|Home | About | Journals | Submit | Contact Us | Français|
To examine whether delivery by caesarean section is a risk factor for childhood obesity.
Prospective pre-birth cohort study (Project Viva).
Eight outpatient multi-specialty practices based in the Boston, Massachusetts area.
We recruited women during early pregnancy between 1999 and 2002, and followed their children after birth. We included 1255 children with body composition measured at 3 years of age.
Body mass index (BMI) z-score, obesity (BMI for age and sex ≥ 95th percentile), and sum of triceps + subscapular skinfold thicknesses, at 3 years of age.
284 children (22.6 percent) were delivered by caesarean section. At age 3, 15.7% of children delivered by caesarean section were obese, compared with 7.5% of children born vaginally. In multivariable logistic and linear regression models adjusting for maternal pre-pregnancy BMI, birth weight, and other covariates, birth by caesarean section was associated with a higher odds of obesity at age 3 (OR 2.10, 95%CI 1.36 to 3.23), higher mean BMI z-score (0.20 units, 95% CI 0.07 to 0.33), and higher sum of triceps + subscapular skinfold thicknesses (0.94 mm, 95% CI 0.36 to 1.51).
Infants delivered by caesarean section may be at increased risk of childhood obesity. Further studies are needed to confirm our findings and to explore mechanisms underlying this association.
Identifiying modifiable risk factors during the perinatal period may offer promising strategies to prevent obesity and its complications throughout the life course.1 Delivery by caesarean section has been identified as a risk factor for child asthma and allergic rhinitis,2,3 but only one previous study has examined the relationship between mode of delivery and child obesity.4 One potential rationale for examining the relationship between mode of delivery and child obesity, is that compared with vaginally-born infants, infants delivered by caesarean section exhibit differences in the composition and timing of acquisition of intestinal flora.5,6 These alterations in intestinal microbial composition in the first year of life may last through childhood, and may contribute to the development of obesity7–9 and other health outcomes. Mode of delivery also has the potential to influence long-term obesity risk through effects on inflammation, immune or endocrine function that are independent of the intestinal microbiota composition.
To our knowledge, no prospective studies have specifically examined whether caesarean delivery is associated with the risk of child obesity. An association between caesarean birth and increased risk of child obesity would provide an important rationale to avoid non-medically indicated caesarean section. In the U.S., the proportion of births by caesarean section increased from 20.7% in 1996 to 32% in 2007,10 probably in part because of increased rates of caesarean birth on maternal request.11 The study goal was to examine whether delivery by caesarean section was associated with a higher risk of child obesity at age 3 in a longitudinal pre-birth cohort.
From April 1999 to July 2002, we enrolled participants into Project Viva, a longitudinal pre-birth cohort of mother-offspring pairs in eastern Massachusetts, USA. Human Subjects Committees of Harvard Pilgrim Health Care, Brigham and Women’s Hospital (BWH), and Beth Israel Deaconess Medical Center (BIDMC) approved study protocols,12 and all mothers provided written informed consent.
We have previously described in detail the study population, enrollment and follow up procedures.12 We recruited women attending their initial prenatal visit before 22 weeks gestational age at Harvard Vanguard Medical Associates, a multi-specialty group practice. Eligibility criteria included fluency in English and singleton pregnancy. All mothers gave birth at one of two hospitals. A trained research assistant conducted in-person study visits with the mother at the end of the 1st and 2nd trimesters of pregnancy, and with both mother and child after delivery and at 6 months and 3 years after birth. At each in-person visit, we measured the infant’s length/height and weight; at 3 years of age, we also measured the child’s skinfold thicknesses. At 1 and 2 years postpartum, participants completed mailed questionnaires.
Of the 2128 women who delivered a live infant, 1579 were eligible for 3-year follow-up on the basis of having completed a prenatal nutrition assessment and providing consent for their children to participate. Of the 1579 participants, 182 were lost to follow-up. We collected follow-up information on 1,397 (88% of 1579), including in-person 3-year examinations on 1,292 (82%). We excluded 16 participants who lacked weight, height and skinfold measures at age 3, 1 participant who was missing exposure data, and 20 participants who were born before 34 weeks gestational age. Thus, our sample size for analysis was 1255 mother-child pairs. For analyses examining BMI z-score, overweight, and obesity outcomes, we excluded 18 participants lacking weight and/or height at age 3, and 6 participants with biologically implausible weight, height, or BMI, leaving 1237 participants. For analyses examining skinfolds, we excluded 56 participants who did not have skinfold measurements, leaving 1199 participants. Compared with the 296 children that were eligible but not included in the present analysis, children in the present study were more likely to have mothers of white race/ethnicity (73 v. 56%) and have college educated mothers (71 v. 54%). Among included participants, mean maternal BMI was slightly lower (24.6 v. 25.3 kg/m2) and birth weight was slightly higher (3517 v. 3474 g) than among excluded participants. Rates of caesarean birth were similar among included (23%) and excluded (24%) participants.
We obtained information about mode of delivery from electronic hospital records. For each participant who had a caesarean section recorded on electronic birth logs, we reviewed the operative report to confirm caesarean delivery and to abstract the primary indication for operative delivery. We defined an unplanned caesarean delivery as a delivery in which the operative report described a failed induction of labor, prolonged latent phase, prolonged active phase, arrest of dilation, “failure to progress”, arrest of descent in the second stage or failed operative vaginal delivery, “non-reassuring fetal heart rate tracing”, non-reassuring testing prompting immediate caesarean delivery, cord prolapse or abruption. We defined planned caesarean deliveries as those in which participants did not undergo a trial of labor (elective repeat caesarean without trial of labor, malpresentation, placenta previa, suspected macrosomia, maternal request, or other indication precluding trial of labor). We defined mode of delivery as a two-category variable: caesarean section vs vaginal delivery. We also performed separate analyses defining mode of delivery as a three-category variable: planned caesarean section, unplanned caesarean section, and vaginal delivery. For analyses using the three-category mode of delivery, we excluded 4 children for whom we were unable to determine whether the caesarean section was planned or unplanned. Unlike planned caesarean section, unplanned caesarean section is frequently accompanied by prior rupture of membranes, which might allow vaginal flora to ascend into the uterus, colonizing the fetus.2 Data on the timing of rupture of membranes were not available for these analyses.
For each child, we measured height using a research-standard stadiometer (Shorr Productions, Olney, MD), and weight using a digital scale (Seca model 881, Seca corporation, Hanover, MD), from which we calculated BMI (weight in kg/(height in m)2). We calculated age- and sex-specific BMI percentiles and z-scores using US national reference data.13 We defined obesity as a BMI (kg/m2) ≥ 95th percentile for age and sex,14,15 overweight as a BMI ≥ 85th and < 95th percentile for age and sex, and we used BMI < 85th percentile as the comparison group. We also calculated the sum (SS+TR) and ratio (SS:TR) of the children’s subscapular (SS) and triceps (TR) skinfold thicknesses, each measured using Holtain calipers (Holtain LTD, Crosswell, United Kingdom) with adequate training and evidence of reproducibility.
We collected sociodemographic and medical data through in-person interviews at enrollment, ages 6 months and 3 years; yearly self-administered questionnaires; and hospital and ambulatory medical records. Mothers reported their age, race/ethnicity, education, parity, pre-pregnancy weight and height, and paternal weight and height. We calculated gestational weight gain by subtracting pre-pregnancy weight from the last prenatal weight. To determine the reporting error using self-reported pre-pregnancy weight, we compared the weights for 170 participants who had clinic visit measurements recorded within 3 months of their last menstrual period for the index pregnancy with self-reported pre-pregnancy weight at the first trimester visit. The correlation coefficient between the two weights was 0.99, with underreporting of pre-pregnancy weight averaging 1 kg. Correlation coefficients and reporting error did not differ by maternal race/ethnicity or gestational age at enrollment into the study. We calculated gestational age at birth using the date of the last menstrual period. If the estimate of gestational age by second-trimester ultrasound assessment differed from the calculated gestational age by more than 10 days, we used the ultrasound dating. We obtained birth weight from medical records. Mothers reported number of hours their children spent in child care,16 timing of solid food introduction,17 breastfeeding duration, child diet,18 television viewing,19 and physical activity habits.
We used unadjusted and multivariable linear regression models to assess the associations between caesarean delivery and BMI z-score, SS+TR, and SS:TR at age 3 years. We used multinomial logistic regression to assess the associations between caesarean delivery and overweight (BMI 85th to <95th percentile) and obesity (BMI ≥95th percentile), and used <85th percentile as the comparison. In our multivariable model, we adjusted for maternal age, race/ethnicity, education, and BMI; and child age, sex and birth weight. For SS:TR models, we additionally adjusted for child BMI z-score. In an additional model, we repeated the analyses defining mode of delivery as a three-category variable: planned caesarean, unplanned caesarean and vaginal delivery. We excluded from our final models potential confounders that did not change our effect estimates, including household income; paternal BMI; maternal smoking, gestational weight gain, parity, and maternal glucose tolerance during pregnancy; and child gestational age at birth, initiation and duration of breastfeeding, timing of solid food introduction, energy intake and television viewing at age 2, and height at age 3. We used birth weight rather than birth weight for gestational age z-score to represent fetal size in our models, because caesarean delivery is more likely to be related to birth weight. Replacement of birth weight with birth weight for gestational age z-score in our final model made no difference to the effect estimates.
Because maternal BMI was likely to be strongly associated with both mode of delivery and child obesity, we examined potential confounding by maternal BMI in several ways. First, we examined the effect of adjustment for maternal BMI as a continuous variable and in deciles. The results were similar, so we defined maternal BMI as a continuous variable in our models. Second, we performed analyses stratified by maternal BMI status, categorized as < 25 or ≥ 25 kg/m2, in which we also adjusted for continuous maternal BMI within each category. We controlled for confounding by fetal size by adjustment for birth weight in our models, and by performing analyses stratified by birth weight, categorized as < 3.5 kg or ≥ 3.5 kg.
We conducted all data analyses using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA).
Participant characteristics are shown in Table 1. Of the 1255 deliveries, 22.6% were by caesarean section and 77.4% were vaginal deliveries. Mean maternal pre-pregnancy BMI was higher among infants delivered by caesarean section than for infants delivered vaginally. Birth weight for gestational age z-score, but not birth weight, was higher for caesarean-delivered than for vaginally-delivered infants. Breastfeeding duration was shorter for infants delivered by caesarean section. Compared with vaginal delivery, caesarean delivery was associated with a higher age 3 mean BMI z-score (0.67 vs 0.39 units), and higher mean sum of SS + TR skinfolds (17.5 mm vs 16.5 mm). Children delivered by caesarean section were more likely to be overweight (18.9% vs 16.7%) or obese (15.7% vs 7.5%), at age 3 than those delivered vaginally (Figure 1).
In multivariable models, caesarean delivery was associated with adverse age 3 adiposity outcomes (Table 2). In unadjusted multinomial logistic regression analyses, caesarean delivery was associated with 2.4-fold higher odds of obesity (95% CI 1.60 to 3.62). After adjustment for maternal age, education, race/ethnicity, and child age and sex (Model 1), the magnitude of the association was hardly changed (OR 2.43, 95% CI 1.60 to 3.68). Additional adjustment for maternal pre-pregnancy BMI and for birth weight (Model 3) slightly attenuated the relationship between caesarean delivery and risk of obesity (OR 2.10, 95% CI 1.36 to 3.23). Caesarean delivery was not significantly associated with age 3 overweight; the odds ratio was 1.24 (95% CI 0.86 to 1.77) after full adjustment.
In the fully adjusted (Model 3 covariates) linear regression models, caesarean delivery was associated with a 0.20 unit increment (95% CI 0.07 to 0.33) in age 3 BMI z-score and with a 0.94 mm increment (95% CI 0.36 to 1.51) in the sum of skinfolds, but it was not associated with the subscapular: triceps skinfold ratio, a measure of central adiposity (β −0.18, 95% CI −2.30 to 1.94). In addition, each kg/m2 increment in maternal BMI was associated with higher odds of child overweight (Model 3 OR 1.04, 95% CI 1.01 to 1.07) and obesity (Model 3 OR 1.10, 95% CI 1.06 to 1.13). Higher birth weight was also associated with child overweight (Model 3 OR 1.96 per kg increment in birth weight, 95% CI [1.44 to 2.68]) and obesity (Model 3 OR 2.02 per kg, 95% CI [1.35 to 3.03]).
In analyses stratified by maternal pre-pregnancy BMI (Table 3), caesarean delivery was associated with a nearly three-fold higher odds of obesity (OR 2.97, 95% CI 1.58 to 5.60) among children born to mothers with a normal pre-pregnancy BMI < 25 kg/m2. Among children born to overweight or obese mothers (pre-pregnancy BMI ≥ 25 kg/m2), caesarean delivery was associated with a somewhat elevated odds of obesity that was not statistically significant (OR 1.61, 95% CI 0.88 to 2.96). Caesarean delivery was associated with a doubling of the odds of obesity at age 3, regardless of birth weight (Table 3).
We performed additional analyses comparing the children with a planned (n=83) or unplanned (n=197) caesarean delivery to those born vaginally (n=971). In unadjusted analyses, planned caesarean section (OR 2.32, 95% CI 1.18 to 4.55) and unplanned caesarean section (OR 2.42, 95% CI 1.52 to 3.83) were both associated with a similar increase in odds of obesity compared with vaginal delivery. These odds ratios were somewhat attenuated by adjustment for covariates in Model 3: only unplanned caesarean delivery was clearly associated with a higher risk of obesity at age 3 (OR 2.19, 95% CI 1.34 to 3.55); the odds of obesity for planned caesarean delivery was less elevated and not statistically significant (Model 3 OR 1.83, 95%CI 0.89 to 3.77), but the confidence interval was wide because of a small sample size. Neither planned (OR 1.53, 95%CI 0.85 to 2.73) nor unplanned (OR 1.15, 95%CI 0.75 to 1.75) caesarean delivery was associated with odds of overweight at age 3 after covariate adjustment. Unplanned (β 0.22, 95% CI 0.06 to 0.37), but not planned (β 0.18, 95%CI −0.05 to 0.40) caesarean delivery was associated with higher mean BMI z-score. Both unplanned (β 0.70, 95%CI 0.03 to 1.36) and planned caesarean delivery (β 1.51, 95%CI 0.54 to 2.48) were associated with a higher sum of SS + TR skinfolds. Neither unplanned nor planned caesarean delivery was associated with the SS:TR skinfold ratio (data not shown).
In this prospective cohort study, we found that children delivered by caesarean section had double the odds of obesity, along with higher BMI (about 0.2 z-score units) and sum of skinfolds (about 1 mm) at age 3 compared with children who had been delivered vaginally. These associations remained even after controlling for key potential confounders, including maternal BMI and birth weight. For a 3-year-old child at the 50th percentile for weight and height, a 0.2 unit increment in BMI z-score would be equivalent to an increment of about 0.23 kg (0.5 lb).
Our findings suggesting that caesarean delivery may be an early life risk factor for obesity development are consistent with a small case-control study of 3 to 6-year-old Chinese children that reported higher odds of obesity (OR 5.23, 95% CI 1.24 to 22.04) associated with a caesarean delivery.4 In that study of 81 obese cases and 81 normal weight controls, the authors relied largely on data collected retrospectively using parental questionnaires, and did not have data regarding maternal pre-pregnancy BMI. In contrast, we had a larger sample size and adjusted for multiple key confounders, including maternal pre-pregnancy BMI, collected prospectively during pregnancy and childhood.
In our study, we were unable to directly examine potential mechanisms underlying the association between caesarean section and child obesity. One possible mechanism is that differences in the composition of intestinal microbiota acquired at birth among caesarean and vaginally-delivered newborns may contribute to their risk of obesity at age 3. Differences in child intestinal flora according to mode of delivery have been noted in the first year of life,5,6,20–22 a period of dramatic changes in number and diversity of gut microbes as well as rapid growth. Most,3–4,23,24 but not all6 studies, suggest that infants delivered by caesarean section have higher stool quantities of members of the Firmicutes group, or lower quantities of the Bacteroidetes group. The Firmicutes and Bacteroidetes bacteria constitute the majority of the microbiota in the adult human intestine.25 Data in mice and humans have shown that obese individuals display a relative abundance of Firmicutes and a lower proportion of Bacteroidetes than do lean individuals.25,26 Experiments in mice support the notion that the composition of intestinal microbiota may alter host body composition.27 Transplantation of intestinal microbiota obtained from obese donor mice (vs lean donors) into germ-free mice resulted in substantially greater percentage increase in recipient body fat (47% vs 27%, p<0.05) and recipient intestinal microbiota with a relative abundance of Firmicutes, resembling the source microbial composition.26 In humans, small prospective intervention studies with follow-up ranging from several weeks to a year, have shown that weight loss is associated with lowering of Firmicutes levels or higher Bacteroidetes levels,25,28,29 although not all studies are in agreement.30 The intestinal microbiota may influence obesity development by increasing energy extracted from the diet, and by effects on host epithelial and endocrine cells that promote insulin resistance, inflammation, and fat deposition.7,27
A few small case-control studies in Finnish children have directly examined whether intestinal microbiota composition in infancy is related to obesity in childhood. 8,9 Children who were overweight (vs normal weight) at ages 7–10 years had lower8 or a trend towards lower9 bifidobacterial quantities in stools collected during infancy. Stool quantities of Clostridia and Bacteroides did not significantly differ by weight status, but the small sample sizes (30 children9, 49 children8) may have limited power to detect differences. Our findings suggest a need for studies examining whether the association between caesarean delivery and child obesity is mediated by the types, quantities and functional effects of intestinal microbiota established in early life.
Other explanations for our findings are possible. Given the routine perioperative antibiotic prophylaxis accompanying caesarean delivery, caesarean delivery may be a proxy for intrapartum antibiotic use, which could influence the composition of neonatal intestinal flora, in turn influencing the development of obesity. One study that reported differences in microbiota composition by mode of delivery specifically excluded subjects who had received intrapartum or perinatal antibiotics.21 Among 1032 Dutch infants with stool microbial composition examined at one month of age, maternal antibiotic use during pregnancy was not associated with infant intestinal microbiota composition;22 antibiotic use during infancy was associated with reduced numbers of Bacteroides and bifidobacteria,22 a pattern that may be obesogenic,7 but other studies have not found any consistent effects of infant antibiotic use on host microbial composition.31 In our study, we were unable to examine the relationships among caesarean delivery, perinatal antibiotic use and childhood obesity.
We hypothesized that unplanned caesarean section might be associated with a risk of child obesity intermediate between the risk associated with planned caesarean section and vaginal delivery. This hypothesis was based on the theory that the rupture of membranes, assumed to occur in most unplanned caesarean sections, would allow the fetus some exposure to vaginal flora. Instead, the evidence was contrary to the hypothesis: we found that there was little difference in the risk of obesity, mean BMI-z, and sum of SS + TR skinfolds between planned and unplanned caesarean births. Perhaps physical passage of the infant through the birth canal is more important than the presence or duration22 of rupture of membranes in determining infant flora composition.
The mode of delivery might influence long-term obesity risk through effects on inflammation, immune or endocrine function that are independent of the intestinal microbiota composition. Labor is associated with many changes in levels of maternal and placental hormones and inflammatory cytokines, which we and others32 have hypothesized could influence the development of obesity. Piglet offspring delivered by caesarean section had greater hepatic steatosis and altered cholesterol metabolism compared with those delivered vaginally.33 In mice, oral exposure to lipopolysaccharide during vaginal, but not caesarean birth, triggered activation of gut epithelial cells.34 Stress-response signaling associated with labor may program the long-term function of the hypothalamic-pituitary-adrenal axis,35 or result in epigenetic modification36,37 of key metabolic genes that might induce an obese phenotype.38
Strengths of this study include a well-characterized cohort with adequate control for a large set of potential confounding variables, and careful measurement of child height and weight using research standards. Our study had several limitations. To calculate maternal BMI, we relied on self-reported maternal pre-pregnancy weight, which we showed to be highly correlated (r=0.99) with pre-pregnancy clinic weights in a subset of Project Viva participants. It is possible that this correlation would be lower for the whole cohort, because participants without recorded pre-pregnancy clinic weights may report their weights with greater error. Our study had some loss to follow-up, raising possible selection bias. Compared with non- participants, participating mothers differed on race/ethnicity, BMI, and infant birth weight. However, rates of caesarean birth were similar among included and excluded participants. Study participants had a relatively high level of education and income, which may limit generalizability. We cannot rule out the possibility of residual confounding as an explanation for our findings. We were particularly concerned about possible residual confounding by fetal size and by maternal BMI. Mean birth weight was only slightly and non-significantly higher for caesarean than for vaginally-delivered infants (p=0.11, Table 1), and the odds of obesity hardly changed when analyses were stratified by birth weight (Table 3). The association between mode of delivery and obesity risk remained robust after adjustment for maternal BMI defined as either a categorical or a continuous variable. Moreover, the OR increased from 2.06 to 2.97 when the cohort was restricted to mothers with a normal pre-pregnancy BMI of < 25 kg/m2 (Table 3). These findings argue against, but do not rule out, residual confounding by fetal size or maternal BMI as an explanation for our results.
The 22.1% caesarean section rate among our study mothers was similar to U.S. national rates reported for 1999 to 2002.39 From 1996 to 2007, the number of caesarean births in the U.S. increased to 32% of births.10,40 One study implication is that further delineation of mechanisms explaining how caesarean section may lead to increased obesity could help with the design of targeted obesity prevention strategies. Another study implication is that prevention of child obesity may be another reason to avoid caesarean section on maternal request, which is estimated to compose between 4 and 18% of caesarean births.11 A mother who chooses caesarean delivery on maternal request should be aware of potential health risks to her and her baby, including child obesity and other potential long-term risks.2,3,32,41–43
In this study, infants delivered by caesarean section had 2-fold higher odds of childhood obesity, even after adjusting for maternal BMI, birth weight, and other confounding variables. Further studies are needed to confirm our findings and to explore mechanisms underlying this association. Expectant mothers choosing caesarean delivery in the absence of an obstetrical or medical indication should be aware that their children may have a higher risk of obesity.
Delivery by caesarean section has been identified as a risk factor for child asthma, but data on child obesity are limited.
An association of caesarean section with child obesity would provide an important additional rationale to avoid non-medically indicated caesarean section.
Infants delivered by caesarean section had 2-fold higher odds of childhood obesity, even after adjusting for maternal BMI, birth weight, and other confounding variables. Expectant mothers choosing caesarean delivery in the absence of a medical indication should be aware that their children may have a higher risk of obesity.
We thank the staff and participants of Project Viva.
This work was supported by NIH grants R01HD034568, R01HD064925 and K24HL068041. The sponsors of the study had no role in the study design, recruitment of participants, data collection, data analysis, data interpretation, writing of the report, or the decision to submit for publication.
All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare that none of the authors or their spouses, partners or children have any financial or non-financial interests that may be relevant to the submitted work.
The Corresponding Author has the right to grant on behalf of all authors and does grant on behalf of all authors, an exclusive licence (or non exclusive for government employees) on a worldwide basis to the BMJ Publishing Group Ltd and its Licensees to permit this article (if accepted) to be published in BMJ editions and any other BMJPGL products and sublicences to exploit all subsidiary rights, as set out in our licence (http://resources.bmj.com/bmj/authors/checklists-forms/licence-for-publication)
ContributorsAll authors had full access to all of the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis. SYH designed the research question, led data analysis and wrote the first draft of the manuscript. CAZ participated in data collection. SRS conducted the statistical analysis. MWG participated in the study design, obtained funding, and directed study operations. All authors contributed to data interpretation and to critical revision of the manuscript and approved the final version. SYH is the guarantor.
Presented in part at the 2010 Obesity Society Annual Meeting, San Diego, CA.
Susanna Y Huh, Pediatrics, Harvard Medical School, and the Division of Gastroenterology and Nutrition, Children’s Hospital Boston.
Sheryl L Rifas-Shiman, Obesity Prevention Program, Department of Population Medicine, Harvard Pilgrim Health Care Institute and Harvard Medical School, Boston, MA.
Chloe A Zera, Obstetrics, Gynecology and Reproductive Biology, Division of Maternal-Fetal Medicine, Brigham and Women’s Hospital, Boston, MA.
Janet W Rich Edwards, Connors Center for Women’s Health and Gender Biology, Brigham and Women’s Hospital and Harvard Medical School and the Department of Epidemiology, Harvard School of Public Health, Boston, MA.
Emily Oken, Obesity Prevention Program, Department of Population Medicine, Harvard Pilgrim Health Care Institute and Harvard Medical School, Boston, MA.
Scott T Weiss, Channing Laboratory, Brigham and Women’s Hospital and Harvard Medical School.
Matthew W Gillman, Obesity Prevention Program, Department of Population Medicine, Harvard Pilgrim Health Care Institute and Harvard Medical School, Boston, MA. Department of Nutrition, Harvard School of Public Health, Boston, MA.