The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Objectives: The increase in consumption of sugar-added beverages over recent decades may be partly responsible for the obesity epidemic among U.S. adolescents. Our aim was to evaluate the relationship between BMI changes and intakes of sugar-added beverages, milk, fruit juices, and diet soda.
Research Methods and Procedures: Our prospective cohort study included >10, 000 boys and girls participating in the U.S. Growing Up Today Study. The participants were 9 to 14 years old in 1996 and completed questionnaires in 1996, 1997, and 1998. We analyzed change in BMI (kilograms per meter squared) over two 1-year periods among children who completed annual food frequency questionnaires assessing typical past year intakes. We studied beverage intakes during the year corresponding to each BMI change, and in separate models, we studied 1-year changes in beverage intakes, adjusting for prior year intakes. Models included all beverages simultaneously; further models adjusted for total energy intake.
Results: Consumption of sugar-added beverages was associated with small BMI gains during the corresponding year (boys: +0.03 kg/m2 per daily serving, p = 0.04; girls: +0.02 kg/m2, p = 0.096). In models not assuming a linear dose-response trend, girls who drank 1 serving/d of sugar-added beverages gained more weight (+0.068, p = 0.02) than girls drinking none, as did girls drinking 2 servings/d (+0.09, p = 0.06) or 3+ servings/d (+0.08, p = 0.06). Analyses of year-to-year change in beverage intakes provided generally similar findings; boys who increased consumption of sugar-added beverages from the prior year experienced weight gain (+0.04 kg/m2 per additional daily serving, p = 0.01). Children who increased intakes by 2 or more servings/d from the prior year gained weight (boys: +0.14, p = 0.01; girls +0.10, p = 0.046). Further adjusting our models for total energy intake substantially reduced the estimated effects, which were no longer significant.
Discussion: Consumption of sugar-added beverages may contribute to weight gain among adolescents, probably due to their contribution to total energy intake, because adjustment for calories greatly attenuated the estimated associations.
Large increases over recent decades in the prevalence of childhood obesity are well documented (1, 2, 3, 4, 5, 6, 7, 8, 9), as are the associated health and social consequences of obesity (3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29). This rapid increase in obesity prevalence implicates environmental factors (27, 30, 31, 32, 33, 34, 35). During this time, physical activity among adolescents has declined, whereas time spent in sedentary activities such as watching television or videos and playing computer games has increased (5, 6). Furthermore, in nationally representative samples of U.S. adolescents, intakes of sugar-added beverages, including soda, have increased (36, 37, 38, 39). Higher soft drink intakes are associated with lower milk and fruit juice intakes and with higher total energy intakes (40). The largest source of added sugars in the U.S. diet is nondiet soft drinks (37).
One cross-sectional study of dietary intakes (41) has reported similar soda and fruit drink intakes for obese vs. nonobese adolescents, whereas another has found a positive correlation between measures of adiposity in adolescents and soft drink intakes (42). However, two other studies have suggested an inverse association between adiposity and intake of sugars (43, 44).
Ludwig et al. (45) published the first longitudinal analysis of sugar-added beverage intakes and body weight changes. They followed 548 ethnically diverse 11- and 12-year-old children in Boston-area public schools for 19 months and found positive associations among sugar-sweetened beverage intakes, weight change, and incident obesity. Whether the critical factor is the sugar, the calories, or behaviors related to beverage consumption is unknown. Aside from the calories within each beverage, some foods may frequently accompany certain beverages (46), and drinking beverages may also lead to higher subsequent energy intakes because compensation for energy consumed in liquid form is less complete, due to lower satiety, than energy consumed in solid form (45, 47, 48). Furthermore, sugar-added beverages may encourage additional energy intake because of their high glycemic index (49). It may be informative to further consider all beverages simultaneously and to study children from a broader age range and with longer follow-up.
Using data from the Growing Up Today Study, an ongoing prospective cohort study of children from all over the U.S., we analyzed the relationship between intakes of beverages (milk, sugar-added beverages, fruit juices, and diet soda) and changes over time in BMI.
Research Methods and Procedures
Established in the fall of 1996, the Growing Up Today Study consists of 16, 771 children, residing in 50 states, who are offspring of Nurses’ Health Study II (NHSII)1 participants (50). The study is described in detail elsewhere (51). These children were ages 9 through 14 years old in 1996. In 1997 and 1998, we sent subjects follow-up questionnaires to update all information. Response rates to at least one of these follow-ups were 92.5% for girls and 87.7% for boys.
Children self-reported their height and weight annually on our questionnaire, which provided specific measuring instructions but suggested that they ask someone for help. Because their mothers are nurses who biennially self-report their own height and weight as part of NHSII, assistance was available to each of them. A previous study reported high validity for self-reported heights and weights for children 12 to 16 years old (52). We assessed adiposity by computing BMI = weight/(height)2 squared (kilograms per meter squared). The International Obesity Task Force supports the use of BMI to assess fatness in children and adolescents (53). Childhood BMI is related to other measures of adiposity that were not feasible to include in our large self-report study (54, 55). A recent study (56) supports the validity of BMI computed from self-reported height and weight, with a correlation of 0.92 between BMI computed from measured values and self-reports by youth in grades 7 through 12.
Before computing BMIs, we excluded any height that was >3 SD beyond the gender-age-specific mean height (0.46% of heights excluded) and any 1-year height change which declined by >1 inch or increased by >3 SD above the mean change (1.65% excluded). We further excluded any BMI < 12.0 kg/m2 as a biological lower limit (clinical opinion, MWG) and BMI > 3 SD above or below the gender-age-specific mean [log(BMI) scale] (0.87% excluded). We then estimated our outcome, annual change in adiposity, by BMI1997 − BMI1996 and BMI1998 − BMI1997, dividing each by the exact time interval between the pair of measurements. Because all these represented realistic 1-year changes in weight, there were no further BMI exclusions: 6871 girls and 5321 boys provided BMI change data. Unfortunately, no validation studies of change in BMI, derived from self-reported data, have been conducted.
We grouped children, based on their BMI at the earlier year of each 1-year time interval, using the Centers for Disease Control and Prevention (CDC) gender- and age-specific percentiles for BMI (57). Children above the 85th percentile were at risk of overweight (85th to 95th percentile), and those above the 95th percentile were overweight (57). Similarly, we grouped together those below the 10th percentile for BMI. For simplicity, we refer to all children whose BMI exceeded the 85th percentile as “overweight,” those below the 10th percentile as “very lean,” and those between the 10th and 85th percentiles as “normal weight.” The CDC standards were also used to assign age-specific z scores to BMIs.
Members of our research group designed a self-administered semiquantitative food frequency questionnaire (FFQ), specifically for older children and adolescents, which is inexpensive and simple to administer to large populations (58). This FFQ for youth has been shown to be valid and reproducible on children 9 through 18 years old (58, 59); the mean correlation for nutrients from the FFQ compared with three 24-hour recalls was r = 0.54, which is comparable with the performance of a similar adult FFQ. The youth FFQ included questions regarding frequency of intake of 132 specific food items over the past year. Beverage questions indicated that the serving size was a can, glass, bottle, or cup (tailored to the particular beverage). The question about “Hawaiian Punch, lemonade, Koolaid, or other noncarbonated fruit drink” preceded questions about “orange juice” and “apple juice and other fruit juices.” For each beverage, we derived typical past year intake (servings per day) and change in intakes between years. We also estimated total energy intake (kilocalories per day) and excluded as implausible intakes <500 or >5000 kcal/d (0.53% excluded).
The beverages we studied were sugar-added beverages (soda, sweetened iced tea, and noncarbonated fruit drinks), fruit juices (orange juice and apple/other juices), diet soda, and milk (white, in a glass or on cereal, and chocolate). Alcohol and coffee intakes were very low; therefore, we did not include them.
We developed a physical activity questionnaire, specifically for youth, which asked the participants to recall the typical amount of time spent, within each season over the past year, in 17 activities and team sports (outside of gym class); response categories ranged from 0 to 10+ h/wk. From each child's responses, we computed his/her typical hours of weekly physical activity within each season and over the entire year. Assessments of an earlier nonseasonal version of this instrument found that estimates of total physical activity were moderately reproducible and reasonably correlated with cardio-respiratory fitness, thus providing evidence of validity (60). Another validation study reported a correlation of r = 0.80 between survey self-reports and 24-hour recalls in sixth to eighth grade children (61). We developed the seasonal version used in this paper to further improve reliability and validity (62). Estimates of total physical activity that exceeded 40 h/wk were deemed implausible and excluded (3.8%).
Another series of questions was designed to measure weekly hours of recreational inactivity: “watching TV,” “watching videos or VCR,” and “Nintendo/Sega/computer games (not homework).” For each of these, children reported their usual number of total hours, separate for weekdays and for weekends, from options ranging from 0 to 31+ hours. From this information, we computed each child's typical hours of recreational inactivity per week. Gortmaker and colleagues (61) reported moderate reproducibility for children in grades six to eight for recalled total inactivity from a similar instrument. We considered totals exceeding 80 h/wk implausible and excluded them (0.94%).
At baseline, children reported their race/ethnic group by marking all of six options that applied. We assigned each child to one of five racial/ethnic groups following U.S. Census definitions, except that we retained Asians as a separate group rather than pooled with “other” (1).
Tanner Stage, Menarche, and Age
Each year, children reported their Tanner maturation stage, a validated self-rating (63) of sexual maturity that uses five categories/illustrations for stage of pubic hair development, and girls reported whether/when their menstrual periods began. We derived a menstrual history variable having three categories: premenarche both before and after the 1-year BMI change, periods that began during the interval, and postmenarche both years. We computed each child's age from dates of birth and questionnaire return.
To assess the potential for selection bias, we compared the baseline (1996) values of age, BMI, individual beverage intakes, and total energy intakes of those children who returned surveys in consecutive follow-up years with those who did not. The differences were small (see “Results”). All models throughout were fit separately for boys and girls.
We reported gender- and age-specific means at baseline for height, weight, total energy intake, and daily intakes of seven beverages. A linear regression model related baseline total daily energy intakes to the intake of each beverage.
To study the effects of beverage intakes during the year of BMI change, we related the past year typical beverage intakes reported in 1997 to change in BMI from 1996 to 1997 and intakes reported in 1998 to BMI change from 1997 to 1998. Because each child can have two BMI changes, the assumption of independent observations required by ordinary regression models was not met, so we used mixed linear regression models (64) with estimation by SAS proc mixed (65).
We also estimated the effects of 1-year change in beverage intakes (the difference between intakes in 1996 and 1997 and between 1997 and 1998) on same-year change in BMI. The prior year intake (reported in 1996 and 1997) was included as a covariate in the mixed model.
All models adjusted for race/ethnicity, and to account for increases in BMI that typically occur during growth and maturation, we included height growth during the same year, menstrual history, Tanner stage, prior BMI z score, and nonlinear age trends (30, 66, 67, 68, 69, 70). Models also adjusted for activity and inactivity during the year of BMI change (51) and for milk type (whole/2%/1%/nonfat/soy). We included total energy intake in further models as a hypothesized intermediary in the pathway between beverages and weight gain.
These children, whose mothers are all participants in the NHSII (50), are mostly white (94.7%). At baseline, 23.2% of the boys and 17.5% of the girls were overweight (>85th percentile on CDC BMI charts), whereas 7.2% of the boys and 8.6% of the girls were very lean (<10th percentile).
Children who did not return surveys in adjacent years (required for inclusion in our longitudinal analyses) were slightly older (girls by 0.20 years; boys by 0.32 years; both p < 0.05). At baseline, they drank slightly less milk (girls by 0.18 servings/d; boys by 0.11 servings/d) but more sugar-added beverages (girls by 0.13 servings/d; boys by 0.10 servings/d) (each age-adjusted p < 0.05). There were no significant differences at baseline in age-adjusted BMI, total energy intake, fruit juice intake, or diet soda intake (each age-adjusted p > 0.05).
Older children drank less milk but more orange juice, soda, iced tea, and punch than younger children (Table 1). Boys reported higher energy intakes and drank more milk, punch, orange juice, and soda than did same-age girls. At baseline, children who drank more milk and less diet soda were leaner, whereas girls who drank more sugar-added beverages were heavier (BMI +0.06 kg/m2 higher per serving, p = 0.04).
Table 1. Baseline means for total energy intakes (kilocalories per day), beverage intakes (servings per day), height (inches), and weight (pounds) for youth participating in the Growing Up Today Study
Milk includes both chocolate and white milk.
To explore whether drinking certain beverages may be linked to higher total energy intakes, we related daily total energy intake to each of the beverages separately (Table 2). As expected, diet soda intakes were not associated with higher total energy intakes. Milk intakes were associated with total energy intakes, with per serving effects slightly more than the energy provided by the milk, whereas the per serving effects for sugar-added beverage and fruit juice intakes were considerably larger than their own energy contents.
Table 2. Cross-sectional association between beverage intakes and total energy intakes at baseline*
Total daily energy increase per daily serving of each beverage
Each beverage was in a separate regression model, which adjusted for age, Tanner stage, menarche (girls), and physical activity and inactivity (TV/videos/video games).
Per serving kilocalories used in deriving total energy:
86 (skim) to 158 (chocolate) kcal/serving.
152 (soda) and 60 (punch) kcal/serving.
84 (orange juice) and 88 (apple juice) kcal/serving.
Among children who completed the FFQ all 3 years, mean milk intake declined significantly each year, whereas soda intake increased significantly (Figure 1). Apple juice intake declined for both boys and girls between 1996 and 1997, diet soda and orange juice intake each increased for girls between 1997 and 1998, and orange juice intake increased for boys each year (all p < 0.05).
Beverage Intakes During Year of BMI Change
We related BMI changes over 1-year periods to beverage intakes during the same year. The regression coefficients β (Table 3) represent the 1-year change in BMI (kilograms per meter squared) expected per usual daily serving of each beverage. For boys, intakes of sugar-added beverages (β = 0.03) and diet soda (β = 0.12) were significantly associated with weight gains; there were suggestions (p < 0.06) that intakes of milk (β = 0.02) and fruit juices (β = 0.04) were also associated. After including total energy intake in the model, the estimated βs for all beverages (except diet soda) were nearly one-half their unadjusted magnitudes and no longer significant (p > 0.31; Table 3). For girls, our analysis suggested (p < 0.10) a linear association between 1-year weight gain and intakes of milk (β = 0.02) and sugar-added beverages (β = 0.02); the corresponding energy-adjusted estimates were slightly smaller (all p > 0.15; Table 3).
Table 3. Longitudinal analysis of beverage intakes and change in BMI (kilograms per meter squared) during the same time period*
1-year change in BMI per daily serving of each beverage during same year
Before energy adjustment
After energy adjustment
All beverages were included simultaneously in each sex-specific mixed model, which was adjusted for age, Tanner stage, race, menarche (girls), prior BMI z score, height growth, milk type, physical activity, and inactivity.
Boys (n = 5067)
Girls (n = 6688)
Figure 2 (far left) presents the association between BMI change and sugar-added beverages analyzed as a categorical variable (0, 1, 2, or 3+ servings/d) to permit nonlinear trends; all Figure 2 models adjusted for all covariates except energy intake. A dose-response trend was confirmed for boys, consistent with the statistically significant per-serving effect (also shown in Figure 2). Girls who reported one (0.5 to 1.5) daily serving of sugar-added beverages gained significantly more BMI (0.068 kg/m2, p = 0.02) during the year than those reporting none (0 to <0.5 servings) (Figure 2, far left). Girls consuming two (+0.09, p = 0.06) or three+ servings (+0.08, p = 0.06) also gained weight compared with nondrinkers.
Beverage Change and BMI Change
For boys, increasing sugar-added beverage intake from one year to the next was significantly associated (β = 0.04 per added daily serving; p = 0.01) with weight gain (Table 4), and increasing milk and diet soda intakes were weakly associated (p < 0.10) with weight gain. With total energy intake (prior year energy and change in energy) in the model, estimates for the energy-containing beverages each declined by >40%, and none remained significant (Table 4). For girls, increasing intake of sugar-added beverages was weakly linearly related to weight gain (β = 0.03, p = 0.08); energy adjustment diminished the estimated effect (p = 0.16).
Table 4. Longitudinal analysis of change in beverage intakes and change in BMI (kilograms per meter squared) over the same year, adjusting for prior beverage intakes*
1-year change in BMI per increase in daily serving of each beverage over same year
Before energy adjustment
After energy adjustment
All beverages were included simultaneously in each sex-specific mixed model, which was adjusted for age, Tanner stage, race, menarche (girls), prior BMI z score, height growth, milk type, physical activity, and inactivity.
Boys (n = 5018)
Girls (n = 6636)
Figure 2 (right half) shows that boys who increased their sugar-added beverage intake by 1 serving/d from the previous year gained more weight (+0.10 kg/m2, p = 0.02) than boys with unchanged intake, and those who increased their intake by 2 or more servings/d gained even more (+0.14, p = 0.01). Girls (Figure 2, right) who increased their intake by 1 serving/d over the previous year gained marginally more BMI (+0.065, p = 0.079) than girls whose intakes were unchanged, and girls whose intakes increased by 2 or more servings/d gained significantly more BMI (+0.10, p = 0.046).
Combining Energy-Containing Beverages
Because the models in Tables 3and 4 suggested that any of the beverages containing calories might contribute to male weight gains, we combined together these beverages (total servings per day of milk, sugar-added beverages, and fruit juices). For boys, this total was associated with weight gain (β = +0.03 kg/m2 per daily serving during the year of BMI change, p = <0.01; and β = +0.03 per increase in daily serving from the prior year, p = <0.01). For girls, because the βs for fruit juice were <0 in Tables 3and 4, combining fruit juice intakes with milk and sugar-added beverages did not provide a significant association (β = +0.01, p = 0.096 per daily serving during the year of BMI change; and β = +0.01, p = 0.13 per increase in daily serving from the prior year). After energy adjustment, significant effects became smaller by at least 31% and were no longer significant (p > 0.12).
Although a previous publication (45) considered whether sugar-added beverages contribute to weight gain among 11- to 12-year-old children, we addressed the effects of several types of beverages on children 9 to 17 years old. Our strongest and most consistent evidence was a linear association between sugar-added beverage intakes (past year and change from prior year) and weight gain in boys (both p < 0.05). The evidence for girls was less compelling but still suggestive (p < 0.10) of a linear association between sugar-added beverages and weight gain. Girls who drank 1 serving/d during the past year gained more weight than nondrinkers (p < 0.05). Both boys and girls who increased their intakes by 2 or more servings/d from the previous year experienced significant weight gain, as did boys who increased their intakes by 1 serving/d from the previous year. However, the magnitudes of these estimated effects were small; a boy consuming 3 servings/d of sugar-added beverages over 10 years is expected to gain only 0.9 BMI more than if he consumed none.
Our finding that drinking diet soda during the past year was associated with weight gain in boys was somewhat unexpected, particularly because a recent double-blind randomized controlled trial of overweight adults compared the effects of sucrose and artificial sweeteners (primarily in beverages) and demonstrated weight loss for the latter group (71). However, our overweight boys were drinking nearly 3 times as much diet soda as our normal weight boys (0.3 vs. 0.1 servings/d) but similar quantities of regular soda (0.6 servings/d for both groups). The correlation between year-to-year changes in regular soda and diet soda intakes was null (r = −0.008; p = 0.70), suggesting that heavier boys were not substituting diet soda for sugared soda. We believe that this explains the diet soda estimate for boys and, furthermore, illustrates the importance of confirming findings using blinded randomized trials whenever feasible and ethical.
Because our estimates became considerably smaller after adjusting for total energy intake, calories probably explain the associations between beverages and weight gain. However, we cannot differentiate between calories in the beverages and calories in foods typically consumed alongside certain beverages (46), or whether drinking beverages leads to higher subsequent energy intakes because compensation for energy consumed in liquid form is less complete than energy consumed in solid form (45, 47, 48). The compensation theory that liquid foods have lower satiety than solid foods would apply to milk and fruit juice as well as to sugar-added beverages. Sugar-added beverages may also encourage further energy intake because of their high glycemic index (49). Under any of these possible mechanisms, consumption of sugar-added beverages encourages higher total energy intakes, which promotes weight gain, so that adjusting models for energy should diminish the estimated associations. The fact that sugar-added beverages lost statistical significance in energy-adjusted models does not imply that sugar-added beverages are not responsible for weight gain because of the pathway.
The literature regarding cross-sectional associations between adiposity and beverage consumption is mixed (41, 42, 43, 44, 72). Our cross-sectional results indicated that heavier children were drinking less milk and more diet soda, presumably to lose weight or prevent further weight gain, although girls who drank sugar-added beverages tended to be heavier. In a nationally representative sample of U.S. children, BMI was positively associated with diet carbonated beverages and, for girls, negatively associated with milk intakes (70). We further presented cross-sectional evidence similar to Harnack et al. (40) that drinking sugar-added beverages was associated with higher total energy intakes.
The first longitudinal study of sugar-sweetened beverages (45), on 548 ethnically diverse 11- and 12-year-old children in Boston-area public schools, reported associations between change in beverage intakes from baseline to 19 months later and BMI change. Their study differed from ours in that they did not have Tanner Stage data, their FFQ and report of activity/inactivity related to past month (ours was past year), and they did not study milk intakes. Their BMIs were measured rather than self-reported, which may partially explain why their estimate for a single serving per day increase [β = +0.20 kg/m2 over 19 months, not energy adjusted; from their Table 2 (45)] is larger than our estimate (over 12 months: boys β = +0.10 kg/m2, p = 0.02; girls +0.07, p = 0.08).
A major strength of our analysis was the longitudinal design, which allowed us to study changes over time in beverage intakes and in BMI while accounting for growth and maturation. BMI typically goes up from year to year among children in this age range, and we took these changes into account. Although our observational study cannot provide conclusive evidence of causality, our evidence is stronger than that obtainable from cross-sectional studies. Baseline differences between children excluded and included in our longitudinal analyses were small, though the loss of some children with higher intakes of sugar-added beverages (0.1 more servings/d) could bias our estimates of those effects. Because we included all beverages together in our models, we minimized confounding by other beverage intakes. However, residual and unmeasured confounding is still possible despite extensive control for many important covariates. A major limitation of our study was the necessity of collecting data (including height, weight, and beverage intakes by FFQ) on youth by self-report on mailed questionnaires, but with our large geographically dispersed cohort, alternatives were not feasible. The impact of random reporting errors should be to bias estimates of true associations toward the null, possibly explaining why our estimates were quite small even when statistically significant. Large soft drink portion sizes complicate the reporting of intakes and encourage overconsumption (35). Data collected by 24-hour recalls from 1994 to 1996 (73) showed that the average soft drink portion size was 19.9 ounces, and differences were noted among eating locations (home, restaurant, and fast food). Our FFQ did suggest portion sizes [“soda, not diet (1 can or glass)”; response category “1 can per day”] but did not specify the number of ounces in a can or glass, so confusion over this may have further biased our estimates toward the null.
Although we cannot claim that our children of nurses are representative of U.S. children, the associations among factors within our cohort should still be internally valid. Our sugar-added beverage intakes for 11- to 12-year-old children (1.35 servings/d for boys and 1.14 for girls) were similar to those of 11- to 12-year-old children studied by Ludwig et al. (1.22 servings/d) (45).
In 1998, Jacobson summarized the history of soft drink consumption, its nutritional value, its potential impact on osteoporosis, tooth decay, heart disease, and kidney stones, and its marketing by the industry, with recommendations for what should be done (74). Here, we extend the evidence (45) that sugar-added beverages (which include soda) may contribute to weight gain. Even if milk and fruit juice also contribute to weight gain, they have nutritional benefits, whereas soda provides only calories (74). The increase in soft drink serving sizes and the increase in energy intakes provided by soft drinks since 1977 have been documented (73, 75, 76), and reversing this trend may help prevent further increases in obesity prevalence.
For both children and adults, prevention of obesity is critical, and for weight loss, recommendations include eating a nutritionally balanced, low-energy diet while increasing energy expenditure through regular physical exercise (77, 78). Beverage intakes, including limiting the consumption of soft drinks, are a potential target for improving diets of adolescents (42, 45, 74, 79). Data from our cohort suggested that children who reduce intakes of sugar-added beverages, along with other behavior modifications such as increasing physical activity and reducing time with TV/videos/computer games (80), may prevent excessive weight gains that can lead to obesity.
This study was funded by NIH Grant DK46834, by Boston Obesity Nutrition Research Center Grant P30 DK46200, by Prevention Research Center Grant U48/CCU115807 from the Centers for Disease Control and Prevention, by Research Grant 43-3AEM-0-80074 from the Economic Research Service of the U.S. Department of Agriculture, and, in part, by Kellogg's. The authors are grateful to Catherine Tomeo Ryan, Karen Corsano, Gary Chase, and Gideon Aweh for ongoing technical support and to all their colleagues in the Growing Up Today Study Research Group. The authors are especially grateful to the children (and their mothers for encouragement) for careful completion of the questionnaires.
Nonstandard abbreviations: NHSII, Nurses’ Health Study II; CDC, Centers for Disease Control and Prevention; FFQ, food frequency questionnaire.