EXBF is smaller, but we can still notice a decrease in the median BMI among the AA genotype carriers as the duration of EXBF increases. This decrease does not seem to be important as the confidence intervals overlap.

Figure 3

Percentage of individuals in different categories of BMI and genotypes at age 14 years in the Raine study

Longitudinal analysis of the SNP (rs9939609) genotypes associated with BMI growth curves

The estimated BMI growth curves obtained from model (1) from birth to age 14 years by SNP genotype and gender are displayed in Figure 1. At baseline (i.e. aged 1.5 years), the parameter estimate for the AA genotype was associated with a higher BMI both in boys and girls but not the AT genotype (Table 2). Boys having the AA genotype have a higher BMI of 0.020 kg/m² [95% confidence interval (CI) 0.012–0.028; ], whereas girls have a higher BMI of 0.028 kg/m² (95% CI 0.020–0.036; ⁠). After 1.5 years of age, there is an interaction between the SNP genotypes (AT and AA) and age in boys only. Boys have an average linear increase of their BMI of 0.127 kg/m²/year (95% CI 0.006–0.191; P = 0.00001) if they carry the AT genotype and 0.138 kg/m²/year (95% CI 0.048–0.228; P = 0.0022) if they carry the AA genotype. There was no interaction with age for the girls.

Table 2

Estimates of SNP (rs9939609) and SNP by age interaction fixed effects, standard errors and P-values for a model with no breastfeeding variables, Perth, Western Australia, 1989–2005

	Boys			Girls
Variables	Estimatea	SE	P-value	Estimatea	SE	P-value
SNP genotype TAb	0.14	0.13	0.29	0.042	0.12	0.73
SNP genotype AA	0.020	0.004	0.0001	0.028	0.004	0.0001
Age, 1.5 years:TA	0.127	0.032	0.0001	−0.009	0.033	0.78
Age, 1.5 years:AA	0.138	0.045	0.0022	−0.026	0.049	0.61

^aEstimates were adjusted for age, quadratic age before and after the break-point, mother’s education, gestational age, mother’s BMI and the interaction age by mother’s education. Age was centred at 1.5 years.

^bThe baseline genotype category is TT, i.e the low-risk genotype.

Impact of the duration of breastfeeding on FTO-induced BMI growth curves

Table 3 gives the estimates of the fixed effects parameters from the linear mixed-effects model [model (1)], their standard errors and associated P-values when EXBF was included in the model. There is a complex pattern of association between BMI growth curves, FTO SNP genotypes and EXBF, which differs between girls and boys (Figure 4). In girls only, an interaction between EXBF and the SNP genotypes was detected, resulting in a substantial decrease in BMI of 0.119 kg/m² (95% CI 0.001–0.237; P = 0.043) for each month of breastfeeding in the AT carriers and 0.180 kg/m² (95% CI 0.002–0.358; P = 0.045) in the AA carriers. The association of the SNP genotypes and EXBF on BMI depends largely also on age. We used the general linear hypothesis tests²⁶ to compare the effect of no breastfeeding vs 6 months of EXBF on BMI at different ages by gender and genotype (Table 4). This table shows clearly that in boys, EXBF affects all SNP genotype categories after 10 years of age and leads to a decrease in BMI varying from 1 to 1.76 kg/m². In contrast, in girls, 6 months of EXBF reduces BMI only in the AA genotype category. The genotype-specific BMI growth curves estimated from the model fixed effects are depicted in Figure 4 for 0, 3 and 6 months of EXBF. In girls, the association with the FTO variant is reversed after 3 months of EXBF. In boys, 6 months of EXBF will put the BMI growth curves in the normal range. Carriers of the AT and AA genotypes breastfed exclusively for 6 months have a similar BMI growth as carriers of the TT genotype who have never been breastfed.

Figure 4

Fitted BMI vs age for different EXBF categories in children from birth to age 14 years in the Raine study

Table 3

Estimates of fixed effects parameters for the SNP (rs9939609) and EXBF main effects and the SNP by age and SNP by EXBF interactions, standard errors and P-values, Perth, Western Australia, 1989–2005

	Boys			Girls
Variables	Estimatea	SE	P-value	Estimatea	SE	P-value
EXBF	−0.054	0.050	0.28	0.097	0.045	0.033
SNP genotype TAb	−0.098	0.22	0.66	0.56	0.22	0.011
SNP genotype AA	0.22	0.36	0.54	0.60	0.31	0.053
Age:EXBF	−0.014	0.008	0.084	0.005	0.008	0.55
Age:TA	0.13	0.034	0.0001	0.010	0.035	0.78
Age:AA	0.13	0.049	0.0088	−0.006	0.052	0.90
EXBF:TA	0.011	0.063	0.86	−0.119	0.059	0.043
EXBF:AA	−0.058	0.100	0.56	−0.180	0.089	0.045

^aEstimates were adjusted for age, quadratic age before and after the break-point, mother’s education, gestational age, mother’s BMI and the interaction age by mother’s education. Age was centred at 1.5 years.

^bThe baseline genotype category is TT, i.e the low risk genotype.

Table 4

Linear estimates for 6 vs 0 months of EXBF at age 1.5, 5, 10 and 14 years by gender and SNP genotype, Perth, Western Australia, 1989–2005

		Boys			Girls
FTO genotypes	Age	Estimate	SE	P-valuea	Estimate	SE	P-valuea
TT	1.5	−0.32	0.30	0.28	0.56	0.27	0.033
	5	−0.63	0.35	0.073	0.69	0.33	0.038
	10	−1.06	0.52	0.043	0.84	0.52	0.11
	14	−1.41	0.70	0.044	0.96	0.70	0.17
TA	1.5	−0.26	0.23	0.26	−0.14	0.23	0.56
	5	−0.56	0.29	0.053	−0.030	0.30	0.92
	10	−1.00	0.49	0.041	0.12	0.50	0.81
	14	−1.35	0.67	0.045	0.24	0.68	0.72
AA	1.5	−0.67	0.52	0.20	−0.50	0.47	0.29
	5	−0.98	0.55	0.076	−0.39	0.50	0.43
	10	−1.41	0.67	0.037	−0.24	0.64	0.70
	14	−1.76	0.82	0.032	−0.12	0.79	0.88

^aSignificant P-values are indicated in bold.

Assessing breastfeeding-related variables

To investigate whether our results are sensitive to the definition of breastfeeding, we refit the models using BF and the two instrumental variables (IV1 and IV2) instead of EXBF. No interaction was detected, in boys or girls, when looking at the duration of breastfeeding regardless of other milk/solids (BF). Neither IV had an interaction with FTO gene variant. If we consider that EXBF is the true variable and BF, IV1 and IV2 are variables measuring breastfeeding but with measurement error, then our analysis emphasizes the need for accurate information on the variables of interest for detecting important associations and gene environment interactions in longitudinal studies.

Assessing distribution assumptions

The distribution of the model residuals shows a slight departure from the normal distribution. Therefore, we also considered alternative models allowing the distribution of the random effects in model (1) to be either skew-normal or skew-t.³⁰ Based on our diagnostic plots, the choice of the normal mixed-effects model was justified. We have also investigated log and sqrt transformations of the response variable, but the qq plots of the model residuals show that the best model fit was obtained when using the raw BMI values.

Discussion

The aim of our study was to investigate the impact of EXBF on FTO-induced BMI growth trajectories from birth to age 14 years. The rationale was that although the FTO gene variant rs9939609 is associated with an increased BMI in children and adults, a longer period of EXBF could reverse the genetic association.

First, our study confirms the role of the FTO gene variant rs9939609 on BMI growth curves both at baseline and over time, but it also suggests an important gender- and age-specific relationship. At baseline (i.e. age 1.5 years), the AA genotype was associated with a higher BMI in both boys and girls than the TT genotype, but no association was detected with the AT genotype. After 1.5 years of age, a linear increase in BMI was only detected in boys among the AT and AA genotype carriers. A recent meta-analysis¹ also suggested an association with FTO gene variant on BMI growth trajectories in children, but it concluded that rs9939609 was associated with a decrease in BMI at the AP and an increase in BMI at the AR. These two studies stress the complex time dependence of the FTO gene variant on BMI growth; however, Sovio et al.¹ did not test the possibility of gender-specific associations.

The beneficial role of breastfeeding on many diseases (obesity, type 2 diabetes, cholesterol and insulin resistance) has been suggested in many publications,^31–34 and there is a general consensus that breastfeeding should be recommended to all women.³³ The protective association found with breast milk could come from its low-protein and high-fat content, which could lead to slower child growth and better nutritional balance between protein and fat during infancy.³⁵ A study has also shown that infants who are bottle-fed in early infancy are more likely to empty the bottle or cup in late infancy than those who are breastfed. A possible reason for this is that parents may encourage an infant to finish the contents of the bottle, whereas when breastfed, an infant naturally develops self-regulation of milk intake.³⁶ Breastfeeding could, therefore, protect against later obesity by reducing the occurrence of high weight gain in infancy.³⁷ This is also consistent with previous findings from the Raine study where an increased duration of breastfeeding was found associated with reduced BMI in adolescence³⁸ and increased weight gain within the first year of life was associated with increased BMI in adolescence. Data from rat models³⁹ also suggest that lactation can mitigate some of the adverse effects of placental insufficiency on the later development of metabolic disease. Another possible role of breastfeeding is that breastfed and formula-fed infants have different hormonal responses to feeding, with formula feeding leading to a greater insulin response resulting in fat deposition and increased number of adipocytes.⁴⁰ Finally, limited evidence suggests that breastfed infants adapt more readily to new foodstuffs, such as vegetables, thus reducing the caloric density of their subsequent diets.⁴¹

The additive genetic model has been widely used in genetic studies of BMI related to FTO gene variants. Most of these studies were cross-sectional in nature and considered BMI in adults.⁴² In longitudinal studies of children's BMI, the genetic model for FTO gene variants has been less studied. The meta-analysis in Sovio et al.¹ showed that although an additive model for the SNP rs99339609 could fit longitudinal BMI profiles in children well in different periods, a more general genetic model was a more appropriate fit for alternative adiposity phenotypes, such as age at AP and AR. Besides, there was also a lot of variability across studies in terms of the best fitting genetic model. In our study, the estimation of FTO-related BMI longitudinal profiles showed that boys with AA or AT genotypes have similar longitudinal BMI profiles but differ greatly from boys with the TT genotype. On the other hand, girls who have the AT genotype have different longitudinal BMI profiles than do girls with a TT or AA genotype. These particular patterns would not have been detected under an additive model. Our study of the SNP by EXBF interaction and estimation of ages at AR and AP also show complex associations with the SNP genotypes, which could not be captured by an additive model. We, therefore, opted for the most general genetic model for the FTO SNP, i.e. the co-dominant model.

It has been previously noted that sex and age are associated with differences in obesity-related traits and body composition. For instance, women tend to store more fat subcutaneously rather than in visceral adipose tissue; hence, at the same BMI, women will tend to carry more body fat than men.⁴³ This finding is of particular interest, as it supports previous hypotheses that there are sex-specific genes contributing to variation of obesity-related traits and that genes account for more of the variability of fat distribution in women than in men.^44,45 Our study confirms the role of age- and gender-specific genetic effects and also adds to this observation that gene by environment interactions could also be gender specific. The biological mechanisms underlying this observation require further investigation.

Despite the importance of breastfeeding in child growth development and prevention of metabolic diseases, its impact on carriers of FTO gene risk alleles has barely been studied. To our knowledge, it has only been investigated by Dedoussis et al.²¹; however, their study was cross-sectional rather than longitudinal, and there was no distinction between EXBF and BF. Their study concluded that a short period of at least one month of breastfeeding was associated with reduced obesity indices in a cohort of Greek children, but this result was not replicated in the ALSPAC child cohort from the UK, even when a longer period of breastfeeding was considered.²¹ The major implication of our results is to support the possible role of EXBF in preventing the increase of BMI because of FTO. To better understand the role of FTO gene variant and EXBF on BMI growth, we also studied their association with the timing of AR. Indeed, it is hypothesized that an earlier age at AR is strongly associated with the risk of obesity later in life.¹ We found that the FTO gene variant was associated with an earlier AR, but only in boys. The mean age at AR was 5.16, 4.11 and 4.17 years in boys for the TT, AT and AA genotypes and was almost constant (∼4.60/4.70 years) in girls. EXBF had no impact on girls’ AR; however, it delays the timing of AR in boys by ∼4 months and 8 months after 3 and 6 months of EXBF, respectively, regardless of their genotype. The association between EXBF and BMI growth found in boy carriers of the SNP risk allele could, therefore, be partly explained by a change in the timing of AR. In girls, this association seems to be before AR occurrence and had little impact on growth after AR. The biological mechanisms explaining the role of breastfeeding on FTO-induced BMI growth curves are not known. A possible explanation is that breastfeeding could prevent FTO-induced fast weight gain during infancy by providing better energy balance at critical periods of development during infancy, but this remains speculative. Further studies are required to determine the complex association of FTO and EXBF with BMI growth trajectories throughout childhood.

Supplementary Data

Supplementary Data are available at IJE online.

Funding

This work was supported by the ALVA foundation, Toronto and by a grant from the Canadian Institute of Health Research (MOP82893). Funding for Core Management of the Raine Study is provided by The University of Western Australia (UWA), the Raine Medical Research Foundation, the Telethon Institute for Child Health Research, UWA Faculty of Medicine, Dentistry and Health Sciences, Women and Infants Research Foundation and Curtin University.

Conflict of interest: None declared.

KEY MESSAGES

• Recent studies have implicated the fat mass and obesity gene (FTO) in the increase of BMI in young adults and children.

• The addition of each minor (A) risk allele at the SNP rs9939609 within the first intron of FTO has been shown to be consistently associated with a higher BMI of up to 0.33 kg/m² or ∼0.1 SD.

• We studied 1096 children from the Raine cohort that were followed up from birth to 14 years of age, and we analysed their longitudinal BMI growth trajectories.

• We showed that the duration of EXBF could attenuate the increase in BMI among carriers of the risk allele of the FTO SNP rs9939609.

• In girls, EXBF can reverse the increase in BMI because of SNP risk allele by age 14 years after 3 months of EXBF.

• In boys, EXBF reduces BMI in both carriers and non-carriers of the risk allele with an association found after 10 years of age. Six months of EXBF will put the boys’ BMI growth curves back to the normal range.

Acknowledgements

The authors are grateful to the Raine study participants and their families, the Raine study research staff for cohort coordination and data collection, the NH and MRC for their long-term contribution to funding the study for the past 20 years, the Telethon Institute for Child Health Research for long-term support of the study and the Fraser Mustard Institute for Human Development, University of Toronto. The authors gratefully acknowledge the assistance of the Western Australian DNA Bank (National Health and Medical Research Council of Australia National Enabling Facility).