Abstract
Importance
Evidence that breastfeeding reduces child obesity risk and lowers blood pressure (BP) is based on potentially confounded observational studies.
Objective
To investigate the effects of a breastfeeding promotion intervention on adiposity and BP at age 16 years, and on longitudinal growth trajectories from birth.
Design
Cluster-randomized Promotion of Breastfeeding Intervention Trial, allocated in 1996–1997 into intervention (n=16) or control arms (n=15).
Setting
Belarusian maternity hospitals and affiliated polyclinics.
Participants
17,046 breastfeeding mother-infant pairs, of whom 13,557 (79.5%) children were followed-up at 16.
Intervention
Breastfeeding promotion, modeled on the Baby-Friendly Hospital Initiative.
Main Outcome Measures
Body mass index (BMI); fat and fat-free mass indices (FMI and FFMI) and percent body fat from bioimpedance; waist circumference; overweight and obesity; height; BP; and longitudinal growth trajectories. The primary analysis was modified intention-to-treat (without imputation for losses to follow-up), accounting for within-clinic clustering.
Results
The intervention substantially increased breastfeeding duration and exclusivity compared with the control arm (exclusively breastfed: 45% vs 6% at 3 months, respectively). Mean differences at 16 years between intervention and control groups were: 0.21 kg/m2 (95%CI:0.06, 0.36) for BMI; 0.21 kg/m2 (−0.03, 0.44) for FMI; 0.00 kg/m2 (−0.21, 0.22) for FFMI; 0.71% (−0.32, 1.74) for percent body fat; −0.73 cm (−2.48, 1.02) for waist circumference; 0.05 cm (−0.85, 0.94) for height; −0.54 mmHg (−2.40, 1.31) for systolic BP; and 0.71 mmHg (−0.68, 2.10) for diastolic BP. The odds ratio for overweight/obesity (BMI≥85th vs <85th percentile) was 1.14 (1.02, 1.28) and for obesity (BMI≥95th vs <95th percentile) was 1.09 (0.92, 1.29). The intervention resulted in a more rapid rate of gain in post-infancy height (1 to 2.8 years), weight (2.8 to 14.5 years), and BMI (2.8 to 8.5 years) compared to the control arm. The intervention had little effect on BMI z-score changes after 8.5 years.
Conclusions
A randomized intervention that increased the duration and exclusivity of breastfeeding did not lower adolescent obesity risk or BP. On the contrary, the prevalence of overweight/obesity were higher in the intervention arm. All mothers initiated breastfeeding, so findings may not apply to comparisons of the effects of breastfeeding versus formula-feeding.
Keywords: Breastfeeding, adiposity, stature, blood pressure, growth, childhood
Introduction
The prevalence of childhood obesity has risen substantially in recent decades around the world.(1) In turn, obese children are more likely to become obese adults(2) and suffer obesity-related chronic illnesses.(3) However, few interventions to prevent childhood obesity have proven effective.(1, 4) Promoting greater uptake and duration of exclusive breastfeeding is a suggested public health measure to reduce childhood obesity(5) and its metabolic consequences (e.g. high blood pressure, BP).(6) This approach is based on mechanistic studies, for example those finding that the lower protein content of breastmilk (in comparison to formula milk) may reduce adipocyte development,(7) and a body of observational human data suggesting inverse associations of breastfeeding and its duration with later obesity.(6, 8–12) However, observational studies are prone to confounding by social patterning of both breastfeeding and growth,(13) the epidemiological evidence is inconsistent(6, 13–23) and publication bias is a concern.(24, 25) Furthermore, weight and BP change dynamically during development, but most previous studies measure these outcomes on a single occasion, rather than on multiple occasions at different ages among the same individuals.
The Promotion of Breastfeeding Intervention Trial (PROBIT) was designed to overcome limitations inherent in observational studies of the long-term effects of breastfeeding on child outcomes, including adiposity and blood pressure. We cluster-randomized 17,046 children from 31 clinics, born in 1996–1997 to either a control arm or breastfeeding promotion intervention (based on the World Health Organization and United Nations Children’s Fund (WHO/UNICEF) Baby-Friendly Hospital Initiative).(26) Trial inclusion criteria required: a) healthy, term (≥ 37 weeks gestation), and normal weight (≥ 2500g) singleton infants with an Apgar score of ≥ 5 at 5 minutes; and b) mothers who initiated breastfeeding with no condition expected to interfere with their ability to breastfeed.(26) The breastfeeding promotion intervention substantially increased breastfeeding duration and exclusivity when compared with the control arm (exclusively breastfed: 45% vs 6% at 3 months and 6.6% vs 0.7% at 6 months).(26) Our trial, therefore, provides a unique opportunity to test, in an intention-to-treat analysis, the extent to which breastfeeding causally influences adiposity, stature and blood pressure, making an important contribution to the debate about whether breastfeeding is protective against obesity.(5, 27)
We previously reported no evidence of a protective effect of the breastfeeding intervention on adiposity or blood pressure at 6.5 and 11.5 years.(28–30) We now analyze these outcomes at 16 years, when the children were beyond adiposity rebound, most had attained (or nearly attained) adult stature, and adiposity and blood pressure measures should better predict adult levels than at earlier ages. In addition to outcomes at single timepoints reported in previous publications, we take advantage of repeated weight and length/height measures taken from birth to adolescence to examine the effects of the intervention on longitudinal growth trajectories, which have not been examined in previous studies of breastfeeding.
METHODS
Randomization
A detailed description of the trial design has previously been published.(26) Briefly, the units of randomization (clusters) were 31 maternity hospitals and their associated outpatient polyclinics (which manage both well and ill children) in Belarus (Figure 1). These units were randomized to either: i) a control arm that continued the breastfeeding practices and policies already in effect at the time of randomization (which was typically characterized by a short duration of exclusive breastfeeding, early introduction of other drinks or foods, and weaning at about 3 months)(31); or ii) an experimental intervention arm based on the Baby-Friendly Hospital Initiative.(26)
Figure 1. Flow diagram of progress of clusters and individuals through PROBIT recruitment and follow-up phases I, II, III and IV.
aDuring PROBIT III, 6 deaths were reported in the intervention arm. Data checking during PROBIT IV found one of these children had been incorrectly reported as deceased and data were amended.
bOf the 13557 seen at PROBIT IV, 12072 were seen at both PROBIT II & III, 274 were not seen at either PROBIT II & III, 449 were seen at PROBIT II but not seen at III, and 762 were seen at PROBIT III but not seen at II. Of the 3489 children randomized but not followed up at 16 years, 267 attended the excluded site, 116 died since randomization, 2674 were lost to follow- up, and 432 were unable or unwilling to come for their clinic visit.
Follow-up
We have previously reported on anthropometry and BP outcomes between birth and 12 months (anthropometry only(32)), at 6.5 years(28) and at 11.5 years.(29, 30) The outcomes reported in the current follow-up are at mean age 16, measured between September 2012 and July 2015. At the 16-year follow-up, all anthropometric outcomes were measured in duplicate at dedicated research clinics using uniform research-specific equipment. These outcomes were: weight, percent body fat, fat mass and fat-free (lean) mass, measured by foot-to-foot bioelectrical impedance (Tanita TBF 300GS body fat analyzer); waist circumference measured using a nonstretchable measuring tape; standing height using a wall-mounted stadiometer (Medtechnika) and systolic and diastolic blood pressure measured in duplicate using a digital oscillometric device (705IT; Omron Healthcare, Milton Keynes, United Kingdom). The outcome measurements and their timing for the earlier research visits are summarized in eTable 1.
Training and quality assurance procedures have been described in detail.(26, 28, 29, 33) Our quality assurance processes raised concerns about the validity of the 16-year follow-up data from one polyclinic (N=267 originally enrolled), and we therefore excluded the 16-year data from this clinic in the current analysis. In the remaining 30 polyclinics (15 intervention, 15 control), the children were followed up by one or two research pediatricians depending on clinic volume.
Derived variables
Duplicate measures of height and waist circumference were taken; if the measurements differed by more than 0.5 cm, third (and fourth, if necessary) measurements were taken and all readings averaged. We calculated body mass (BMI), fat mass (FMI), and fat-free mass (FFMI) indices as weight, fat mass, and fat-free mass in kilograms divided by height in meters squared. We calculated waist-to-height ratio by dividing waist circumference in centimeters by standing height in centimeters. We defined overweight as BMI between the 85th to <95th percentiles and obesity as BMI at or above the 95th percentile, based on the Centers for Disease Control and Prevention (CDC) 2000 age- and sex-specific reference data.(34) We used three dichotomous outcomes: BMI ≥95th percentile vs <95th percentile, BMI ≥85th percentile vs <85th percentile, and BMI ≥25 kg/m2 vs <25 kg/m2. For the longitudinal trajectory analysis, BMI z-scores were calculated using the WHO standard/reference.(35–37)
The analyses of growth trajectories included 17,042 (99%) children with at least one measurement of weight, length (to age 2 years) or height (after 2 years). We parameterized the relationships of weight, stature or BMI z-score with age using linear splines with 5 knot points at 3 months, 12 months, 2.8 years, 8.5 years and 14.5 years to describe periods of approximately linear growth based on the data.(38) Although a linear spline model is an approximation of the true non-linear growth function, its coefficients are easily interpretable and have produced a good model fit in this and other cohorts.(39–44) The knot points at 8.5 and 14.5 years were chosen because those were the oldest ages at the 6.5- and 11.5-year follow-ups. Setting the knot points at the median (or 25% or 75%) age of the 6.5- and 11.5-year follow-ups, resulted in similar findings.
The parent or guardian (usually the mother) who accompanied the child at the 6.5-year follow-up reported weight and height for herself and her partner; at the 11.5-year follow-up we measured mother’s weight and height if she attended. The most recent measurements of parental height and calculated BMI were used for analysis.
Reproducibility
Audit visits were conducted to assess inter-observer reproducibility of the outcome measurements, an important feature, given that blinding of pediatricians to the intervention vs control group assignment was not feasible. In the 24 lower-volume polyclinics with a single pediatrician, 4 children were randomly selected to return for re-measurement of all variables. For the 6 higher-volume clinics with 2 study pediatricians, 3 children per pediatrician were selected. Thus, a total of 132 children were audited. So that all children seen in follow-up were eligible for the repeated measurements, the selection was carried out after completion of primary data collection, an average of 1.2 years (range, 0.02–2.5) after the initial visit. The audit was carried out by 1 of 3 Minsk-based pediatricians not involved in primary data collection and blinded to the measures obtained at the initial visit but not to trial arm. Because of the time elapsed between the audit and initial visits, reproducibility was assessed using Pearson correlation coefficients.
Governance and ethics
The 16-year follow-up was approved by the Belarussian Ministry of Health and received ethical approval from the McGill University Health Centre Research Ethics Board, the Institutional Review Board at Harvard Pilgrim Health Care and the Avon Longitudinal Study of Parents and Children (ALSPAC) Law and Ethics Committee. A parent or legal guardian provided informed consent in Russian at enrollment and at all follow-up visits, and all children provided written assent at the 11.5- and 16-year clinic visits.
Statistical analysis
Comparisons between the intervention and control groups were based on a modified intention-to-treat analysis without imputation for missing outcome data (i.e., based on the 13,557 children with observed outcomes). We accounted for possible non-independence of measurements within individual clinics (clustering) using mixed-effects models. In a sensitivity analysis, we used SAS multiple imputation for N=17,046 individuals to impute 20 values for each missing observation (including outcomes at 16 years) and combined multivariable modeling estimates using Proc MI ANALYZE in SAS (see Supplementary materials(45, 46)). For the trajectory analyses we employed a three-level multilevel model: i) measurement occasion; ii) individual child; and iii) clinic site where the child was examined; these analyses were conducted in STATA version 13.1 (StataCorp. 2013. College Station, TX: StataCorp LP)(47) and MLwiN version 2.36.(48)
Results are presented for: i) the simple cluster-adjusted model; ii) the model after additional adjustment for age at follow-up and baseline characteristics: stratum-level variables (urban vs rural and East vs West Belarus residence), maternal and paternal education, child sex, and birth weight (for adiposity, standing height and blood pressure outcomes); iii) the model after further adjustment for measured parental BMI (for adiposity and weight gain), parental height (for child height outcomes) or parental BMI and height (for blood pressure and BMI gain). Models ii) and iii) were implemented in case of baseline imbalances, given the relatively small number of randomized clusters.
The intention-to-treat analysis likely underestimates the magnitude of effect of breastfeeding exclusivity and duration, owing to overlap in breastfeeding between the randomized groups – many intervention mothers did not exclusively breastfeed for 3 or 6 months, and some control mothers did. In a secondary analysis, we applied instrumental variable methods(49) to account for non-adherence. The instrumental variable analysis robustly estimates the causal effect of having been exclusively breastfed for ≥3 months (versus <3 months), using randomization status as an instrument (i.e., a variable causally related to exclusive breastfeeding but not to the adiposity outcomes, except through breastfeeding), assuming that randomization status is independent of any confounders of the exposure-outcome relationships. As such, the effect estimates from instrumental variable analyses are not affected by measured or unknown variables that may confound the exposure-outcome association. We performed instrumental variable estimation of continuous outcomes using the generalized 2-stage least squares estimator, and of dichotomous outcomes using a probit model for instrumental variable analysis,(50) both of which account for within-clinic clustering.
For comparison with previous observational studies, we conducted observational analyses (i.e., disregarding randomization status) to estimate associations of the duration of any or exclusive breastfeeding on the same outcomes as the intention-to-treat analysis, also accounting for clustering and the same baseline characteristics as described above, using multiple linear regression for continuous outcomes and multiple logistic regression for dichotomous outcomes. In a sensitivity analysis, we stratified the results by whether or not the children correctly identified their originally allocated trial arm to determine if this knowledge biased the outcomes.
To provide context, we also present the observational associations of study outcomes with other non-breastfeeding baseline characteristics previously suggested to be early-life determinants of overweight and obesity.
RESULTS
A total of 13,557 children were examined at a median age of 16.2 years (SD, 0.5; IQR, 15.8–16.4), representing 79.5% of the 17,046 originally randomized and 80.8% of the 16,779 from included sites (Figure 1). Follow-up rates were similar in the intervention (82%) and control (79%) arms overall, although they varied from 41% to 98% at the different clinics. The children followed-up at 16 years in the intervention and control groups were similar in baseline characteristics, with small differences paralleling those previously reported at randomization (Table 1).(26) The audit showed high correlations (Pearson r ≥0.83) between initial clinic results and blinded repeat measures of weight, fat mass, fat-free mass, percent fat, waist circumference and standing height. The correlations were lower for systolic (r=0.55) and diastolic (r=0.37) blood pressure (eTable 2). All 16-year outcome measures showed a low degree of within-polyclinic clustering (ICC range 0.003–0.09).(33)
Table 1.
Baseline characteristics (N = 13557)
| Characteristic | Total N = 13557 |
Intervention N = 7064 |
Control N = 6493 |
|---|---|---|---|
| Measured at child’s birth | |||
| Maternal age, years: N (%) | |||
| <20 | 1820 (13.4) | 979 (13.9) | 841 (13.0) |
| 20–34 | 11173 (82.4) | 5792 (82.0) | 5381 (82.9) |
| ≥35 | 564 (4.2) | 293 (4.1) | 271 (4.2) |
| Maternal education: N (%) | |||
| Completed university | 1842 (13.6) | 1002 (14.2) | 840 (12.9) |
| Advanced secondary or partial university | 6925 (51.1) | 3365 (47.6) | 3560 (54.8) |
| Common secondary | 4318 (31.9) | 2406 (34.1) | 1912 (29.4) |
| Incompleted secondary | 472 (3.5) | 291 (4.1) | 181 (2.8) |
| Paternal education: N (%) | |||
| Completed university | 1737 (12.8) | 936 (13.3) | 801 (12.3) |
| Advanced secondary or partial university | 6205 (45.8) | 2910 (41.2) | 3295 (50.7) |
| Common secondary | 4883 (36.0) | 2828 (40.0) | 2055 (31.6) |
| Incompleted secondary or unknown | 732 (5.4) | 390 (5.5) | 342 (5.3) |
| Stratum-level variable: N (%) | |||
| East/Urban | 4150 (30.6) | 2215 (31.4) | 1935 (29.8) |
| East/Rural | 2152 (15.9) | 1075 (15.2) | 1077 (16.6) |
| West/Urban | 3524 (26.0) | 2296 (32.5) | 1228 (18.9) |
| West/Rural | 3731 (27.5) | 1478 (20.9) | 2253 (34.7) |
| Number of older children in household: N (%) | |||
| 0 | 7707 (56.8) | 4152 (58.8) | 3555 (54.8) |
| 1 | 4717 (34.8) | 2365 (33.5) | 2352 (36.2) |
| 2+ | 1133 (8.4) | 547 (7.7) | 586 (9.0) |
| Maternal smoking during pregnancy: N (%) | |||
| No | 13287 (98.0) | 6898 (97.7) | 6389 (98.4) |
| Yes | 270 (2.0) | 166 (2.3) | 104 (1.6) |
| Child sex: N (%) | |||
| Female | 6576 (48.5) | 3474 (49.2) | 3102 (47.8) |
| Male | 6981 (51.5) | 3590 (50.8) | 3391 (52.2) |
| Birth weight, kg: Mean (SD) | 3.44 (0.42) | 3.44 (0.42) | 3.44 (0.42) |
SD = standard deviation.
The results of the primary analysis are shown in Table 2. There was little consistent evidence that the intervention-effects differed in boys compared to girls (P values for sex-interactions in Table 2). The raw mean values of BMI, fat mass index, percent body fat, standing height and blood pressure, and the prevalence of overweight and obesity, were slightly higher in the intervention vs control arms. The cluster-adjusted odds ratio for overweight/obesity was 1.14 (95% CI, 1.02 to 1.28) and for obesity was 1.09 (95% CI, 0.92 to 1.29). Further controlling for baseline (Table 2) and parental characteristics (eTable 3), or multiply imputing outcomes (eTable 4 and eMethods), did not alter these conclusions.
Table 2.
Modified intention-to-treat analysis (without imputation) showing differences in adiposity measures, height and blood pressure comparing intervention vs control groups
| Outcome at 16 years (N = 13557) |
Intervention | Control | Difference in mean (95% CI)
|
||||||
|---|---|---|---|---|---|---|---|---|---|
| N | Mean (SD) | N | Mean (SD) | Cluster adjusted | p-value | P for sex int. | Further adjusted for baseline factors and age at follow-upa | p-value | |
| BMI, kg/m2 | 7057 | 21.5 (3.4) | 6480 | 21.2 (3.3) | 0.21 (0.06, 0.36) | 0.01 | 0.92 | 0.19 (0.04, 0.34) | 0.01 |
| FMI, kg/m2 | 6997 | 4.2 (2.6) | 6462 | 4.0 (2.5) | 0.21 (−0.03, 0.44) | 0.09 | 0.41 | 0.20 (−0.01, 0.42) | 0.06 |
| FFMI, kg/m2 | 6997 | 17.2 (2.1) | 6462 | 17.2 (2.1) | 0.00 (−0.21, 0.22) | 0.98 | 0.13 | 0.00 (−0.21, 0.21) | 1.00 |
| Body fat, % | 7043 | 18.9 (9.0) | 6462 | 18.2 (8.8) | 0.71 (−0.32, 1.74) | 0.18 | 0.02 | 0.69 (−0.26, 1.64) | 0.16 |
| Waist circumference, cm | 7061 | 73.6 (8.5) | 6482 | 74.7 (8.1) | −0.73 (−2.48, 1.02) | 0.41 | <.001 | −0.44 (−2.13, 1.24) | 0.60 |
| Waist-to-height ratio (× 100) | 7059 | 43.2 (4.7) | 6479 | 43.9 (4.6) | −0.45 (−1.50, 0.59) | 0.92 | 0.97 | −0.30 (−1.25, 0.64) | 0.82 |
| Standing height, cm | 7061 | 170.4 (8.5) | 6489 | 170.3 (8.5) | 0.05 (−0.85, 0.94) | 0.40 | <.001 | 0.08 (−0.60, 0.76) | 0.53 |
| Systolic BP, mm Hg | 7061 | 120.5 (11.7) | 6484 | 119.9 (11.1) | −0.54 (−2.40, 1.31) | 0.57 | 0.64 | −0.48 (−2.10, 1.13) | 0.56 |
| Diastolic BP, mm Hg | 7061 | 68.8 (7.6) | 6484 | 67.8 (7.0) | 0.71 (−0.68, 2.10) | 0.32 | 0.14 | 0.52 (−0.73, 1.76) | 0.41 |
| N (%) | N (%) | Odds Ratio (95% CI) | Odds Ratio (95% CI) | ||||||
| BMI ≥25 vs <25 kg/m2 | 892 (12.6) | 696 (10.7) | 1.20 (1.06, 1.37) | 0.004 | 0.85 | 1.19 (1.04, 1.35) | 0.01 | ||
| BMI ≥85th vs <85th %ileb | 1026 (14.5) | 842 (13.0) | 1.14 (1.02, 1.28) | 0.03 | 0.76 | 1.15 (1.01, 1.30) | 0.04 | ||
| BMI ≥95th vs <95th %ileb | 319 (4.5) | 270 (4.2) | 1.09 (0.92, 1.29) | 0.31 | 0.53 | 1.12 (0.94, 1.33) | 0.20 | ||
BMI = body mass index; BP = blood pressure; CI = confidence interval; FMI = fat mass index; FFMI = fat free mass index; SD = standard deviation; Sex int. = sex interaction.
Additionally adjusted for stratum-level variables (urban vs rural and East vs West Belarus residence), maternal and paternal education, child sex, birth weight and age at follow-up
Based on Centers for Disease Control and Prevention (CDC) 2000 reference data.(34)
Compared to the control arm, infants in the intervention arm had more rapid weight and length gain in the first 3 months, followed by lower weight and length gain between 3–12 months (Table 3 and Figure 2), as reported previously(32). In the present updated analysis, we found that the intervention resulted in more rapid growth in length than the control arm between 1 and 2.8 years, and in more rapid weight gain between 2.8 and 14.5 years. The rate of BMI change between 2.8 and 8.5 years was slightly higher in the intervention arm, reflecting the greater weight gain during this period. Sex-specific results are presented in eTable 5.
Table 3.
Modified intention-to-treat analysis (without imputation) showing differences in growth trajectories comparing intervention vs control groups
| Growth (N = 17042) | Intervention | Control | Difference in mean (95% CI) | ||||
|---|---|---|---|---|---|---|---|
| Mean (95% CI) N = 8864 |
Mean (95% CI) N = 8178 |
Cluster adjusted | p-value | P for sex int. | Further adjusted for baseline factorsa | p-value | |
| Birth weight, kg | 3.4 (3.38, 3.41) | 3.38 (3.37, 3.4) | 0.01 (−0.01, 0.04) | 0.38 | 0.02b | 0.02 (−0.01, 0.04) | 0.17 |
| Weight gain, kg/year: | |||||||
| Birth-3m | 11.4 (11.35, 11.45) | 11.05 (11, 11.1) | 0.35 (0.28, 0.42) | <.001 | 0.32 (0.25, 0.39) | <.001 | |
| >3–12m | 6.17 (6.14, 6.19) | 6.29 (6.26, 6.31) | −0.12 (−0.16, −0.08) | <.001 | −0.09 (−0.12, −0.05) | <.001 | |
| >1–2.8y | 1.81 (1.79, 1.83) | 1.8 (1.78, 1.82) | 0.01 (−0.02, 0.04) | 0.48 | 0.01 (−0.03, 0.04) | 0.69 | |
| >2.8–8.5y | 2.26 (2.24, 2.28) | 2.18 (2.16, 2.19) | 0.08 (0.06, 0.11) | <.001 | 0.07 (0.04, 0.1) | <.001 | |
| >8.5–14.5y | 4.65 (4.6, 4.7) | 4.58 (4.53, 4.63) | 0.07 (0, 0.14) | 0.04 | 0.09 (0.02, 0.16) | 0.01 | |
| >14.5–18.9y | 4.62 (4.45, 4.78) | 5.01 (4.83, 5.19) | −0.39 (−0.63, −0.15) | 0.002 | −0.29 (−0.52, −0.07) | 0.01 | |
| Birth length, cm | 51.69 (51.46, 51.91) | 51.86 (51.63, 52.09) | −0.17 (−0.49, 0.16) | 0.31 | <.001b | −0.05 (−0.35, 0.24) | 0.73 |
| Stature gain, cm/year: | |||||||
| Birth-3m | 37.9 (37.7, 38.09) | 36.12 (35.91, 36.32) | 1.78 (1.5, 2.06) | <.001 | 1.24 (0.97, 1.52) | <.001 | |
| >3–12m | 20.98 (20.9, 21.07) | 21.24 (21.15, 21.32) | −0.25 (−0.37, −0.14) | <.001 | −0.3 (−0.42, −0.18) | <.001 | |
| >1–2.8y | 9.83 (9.76, 9.91) | 9.62 (9.54, 9.7) | 0.22 (0.11, 0.33) | <.001 | 0.16 (0.05, 0.27) | 0.005 | |
| >2.8–8.5y | 6.87 (6.84, 6.91) | 6.86 (6.82, 6.9) | 0.01 (−0.04, 0.07) | 0.65 | 0.04 (−0.01, 0.1) | 0.11 | |
| >8.5–14.5y | 5.2 (5.16, 5.24) | 5.23 (5.18, 5.27) | −0.03 (−0.09, 0.04) | 0.39 | −0.06 (−0.12, 0.01) | 0.08 | |
| >14.5–18.9y | 3.23 (3.07, 3.39) | 3.67 (3.5, 3.85) | −0.44 (−0.68, −0.21) | <.001 | −0.26 (−0.45, −0.07) | 0.007 | |
| BMI at birth, z-score | −0.52 (−0.58, −0.46) | −0.62 (−0.68, −0.56) | 0.1 (0.01, 0.19) | 0.03 | 0.3b | 0.06 (−0.02, 0.15) | 0.15 |
| BMI gain, z-score/year: | |||||||
| Birth-3m | 1.66 (1.56, 1.76) | 1.79 (1.69, 1.9) | −0.13 (−0.27, 0.01) | 0.07 | 0.01 (−0.14, 0.15) | 0.92 | |
| >3–12m | 1.8 (1.76, 1.83) | 1.95 (1.92, 1.99) | −0.16 (−0.21, −0.11) | <.001 | −0.12 (−0.16, −0.07) | <.001 | |
| >1–2.8y | −0.5 (−0.52, −0.49) | −0.49 (−0.51, −0.48) | −0.01 (−0.04, 0.02) | 0.47 | −0.01 (−0.03, 0.02) | 0.67 | |
| >2.8–8.5y | −0.07 (−0.08, −0.07) | −0.1 (−0.11, −0.09) | 0.03 (0.02, 0.04) | <.001 | 0.01 (0, 0.03) | 0.02 | |
| >8.5–14.5y | 0.06 (0.05, 0.07) | 0.07 (0.06, 0.08) | −0.01 (−0.02, 0) | 0.07 | 0 (−0.01, 0.02) | 0.56 | |
| >14.5–18.9y | −0.08 (−0.1, −0.06) | −0.11 (−0.13, −0.08) | 0.02 (0, 0.05) | 0.10 | 0 (−0.03, 0.03) | 0.81 | |
BMI = body mass index; CI = confidence interval; m = months; y = years; Sex int. = sex interaction.
Additionally adjusted for stratum-level variables (urban vs rural and East vs West Belarus), both maternal and paternal education and child sex.
P-value for sex interaction for entire trajectory.
Figure 2. Predicted difference in mean weight, height and BMI z-score (with 95% confidence intervals) in the intervention arm compared to control arm.
Predicted size in the intervention compared to the control at age 1, 6.5, 11.5 and 16.2 years
Using multilevel models to estimate mean differences in weight, height and BMI z-scores between intervention and control groups at the mean clinic age revealed differences broadly in line with the cluster-adjusted estimates (Figure 2). Although children were heavier in the intervention compared to the control arm, they were also taller, and their BMI z-scores showed little overall difference from mid-childhood.
The instrumental variable results are in line with those of the primary analysis (Table 4). Overall, 32.1% of the intervention group and 25.4% of the control group correctly identified the randomization arm to which they belonged, but such knowledge made little difference to the effect estimates (data not shown). In observational analyses, increased duration of exclusive (eTable 6) or any (eTable 7) breastfeeding was positively associated with several measures of adiposity, in line with the intention-to-treat results. eTable 8 presents the association of baseline characteristics with BMI category at 16 years. Estimates were in the expected direction for several sociodemographic and early-life variables.
Table 4.
Instrumental variable and observational associations analysis of duration of exclusive breastfeeding (≥3 months vs <3 months) with adiposity measures, height and blood pressure at 16 years
| (N = 13557) |
Instrumental variable results Exclusive breastfeeding ≥3 vs <3 months |
Observational analysis Exclusive breastfeeding ≥3 vs <3 months |
||
|---|---|---|---|---|
| Cluster adjusted | Further adjusted for baseline factors and age at follow-upa | Cluster adjusted | Further adjusted for baseline factors and age at follow-upa | |
|
|
|
|||
| Continuous outcomes | Difference in mean (95% CI) | Difference in mean (95% CI) | ||
| BMI, kg/m2 | 0.55 (0.09, 1.00) | 0.49 (−0.01, 0.99) | 0.19 (0.05, 0.33) | 0.19 (0.05, 0.33) |
| FMI, kg/m2 | 0.54 (−0.12, 1.20) | 0.52 (−0.28, 1.31) | 0.09 (−0.03, 0.20) | 0.06 (−0.04, 0.16) |
| FFMI, kg/m2 | 0.01 (−0.59, 0.61) | 0.01 (−0.65, 0.66) | 0.09 (0.00, 0.18) | 0.11 (0.04, 0.18) |
| Body fat, % | 1.87 (−1.05, 4.79) | 1.76 (−1.71, 5.23) | 0.23 (−0.16, 0.62) | 0.12 (−0.18, 0.42) |
| Waist circumference, cm | −1.93 (−6.74, 2.88) | −1.15 (−5.31, 3.02) | 0.23 (−0.13, 0.59) | 0.22 (−0.13, 0.57) |
| Waist-to-height ratio (× 100) | 0.12 (−2.08, 2.33) | 0.19 (−1.39, 1.78) | 0.10 (−0.28, 0.47) | 0.05 (−0.23, 0.33) |
| Standing height, cm | −1.20 (−4.16, 1.76) | −0.78 (−3.12, 1.55) | 0.10 (−0.10, 0.30) | 0.11 (−0.09, 0.31) |
| Systolic BP, mm Hg | −1.46 (−6.67, 3.76) | −1.21 (−4.93, 2.52) | −0.08 (−0.57, 0.42) | −0.04 (−0.50, 0.42) |
| Diastolic BP, mm Hg | 1.88 (−2.04, 5.80) | 1.32 (−2.71, 5.35) | 0.22 (−0.10, 0.53) | 0.20 (−0.11, 0.52) |
| Dichotomous outcomes | Odds Ratio (95% CI) | Odds Ratio (95% CI) | ||
| BMI ≥25 vs <25 kg/m2 | 1.51 (1.11, 1.17) | 1.43 (1.08, 1.16) | 1.20 (1.06, 1.36) | 1.20 (1.06, 1.36) |
| BMI ≥85th vs <85th %ileb | 1.36 (1.03, 1.15) | 1.35 (1.02, 1.16) | 1.17 (1.04, 1.31) | 1.17 (1.04, 1.31) |
| BMI ≥95th vs <95th %ileb | 1.18 (0.87, 1.17) | 1.24 (0.89, 1.18) | 1.00 (0.83, 1.21) | 1.02 (0.84, 1.23) |
BMI = body mass index; FMI = fat mass index; FFMI = fat free mass index.
Adjusted for stratum-level variables (urban vs rural and East vs West Belarus residence), maternal and paternal education, child sex, birth weight and age at follow-up
Based on Centers for Disease Control and Prevention (CDC) 2000 reference data.(34)
COMMENT
In this large cluster randomized controlled trial, an intervention to promote increased duration and exclusivity of breastfeeding did not reduce levels of general or central adiposity or lower BP in children aged 16 years. Beyond infancy, the intervention resulted in more rapid growth in height and then more rapid weight gain in early and mid-childhood, respectively, but the intervention had little effect on BMI z-scores after 8.5 years.
Our findings are similar to results in the same children at age 6.5 years and 11.5 years.(28, 29) The minimal imbalances in baseline characteristics at enrollment and amongst those followed up provide reassurance that the randomization was successful and that confounding and selection bias are unlikely explanations for the results. In an observational analysis, we did not observe the inverse associations of increased breastfeeding with overweight and obesity reported in previous observational studies, possibly due to differences in confounding structures in Belarus compared to Western countries. The Pelotas (Brazil) cohort found no association of socioeconomic position with breastfeeding, and no strong association of breastfeeding with BMI or BP (similar to our observational analysis in Belarus).(13) This contrasts with the ALSPAC cohort, UK, in which higher socioeconomic position was strongly associated with increased breastfeeding, and breastfeeding was associated with lower BMI and BP, even after adjusting for socioeconomic position.(13) Such cross-cohort comparisons suggest that reported associations of breastfeeding with child BMI and BP in ALSPAC are likely to reflect residual confounding.(13)
Higher-than-expected breastfeeding duration was observed in the control group, which may have been due to deteriorating economic conditions in Belarus during the trial and the higher cost of formula.(26) Nonetheless, the intervention led to a substantial increase in breastfeeding duration and exclusivity compared to the control arm.(26) Breastfeeding was initiated in all study participants, so our findings may not apply to comparisons of breastfeeding versus formula feeding. Given the (expected) overlap in breastfeeding in the intervention and control arms, we used instrumental variable analysis to estimate unconfounded associations of the difference in breastfeeding exclusivity and duration achieved between the two randomized groups with adiposity and BP. The instrumental variable analysis supports our primary findings that the breastfeeding promotion intervention did not substantially lower the outcomes of interest. The small positive associations of the intervention with overweight and obesity could be a chance finding, but we cannot exclude a true increase in risk caused by the intervention. One suggested physiological mechanism whereby prolonged breastfeeding could increase the risk of obesity is longer exposure to maternal hormones present in breastmilk, which could theoretically alter the infant’s lipid metabolism and increase body fat composition in later life.(51)
Belarus has low overall levels of obesity and overweight (in our study, 4–5% were obese and 13–15% were overweight or obese at 16 years). Hence, our findings may not be generalizable to other settings with higher prevalence of overweight and obesity. Whilst many observational studies suggest that longer-term and exclusive breastfeeding reduces childhood obesity risk,(6) these studies are prone to confounding by life-style factors and publication bias.(27)
Conclusions
An intervention that achieved substantially greater duration and exclusivity of breastfeeding in Belarus did not prevent overweight or obesity, or lower BP levels at age 16 years, despite differences in growth rates between the trial arms at various ages. On the contrary, overweight and obesity were more prevalent in the breastfeeding promotion intervention arm. While there are many reasons for promoting breastfeeding duration and exclusivity, our trial does not indicate that breastfeeding prevents obesity or lowers BP in childhood or adolescence.
Supplementary Material
Key points.
Question
What is the effect of a randomized intervention that increased breastfeeding duration and exclusivity on growth, adiposity and blood pressure (BP) at age 16 years?
Findings
Cluster-adjusted mean differences between intervention versus control groups were: BMI 0.21 kg/m2 (95% CI: 0.06, 0.36) (with similarly positive effects for other adiposity measures); systolic BP −0.54 mmHg (−2.40, 1.31); and diastolic BP 0.71 mmHg (−0.68, 2.10). The cluster-adjusted odds ratio for overweight/obesity was 1.14 (1.02, 1.28) and for obesity was 1.09 (0.92, 1.29).
Meaning
A randomized intervention that increased breastfeeding intensity did not reduce obesity or lower BP levels at 16.
Acknowledgments
We are grateful to the cohort members, their parents and the study pediatricians and auditors who participated so willingly in the study.
Funding/Support: This study was supported by grants from: Canadian Institutes of Health Research (MOP-53155); and USA National Institutes of Health (R01 HD050758). Prof Oken was supported by US National Institute of Health (K24 HD069408, P30 DK092924). Profs Martin and Tilling work in the Integrative Epidemiology Unit (IEU) supported by the United Kingdom Medical Research Council (MRC) and the University of Bristol (Grant Code: MC_UU_12013/1-9). The NIHR Bristol Nutrition Biomedical Research Unit is funded by the National Institute for Health Research (NIHR) and is a partnership between the University Hospitals Bristol NHS Foundation Trust and the University of Bristol.
Role of the Sponsors: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Footnotes
Trial Registration isrctn.org: ISRCTN37687716; and clinicaltrials.gov: NCT01561612
Author Contributions: Professor Martin and Ms. Rifas-Shiman had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Martin, Kramer, Oken.
Acquisition of data: Martin, Patel, Kramer, Vilchuck, Bogdanovich, Hameza, Thompson, Oken.
Analysis and interpretation of data: Martin, Kramer, Tilling, Rifas-Shiman, Patel, Yang, Oken.
Drafting of the manuscript: Patel, Martin, Rifas-Shiman, Tilling
Critical revision of the manuscript for important intellectual content: Martin, Kramer, Tilling, Patel, Vilchuck, Bogdanovich, Rifas-Shiman, Yang, Thompson, Oken.
Statistical analysis: Martin, Rifas-Shiman, Patel, Tilling.
Obtained funding: Martin, Kramer, Oken.
Administrative, technical, or material support: Martin, Patel, Vilchuck, Bogdanovich, Hameza, Thompson, Oken.
Study supervision: Martin, Kramer, Vilchuck, Bogdanovich, Oken.
Conflict of Interest Disclosures: None
References
- 1.Lobstein T, Jackson-Leach R, Moodie ML, et al. Child and adolescent obesity: part of a bigger picture. The Lancet. 2015;385(9986):2510–20. doi: 10.1016/S0140-6736(14)61746-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Simmonds M, Llewellyn A, Owen C, Woolacott N. Predicting adult obesity from childhood obesity: a systematic review and meta-analysis. Obesity reviews. 2016;17(2):95–107. doi: 10.1111/obr.12334. [DOI] [PubMed] [Google Scholar]
- 3.Kelsey MM, Zaepfel A, Bjornstad P, Nadeau KJ. Age-Related Consequences of Childhood Obesity. Gerontology. 2014;60(3):222–8. doi: 10.1159/000356023. [DOI] [PubMed] [Google Scholar]
- 4.Ng M, Fleming T, Robinson M, et al. Global, regional, and national prevalence of overweight and obesity in children and adults during 1980–2013: a systematic analysis for the Global Burden of Disease Study 2013. The Lancet. 2014;384(9945):766–81. doi: 10.1016/S0140-6736(14)60460-8. doi: http://dx.doi.org/10.1016/S0140-6736(14)60460-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.World Health Organization Commission on Ending Childhood Obesity. Final report of the Commission on Ending Childhood Obesity. World Health Organization; Geneva: 2016. http://www.who.int/end-childhood-obesity/news/launch-final-report/en/ (accessed January 26, 2016). 2016. [Google Scholar]
- 6.Horta BL, Loret de Mola C, Victora CG. Long-term consequences of breastfeeding on cholesterol, obesity, systolic blood pressure and type 2 diabetes: a systematic review and meta-analysis. Acta Paediatrica. 2015;104(S467):30–7. doi: 10.1111/apa.13133. [DOI] [PubMed] [Google Scholar]
- 7.Arenz S, Ruckerl R, Koletzko B, von Kries R. Breast-feeding and childhood obesity-a systematic review. International Journal of Obesity & Related Metabolic Disorders: Journal of the International Association for the Study of Obesity. 2004;28(10):1247–56. doi: 10.1038/sj.ijo.0802758. [DOI] [PubMed] [Google Scholar]
- 8.Weng SF, Redsell SA, Swift JA, Yang M, Glazebrook CP. Systematic review and meta-analyses of risk factors for childhood overweight identifiable during infancy. Archives of Disease in Childhood. 2012;97(12):1019–26. doi: 10.1136/archdischild-2012-302263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Yan J, Liu L, Zhu Y, Huang G, Wang PP. The association between breastfeeding and childhood obesity: a meta-analysis. BMC Public Health. 2014;14(1):1. doi: 10.1186/1471-2458-14-1267. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Li C, Goran MI, Kaur H, Nollen N, Ahluwalia JS. Developmental Trajectories of Overweight During Childhood: Role of Early Life Factors. Obesity. 2007;15(3):760–71. doi: 10.1038/oby.2007.585. [DOI] [PubMed] [Google Scholar]
- 11.Grube MM, von der Lippe E, Schlaud M, Brettschneider A-K. 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(3):e0122534. doi: 10.1371/journal.pone.0122534. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Gillman MW, Rifas-Shiman SL, Camargo CA, Jr, et al. Risk of overweight among adolescents who were breastfed as infants. JAMA. 2001;285(19):2461–7. doi: 10.1001/jama.285.19.2461. [DOI] [PubMed] [Google Scholar]
- 13.Brion M-JA, Lawlor DA, Matijasevich A, et al. What are the causal effects of breastfeeding on IQ, obesity and blood pressure? Evidence from comparing high-income with middle-income cohorts. International Journal of Epidemiology. 2011 doi: 10.1093/ije/dyr020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hancox R, Stewart A, Braithwaite I, Beasley R, Murphy R, Mitchell E. Association between breastfeeding and body mass index at age 6–7 years in an international survey. Pediatric obesity. 2015;10(4):283–7. doi: 10.1111/ijpo.266. [DOI] [PubMed] [Google Scholar]
- 15.Fall CH, Borja JB, Osmond C, et al. Infant-feeding patterns and cardiovascular risk factors in young adulthood: data from five cohorts in low-and middle-income countries. International journal of epidemiology. 2011;40(1):47–62. doi: 10.1093/ije/dyq155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Zheng J-S, Liu H, Li J, et al. Exclusive breastfeeding is inversely associated with risk of childhood overweight in a large chinese cohort. The Journal of nutrition. 2014;144(9):1454–9. doi: 10.3945/jn.114.193664. [DOI] [PubMed] [Google Scholar]
- 17.Estévez-González M, del Pino AS, Henríquez-Sánchez P, Peña-Quintana L, Saavedra-Santana P. Breastfeeding during the first 6 months of life, adiposity rebound and overweight/obesity at 8 years of age. International Journal of Obesity. 2016;40(1):10–3. doi: 10.1038/ijo.2015.228. [DOI] [PubMed] [Google Scholar]
- 18.Van der Willik EM, Vrijkotte TG, Altenburg TM, Gademan MG, Kist-van Holthe J. Exclusively breastfed overweight infants are at the same risk of childhood overweight as formula fed overweight infants. Archives of Disease in Childhood. 2015;100(10):932–7. doi: 10.1136/archdischild-2015-308386. [DOI] [PubMed] [Google Scholar]
- 19.Victora CG, Barros F, Lima RC, Horta BL, Wells J. Anthropometry and body composition of 18 year old men according to duration of breast feeding: birth cohort study from Brazil. Bmj. 2003;327(7420):901. doi: 10.1136/bmj.327.7420.901. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Durmuş B, Heppe DH, Gishti O, et al. General and abdominal fat outcomes in school-age children associated with infant breastfeeding patterns. The American journal of clinical nutrition. 2014;99(6):1351–8. doi: 10.3945/ajcn.113.075937. [DOI] [PubMed] [Google Scholar]
- 21.Martin RM, Gunnell D, Smith GD. Breastfeeding in infancy and blood pressure in later life: systematic review and meta-analysis. American Journal of Epidemiology. 2005;161(1):15–26. doi: 10.1093/aje/kwh338. [DOI] [PubMed] [Google Scholar]
- 22.Lawlor D, Riddoch C, Page A, et al. Infant feeding and components of the metabolic syndrome: findings from the European Youth Heart Study. Archives of disease in childhood. 2005;90(6):582–8. doi: 10.1136/adc.2004.055335. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Victora CG, Bahl R, Barros AJ, et al. Breastfeeding in the 21st century: epidemiology, mechanisms, and lifelong effect. The Lancet. 2016;387(10017):475–90. doi: 10.1016/S0140-6736(15)01024-7. [DOI] [PubMed] [Google Scholar]
- 24.Owen CG, Martin RM, Whincup PH, Smith GD, Cook DG. Effect of infant feeding on the risk of obesity across the life course: a quantitative review of published evidence. Pediatrics. 2005;115(5):1367–77. doi: 10.1542/peds.2004-1176. [DOI] [PubMed] [Google Scholar]
- 25.Owen CG, Whincup PH, Gilg JA, Cook DG. Effect of breast feeding in infancy on blood pressure in later life: systematic review and meta-analysis. BMJ. 2003;327(7425):1189–95. doi: 10.1136/bmj.327.7425.1189. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kramer MS, Chalmers B, Hodnett ED, et al. Promotion of Breastfeeding Intervention Trial (PROBIT): a randomized trial in the Republic of Belarus. JAMA. 2001;285(4):413–20. doi: 10.1001/jama.285.4.413. [DOI] [PubMed] [Google Scholar]
- 27.Casazza K, Fontaine KR, Astrup A, et al. Myths, presumptions, and facts about obesity. New England Journal of Medicine. 2013;368(5):446–54. doi: 10.1056/NEJMsa1208051. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Kramer MS, Matush L, Vanilovich I, et al. Effects of prolonged and exclusive breastfeeding on child height, weight, adiposity, and blood pressure at age 6.5 y: evidence from a large randomized trial. American Journal of Clinical Nutrition. 2007;86(6):1717–21. doi: 10.1093/ajcn/86.5.1717. [DOI] [PubMed] [Google Scholar]
- 29.Martin RM, Patel R, Kramer MS, 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(10):1005–13. doi: 10.1001/jama.2013.167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Martin RM, Patel R, Kramer MS, et al. Effects of Promoting Longer Term and Exclusive Breastfeeding on Cardiometabolic Risk Factors at Age 11.5 Years: A Cluster-Randomized, Controlled Trial. Circulation. 2013 doi: 10.1161/circulationaha.113.005160. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Lawrence RA. Breastfeeding in Belarus. JAMA: The Journal of the American Medical Association. 2001;285(4):463–4. doi: 10.1001/jama.285.4.463. [DOI] [PubMed] [Google Scholar]
- 32.Kramer MS, Guo T, Platt RW, et al. Breastfeeding and infant growth: biology or bias? Pediatrics. 2002;110(2 Pt 1):343–7. doi: 10.1542/peds.110.2.343. [DOI] [PubMed] [Google Scholar]
- 33.Guthrie L, Oken E, Sterne J, et al. Ongoing monitoring of data clustering in multicenter studies. BMC Medical Research Methodology. 2012;12(1):29. doi: 10.1186/1471-2288-12-29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Ogden CL, Kuczmarski RJ, Flegal KM, et al. Centers for Disease Control and Prevention 2000 growth charts for the United States: improvements to the 1977 National Center for Health Statistics version. Pediatrics. 2002;109(1):45–60. doi: 10.1542/peds.109.1.45. [DOI] [PubMed] [Google Scholar]
- 35.Vidmar SI, Cole TJ, Pan H. Standardizing anthropometric measures in children and adolescents with functions for egen: Update. Stata Journal. 2013;13(2):366–78. [Google Scholar]
- 36.World Health Organization. WHO child growth standards: Length/height-for-age, weight-for-age, weight-for-length, weight-for-height and body mass index-for-age: Methods and development. Geneva: World Health Organization; 2006. http://www.who.int/childgrowth/standards/technical_report/en/ [Google Scholar]
- 37.World Health Organization. WHO Reference. Geneva: World Health Organization; 2007. 2007. http://www.who.int/growthref/en/ [Google Scholar]
- 38.Patel R, Tilling K, Lawlor D, et al. Socioeconomic differences in childhood length/height trajectories in a middle-income country: a cohort study. BMC Public Health. 2014;14(1):932. doi: 10.1186/1471-2458-14-932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Howe LD, Tilling K, Galobardes B, Davey Smith G, Gunnell D, Lawlor DA. Socioeconomic differences in childhood growth trajectories: at what age do height inequalities emerge? Journal of Epidemiology and Community Health. 2012;66(2):143–8. doi: 10.1136/jech.2010.113068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Matijasevich A, Howe LD, Tilling K, Santos IS, Barros AsJD, Lawlor DA. Maternal education inequalities in height growth rates in early childhood: 2004 Pelotas birth cohort study. Paediatric and Perinatal Epidemiology. 2012;26:236–49. doi: 10.1111/j.1365-3016.2011.01251.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Tilling K, Davies NM, Nicoli E, et al. Associations of growth trajectories in infancy and early childhood with later childhood outcomes. The American Journal of Clinical Nutrition. 2011;94(6 Suppl):1808S–13S. doi: 10.3945/ajcn.110.001644. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Anderson EL, Fraser A, Martin RM, et al. Associations of postnatal growth with asthma and atopy: the PROBIT Study. Pediatric Allergy and Immunology. 2013;24(2):122–30. doi: 10.1111/pai.12049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Fairley L, Petherick ES, Howe LD, et al. Describing differences in weight and length growth trajectories between white and Pakistani infants in the UK: analysis of the Born in Bradford birth cohort study using multilevel linear spline models. Archives of Disease in Childhood. 2013;98(4):274–9. doi: 10.1136/archdischild-2012-302778. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Howe LD, Tilling K, Matijasevich A, et al. Linear spline multilevel models for summarising childhood growth trajectories: A guide to their application using examples from five birth cohorts. Statistical Methods in Medical Research. 2013 doi: 10.1177/0962280213503925. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.White IR, Royston P, Wood AM. Multiple imputation using chained equations: issues and guidance for practice. Statistics in medicine. 2011;30(4):377–99. doi: 10.1002/sim.4067. [DOI] [PubMed] [Google Scholar]
- 46.Rubin DB. Multiple imputation for nonresponse in surveys. John Wiley & Sons; 2004. [Google Scholar]
- 47.Leckie G, Charlton C. runmlwin - A Program to Run the MLwiN Multilevel Modelling Software from within Stata. Journal of Statistical Software. 2013;52(11):1–40. [Google Scholar]
- 48.Rasbash J, Charlton C, Browne WJ, Healy M, Cameron B. MLwiN Version 2.32. Centre for Multilevel Modelling. University of Bristol; Bristol, UK: 2005. [Google Scholar]
- 49.Angrist JD, Imbens GW, Rubin DB. Identification of causal effects using instrumental variables. Journal of the American statistical Association. 1996;91(434):444–55. [Google Scholar]
- 50.Rassen JA, Schneeweiss S, Glynn RJ, Mittleman MA, Brookhart MA. Instrumental variable analysis for estimation of treatment effects with dichotomous outcomes. American Journal of Epidemiology. 2009;169(3):273–84. doi: 10.1093/aje/kwn299. [DOI] [PubMed] [Google Scholar]
- 51.O’Tierney PF, Barker DJ, Osmond C, Kajantie E, Eriksson JG. Duration of breast-feeding and adiposity in adult life. The Journal of Nutrition. 2009;139(2):422S–5S. doi: 10.3945/jn.108.097089. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.