Associations between fruit juice and milk consumption and change in BMI in a large prospective cohort of US adolescents and preadolescents

Abstract

Background:

There are concerns that fruit juice and milk contribute to childhood obesity.

Objective:

Determine the relationship between fruit juice and milk intakes and body mass index (BMI) change among preadolescents/adolescents.

Methods:

Participants aged 9–16 years old from the Growing Up Today Study II completed surveys including validated food frequency questionnaires in 2004, 2006 and 2008. The contributions of one serving of juice or milk to total energy intake and 2-year change in BMI were evaluated using multiple linear regression. Additional analyses were conducted with subgroups of juice (orange juice and other fruit juice) and milk (low fat and high fat). Missing values for BMI were imputed using a multiple imputation approach, after which data from 8,173 participants and 13,717 2-year interval observations were analyzed.

Results:

Baseline fruit juice consumption was inversely associated with BMI change in girls (β=−0.102 kg/m², SE=0.038, P-value=0.008) but not boys after controlling for race, age, baseline BMI, and baseline and 2-year changes in total energy intake and physical activity. Orange juice was inversely associated with BMI change among girls (β=−0.137 kg/m², SE=0.053, P-value=0.010) while other fruit juice, low fat and high fat milk were not associated with BMI change.

Conclusion:

Orange juice was inversely associated with 2-year BMI change among preadolescent/adolescent girls but not boys and there were no significant associations with other juices or milk among either gender.

Introduction

Fruit juice and milk are nutrient-rich beverages that are commonly consumed among children and adolescents, with 21.6% regularly consuming juice and 37.5% regularly consuming milk¹. One hundred precent fruit juice contains abundant amounts of vitamin C, folate, potassium, B vitamins and antioxidants, in addition to calcium and vitamin D in the case of fortified orange juice (OJ). Milk is rich in protein, calcium, vitamin D, vitamin B12, potassium and other minerals. Together, fruit juice and milk can contribute to overall fruit and dairy consumption and help achieve adequacy in important nutrients of concern².

However, there are concerns that fruit juice and milk contribute to childhood obesity, a national epidemic wherein 18.5% of US children are affected with obesity³. Authorities such as the American Academy of Pediatrics and the U.S. Department of Agriculture recommend limiting juice intake to 8 fl oz. per day⁴ or contributing to less than half of one’s daily total fruit intake. U.S. Dietary guidelines² recommend that fruit should be consumed either as whole fruit—including fresh, canned, frozen or dried—or as 100% fruit juice, although warn that juice contributes less dietary fiber than whole fruit and can contribute excess calories if consumed in excess. In an effort to moderate calorie and saturated fat intake, these authorities also recommend consuming low fat dairy in favor of high fat dairy². These guidelines are built on evidence suggesting that sugar and fat play a role in the development of obesity and other related chronic diseases^5,6. However, evidence regarding the contribution of juice^7,8 and milk⁹ intakes to obesity has been inconclusive, which may be partly due to methodologic differences such as assessing total fruit juice^10,11 or dairy¹² rather than specifically assessing sub-types of juice and milk. Furthermore, evidence has indicated possible differences in adiposity by types of juice¹³. OJ is abundant in vitamin C and the flavanones hesperidin and naringenin, which have been posited to reduce the risk of obesity^14,15. Apple juice and other non-citrus fruit juices, on the other hand, are a poor source of flavanones and vitamin C. Therefore, in order to ascertain public health guidelines for beverage intake in youth, clarification is needed. The objective of this study was to determine whether intakes of fruit juice and milk, both in total and by sub-types, were associated with body mass index (BMI) change in preadolescents and adolescents. Furthermore, due to gender-based differences in expected BMI change from normal growth¹⁶ and BMI change reported in a previous study conducted in a similar cohort¹⁰, gender-specific effects were evaluated.

Methods

Study population

The Growing Up Today Study II (GUTS II) is an ongoing prospective cohort study of US children and adolescents who were offspring of the participants in the Nurses’ Health Study II (NHS II). Participants were recruited from all 50 US states and were 9–16 years old at enrollment which began in 2004. An initial questionnaire detailing demographic, social, behavioral, dietary and health-related characteristics at baseline was mailed to participants, and follow-up surveys were sent in 2006 and 2008. Participants’ mothers provided informed consent and the participants assented by completing the initial questionnaire. For inclusion in this study, participants must have completed the baseline questionnaire administered in 2004 in addition to the follow-up questionnaire in 2006. Additionally, data from the 2008 questionnaire were analyzed if available. Participants were excluded for missing/incomplete data on key demographic, anthropometric, behavioral and dietary characteristics including age, race, physical activity, total energy intake, and the intakes of orange juice, other fruit juice and milk (n=2,589). Participants were also excluded for implausible energy intake (<500 or >5,000 kcal/day or a change in total daily energy intake in excess of 2,000 kcal between survey years) (n=92) and extremely low weight (BMI <12 kg/m²) (n=64)¹⁷. After exclusions, the final analytic cohort consisted of 8,173 participants (3,605 boys and 4,568 girls).

Assessment of anthropometric variables

The primary outcome measurement was BMI (kg/m²), which was calculated from self-reported height and weight. Compared to change in BMI percentile or BMI z-score, change in BMI has less within-child variability in children and adolescents^16,18. Specific measuring instructions indicated to stand up straight against the wall with feet flat on the floor without shoes or hats, and participants were encouraged to seek assistance with measurements. Weight status was classified using the Centers for Disease Control and Prevention growth charts¹⁹ and BMI percentile cutoffs²⁰. Missing BMI values were imputed using a multiple imputation approach performed previously in this cohort²¹, utilizing the SAS procedure PROC MI (SAS Institute Inc. SAS/STAT® 9.2 User’s Guide. Cary, NC: SAS Institute Inc., 2008). Briefly, imputed BMI values were based off predictor variables including valid BMI values assessed at other follow-up questionnaires, age, gender, diet, physical activity, sedentary time and total energy intake at baseline and during follow-up. The validity of imputations was demonstrated in a previous study²¹. BMI values between those with measured BMI and imputed BMI were not significantly different. After BMI imputation, data from 8,173 participants (3,605 boys and 4,568 girls) – from which 13,717 2-year observations were created – were used for analysis.

Assessment of dietary variables

Dietary data were collected using the Youth/Adolescent Questionnaire, a semi-quantitative food frequency questionnaire (FFQ) specifically developed for and validated in children and adolescents²³. The FFQ asked participants to quantify their dietary intake over the past 12 months. For fruit juice, there were distinct questions for OJ and apple/other fruit juices. These were considered both individually and collectively as fruit juice (sum of OJ and apple/other fruit juices). There were separate questions for fruit juice and for sugar-added beverages such as Hawaiian Punch, lemonade, Koolaid or other non-carbonated fruit drinks, which is likely to reduce the chance of misclassification of fruit juice due to the additional options. Milk consumption, either as a beverage or consumed with cereal, was evaluated by selecting the type of milk consumed followed by describing frequency of consumption. Options for milk type included whole, 2%, 1%, skim/non-fat, soy milk, “don’t know,” and “don’t drink milk.” In this study, only whole, 2%, 1% and skim/non-fat milk were considered and re-categorized: skim/non-fat and 1% milk were considered as low fat milk and whole and 2% milk were considered as high fat milk. Intakes of each milk type were then estimated by grouping the milk type by the frequency of consumption.

Our analysis was limited to white cow milk as it is the predominant form of milk consumed by children/adolescents²⁴. Consequently, neither chocolate milk nor plant-based milk such as almond or soy milk were included in the analysis. The consumption of plant-based milk in this cohort was very low (data not shown), and together with the large diversity in types of plant-based milk (e.g. soy, almond, rice, oat, coconut, flavored vs. unflavored), it would be difficult to consolidate a singular analysis for plant-based milk in this study given the constraints of the FFQ. Other dairy-based foods such as milkshakes, ice cream and cream-based soups were not considered. The response options were converted to mean daily intake (servings/day). When a range of intakes was selected, the midpoint value was used (e.g. “1–3 glasses per month” was considered to be 2 glasses/month and consequently 0.067 glasses/day). The lowest possible frequency (e.g. “Never/<1 glass per month”) was considered to be 0 servings, and the highest possible frequency was considered to be the lowest possible frequency beyond the given option (e.g. “≥1 glass per day” was considered as 2 glasses/day).

Assessment of physical activity

Moderate/vigorous physical activity was calculated by first summing the number of hours participating in each sport or activity such as soccer, running, mowing the lawn, gardening or strength training that qualified as ≥ 4 metabolic equivalent of task (MET)²⁵, multiplying the number of hours by each sport/activity’s MET score, then summing the values to produce total MET-hours/week.

Statistical analysis

Statistical analyses were conducted using SAS® software, version 9.4 (SAS Institute, Inc., Cary, NC, USA). To determine the contribution of milk and fruit juice to total energy intake in this cohort, the consumption of each beverage was related to its calories per serving information and presented as absolute (kcal/day) and relative (percent of total energy intake) caloric contribution. For the calculation of contribution of beverages to total energy intake, one serving was assumed to be 8 fl oz. and 10 fl oz. for milk and fruit juice, respectively, based on the mean portion sizes consumed by 6–11 and 12–19 year old children and adolescents in the US²⁶.

To determine whether the consumption of milk or fruit juice was associated with BMI change, baseline intakes of milk and fruit juice were included in a multiple regression model with 2-year change in BMI as the outcome. This model was used because beverage intake acts as a predictor of BMI change, allowing for a simple and practical interpretation: the amount of BMI change expected from drinking one daily serving of fruit juice or milk at baseline. We used the SAS procedure proc genmod with repeated measurements using the participant identifier since each participant could potentially have up to two 2-year intervals (2004 to 2006 and 2006 to 2008) and therefore not all observations were independent. The fully adjusted model included age, race, baseline BMI, baseline and 2-year change in total energy intake, and baseline and 2-year change in physical activity. There were no meaningful differences observed when total energy intake was removed in a sensitivity analysis. Results were stratified by gender as well as baseline weight status. Significance testing was used to determine whether fruit juice and milk contributed to 2-year change in BMI and P-values<0.05 were considered statistically significant. Interaction tests were performed according to gender using cross-product terms (gender*dietary components). To account for the inflated family-wise type 1 error rate in the gender interaction model given the multitude of terms, we applied the Holm-Bonferroni correction method for significance testing.

A sensitivity test was performed between the imputed BMI group (n=677; 240 boys, 437 girls) and the group with measured BMI to determine potential bias in baseline characteristics. The non-parametric Wilcoxon two-sample test was used to determine significance in differences. The imputed group was slightly younger (mean ± SD, boys imputed vs. non-imputed: 12.8 ± 1.8 vs. 13.3 ± 1.8 years old, P-value=0.0001; girls imputed vs. non-imputed: 12.9 ± 1.7 vs. 13.5 ± 1.8 years old, P-value<0.0001), had a lower proportion of non-Hispanic whites (boys imputed vs. non-imputed: 85% vs. 96%; girls imputed vs. non-imputed: 91% vs. 96%, P-value<0.0001 for both genders), and were less physically active for girls (mean ± SD, imputed vs. non-imputed: 63.0 ± 49.8 vs. 69.4 ± 51.4, P-value=0.0020). All analyses were additionally conducted with only those with measured BMI values, and there were no differences in the overall magnitude and direction of results.

Results

Baseline characteristics

The unadjusted mean values of the GUTS II participants’ baseline characteristics are presented in Table 1. The average age was approximately 13 years old and the majority of participants were non-Hispanic white (>95%). On average, participants consumed approximately 2/3^rd of a serving per day of fruit juice and 1-½ servings per day of milk. Fruit juice consumption was common, with approximately 75% of participants reporting consuming OJ and 75% consuming other fruit juices at least monthly. Similarly, milk consumption was common, with 95% of participants having reported consuming milk at least monthly (data not shown). Low fat milk was the most commonly consumed (59–63%) milk type, followed by high fat milk (36–40%).

Table 1.

Baseline characteristics of the Growing Up Today Study II participants in 2004

Characteristics	Boys (n=3,605)		Girls (n=4,568)
	Mean or %	SD	Mean or %	SD
Age (y)	13.28	1.81	13.45	1.81
Non-Hispanic white (%)	95.45		95.97
BMI1 (kg/m2)	20.33	3.85	20.16	3.66
Total energy intake (kcal/day)	2324.24	724.20	1953.93	638.99
Fruit juice2 (svgs/day)	0.67	0.73	0.60	0.65
Orange juice (svgs/day)	0.38	0.49	0.31	0.43
Other fruit juice (svgs/day)	0.29	0.44	0.29	0.41
Milk (svgs/day)	1.69	1.19	1.42	1.14
Milk type (%)
Low fat3	59.31		63.22
High fat4	40.69		36.78
Physical activity5 (MET-hrs/week)	91.99	66.02	70.35	52.13

¹Mean and standard deviation calculated by multiple imputation approach.

²Includes orange juice and other fruit juice.

³Includes skim and 1% milk.

⁴Includes 2% and whole milk.

⁵Physical activity was calculated by summing the hours per week spent participating in various sports, games and intentional exercise of moderate or vigorous intensity (MET ≥ 4) and multiplying by the MET score.

Abbreviations: BMI, body mass index; MET, metabolic equivalent of tasks.

Contribution of fruit juice and milk intake to total energy intake

Table 2 displays the caloric contribution of fruit juice and milk. In boys and girls milk contributed modestly to total energy intake (mean ± SD: 183.7 ± 125.6 kcal/day or 8.2 ± 5.8% of total energy for boys or 154.9 ± 117.6 kcal/day (8.1 ± 6.3%) for girls), with low fat milk contributing slightly less than high fat milk to total energy for both boys and girls. The caloric contribution to total energy for fruit juice was approximately half of that of milk (4.0 ± 4.2% for boys, 4.2 ± 4.3% for girls), and the caloric contributions for each type of fruit juice was approximately 40–50 kcal/day.

Table 2.

Contribution of fruit juice and milk (per daily serving) to total energy intake (kcal) in Growing Up Today Study II participants at baseline in 2004

		Boys		Girls
		Contribution to total energy intake (mean ± SD)		Contribution to total energy intake (mean ± SD)
	Kcal/serving1	Kcal/day	%	Kcal/day	%
Fruit juice2	142.8	95.6 ± 104.4	4.0 ± 4.1	85.7 ± 93.0	4.2 ± 4.3
Orange juice	139.8	53.1 ± 67.8	2.3 ± 2.8	43.9 ± 59.8	2.2 ± 3.0
Other fruit juice	145.8	42.5 ± 63.9	1.8 ± 2.6	41.7 ± 59.7	2.0 ± 2.8
Milk3	113.5	183.7 ± 125.6	8.2 ± 5.8	154.9 ± 117.6	8.1 ± 6.3
Low fat milk4	92.9	173.9 ± 107.1	7.9 ± 5.3	145.1 ± 103.2	7.9 ± 5.9
High fat milk5	134.2	198.1 ± 147.4	8.6 ± 6.5	172.3 ± 137.8	8.6 ± 6.9

¹Servings sizes for fruit juice and milk are 10 fl oz. and 8 fl oz., respectively.

²Fruit juice consists of orange juice and other fruit juices and the kcal/serving value is an average of the sub-types’ kcal/serving values.

³Milk consists of low fat and high fat milk and the kcal/serving value is an average of the sub-types’ kcal/serving values.

⁴Low fat milk consists of skim and 1% milk and the kcal/serving value is an average of their kcal/serving values (skim: 83.3 kcal/serving, 1%: 102.5 kcal/serving).

⁵High fat milk consists of 2% and whole milk and the kcal/serving value is an average of their kcal/serving values (2%: 122.0 kcal/serving, whole: 146.4 kcal/serving).

Abbreviations: SD, standard deviation.

Association between fruit juice and milk intake and 2-year change in BMI

Unadjusted baseline intakes of fruit juice and milk were related to 2-year change in BMI. After adjusting for potential confounders, one daily serving of fruit juice was negatively associated with BMI change in girls (β=−0.102 kg/m², SE=0.038, P-value=0.008) but not in boys (β=−0.020 kg/m², SE=0.038, P-value=0.592; P-value for gender interaction=0.146), while one daily serving of milk was not associated with BMI change in either gender (Table 3). The negative association was primarily attributed to OJ while other juices, low fat and high fat milk were not significantly associated with BMI change (Table 4). Baseline and 2-year changes in physical activity and total energy intake were not significant factors in the regression models in either the juice/milk or juice/milk sub-types analyses (data not shown). There was no significant gender interaction for OJ when using the Holm-Bonferroni-corrected α. Stratifying the results by baseline weight status did not reveal any significant observations; however, the magnitude of association between OJ intake and BMI change was nearly four times greater in girls with overweight/obesity (β=−0.388, SE: 0.206, P-value=0.060) compared to girls with normal weight (β=−0.105, SE: 0.054, P-value=0.052) (Table 5).

Table 3.

Average 2-year change in BMI in relation to baseline fruit juice and milk intakes (per daily serving) in Growing Up Today Study II participants (2004–2008)

	Boys (n=5,968 observations)						Girls (n=7,749 observations)
Beverages	Crude model			Fully adjusted			Crude model			Fully adjusted			P for gender interaction
	β	SE	P	β	SE	P	β	SE	P	β	SE	P
Fruit juice1	0.026	0.038	0.487	−0.020	0.038	0.592	−0.072	0.037	0.057	−0.102	0.038	0.008	0.146
Milk	0.041	0.023	0.071	0.027	0.022	0.229	0.029	0.020	0.137	−0.007	0.019	0.723	0.255

¹Includes orange juice and other fruit juice. Standard error was calculated from participants with complete BMI only. Baseline intakes of fruit juice and milk were included simultaneously in a multiple mixed regression model with 2-year BMI change as the outcome. Fully adjusted model adjusts for race, age, baseline BMI, baseline and 2-year change in total energy intake, and baseline and 2-year change in physical activity. Interaction tests were performed according to gender, using cross-product terms (gender*dietary components). Significance testing assesses whether the beverages contribute to 2-year change in BMI.

Abbreviations: BMI: body mass index.

Table 4.

Average 2-year change in BMI in relation to baseline fruit juice and milk sub-types (per daily serving) in Growing Up Today Study II participants (2004–2008)

	Boys (n=6,022 observations)						Girls (n=7,824 observations)
Beverages	Crude model			Fully adjusted			Crude model			Fully adjusted			P for gender interaction
	β	SE	P	β	SE	P	β	SE	P	β	SE	P
Orange juice	0.053	0.058	0.361	0.028	0.057	0.615	−0.128	0.054	0.017	−0.137	0.053	0.010	0.034
Other fruit juice	−0.005	0.069	0.947	−0.084	0.068	0.219	−0.009	0.067	0.894	−0.065	0.064	0.316	0.856
Low fat milk1	0.032	0.023	0.167	0.026	0.023	0.257	0.025	0.021	0.250	−0.001	0.020	0.946	0.379
High fat milk2	0.034	0.030	0.262	0.000	0.030	0.990	0.036	0.024	0.144	−0.016	0.024	0.499	0.679

¹Includes skim and 1% milk.

²Includes whole and 2% milk. Standard error was calculated from participants with complete BMI only. Baseline intakes of fruit juice and milk were included simultaneously in a multiple mixed regression model with 2-year BMI change as the outcome. Fully adjusted model adjusts for race, age, baseline BMI, baseline and 2-year change in total energy intake, and baseline and 2-year change in physical activity. Interaction tests were performed according to gender, using cross-product terms (gender*dietary components). Significance testing assesses whether the beverages contribute to 2-year change in BMI.

Abbreviations: BMI: body mass index.

Table 5.

Average 2-year change in BMI in relation to baseline fruit juice and milk sub-types (per daily serving) in Growing Up Today Study II participants, stratified by baseline weight status (2004–2008)

	Boys						Girls
Beverages	Normal weight (n=4,082 observations)			Obese/overweight (n=1,434 observations)			Normal weight (n=5,844 observations)			Obese/overweight (n=1,249 observations)
	β	SE	P	β	SE	P	β	SE	P	β	SE	P
Orange juice	0.079	0.057	0.169	−0.207	0.182	0.255	−0.105	0.054	0.052	−0.388	0.206	0.060
Other fruit juice	−0.073	0.063	0.247	0.019	0.199	0.924	−0.047	0.059	0.420	−0.154	0.215	0.472
Low fat milk1	0.002	0.024	0.941	0.000	0.062	1.000	0.006	0.019	0.753	0.009	0.078	0.906
High fat milk2	−0.012	0.028	0.666	0.004	0.096	0.965	−0.016	0.024	0.501	0.081	0.115	0.480

¹Includes skim and 1% milk.

²Includes whole and 2% milk. Weight status was classified among participants with complete BMI data using cutoffs defined by the U.S. Centers for Disease Control and Prevention (normal: 5^th≤BMI percentile<85; overweight/obese: BMI percentile≥85^th). Standard error was calculated from participants with complete BMI only. Baseline intake of fruit juice and milk were included simultaneously in a multiple mixed regression model with 2-year BMI change as the outcome. Adjusted for race, age, baseline BMI, baseline and 2-year change in total energy intake, and baseline and 2-year change in physical activity. Significance testing assesses whether the beverages contribute to 2-year change in BMI.

Abbreviations: BMI: body mass index.

Discussion

In this longitudinal analysis evaluating the association between fruit juice and milk intake and BMI change in preadolescents and adolescents over two years, consumption of fruit juice was inversely associated with BMI change in girls. In particular, this association was evident only with OJ, although the direction of BMI change was similarly (but non-significantly) negative for other fruit juices. On the other hand, neither high fat nor low fat milk were associated with BMI change. Furthermore, intakes of either fruit juice or milk were not associated with BMI change in boys, suggesting potential gender-based differences in anthropomorphic responses to fruit juice and milk intake.

Longitudinal studies have generally shown a positive association between 100% fruit juice intake and adiposity in younger children but not in older children^27,28. Fruit juice intake was not associated with excess weight in older children and adolescents in GUTS I participants¹⁷ and in Australian schoolchildren²⁹, although sub-types of juice were not assessed, which potentially masks the simple effects of each juice type. One longitudinal study that specifically assessed OJ and apple juice reported no change in BMI over 5 years in adolescents for either juice²⁴. Apart from that study, however, evaluations of OJ in longitudinal cohort studies are sparse and data regarding OJ and adiposity are mostly derived from cross-sectional studies. Cross-sectional analyses conducted in children/adolescents determined that OJ consumption was either not associated or inversely associated with BMI or weight status^30–33. However, total energy intake^24,30–32 and physical activity^24,32,34 have consistently been shown to be greater in those consuming more OJ, suggesting that OJ consumers may be drinking more OJ in order to meet greater metabolic demands but not drinking in excess. As the current study did not measure energy expenditure beyond moderate/vigorous physical activity we are unable to explicitly determine whether this is indeed the case. However, in the present study each serving of fruit juice was associated with a relatively large number of additional calories but not excess weight gain; furthermore, adjusting for the variables directly contributing to energy expenditure including age, physical activity, total energy intake and BMI in the fully adjusted model did not attenuate the association between OJ intake and change in BMI. Together, this suggests that OJ consumption is linked with other behavioral or physiologic characteristics implicated in healthy weight management, and that the true relationship between juice consumption and excess weight gain is complex and extends beyond simply total energy balance. OJ may possess unique benefits over other fruit juices as it has been found to reduce postprandial inflammation and oxidative stress in clinical trials^35,36, likely due to its higher quantities of antioxidants and the flavonoids hesperidin and naringenin. The results from our analysis should be interpreted with caution, however. Baseline intake was used to model a prediction of BMI change; i.e., estimated the amount of BMI change over 2 years that would be associated with each serving of juice consumed at baseline. This method provides a simple clinical interpretation but is limited by the fact that intake and outcome measurements were taken 2 years apart. Furthermore, we were unable to control for some potential confounders such as socioeconomic status. Hypothetically, consumption of any caloric beverage, even a nutrient dense beverage such as juice, contributes to overall caloric intake which can lead to accumulation of fat stores when the caloric balance is weighted towards energy intake. The preponderance of evidence from observational studies, most of which is derived from having measured realistic and reasonable intakes of juice, therefore supports consumption of fruit juice in moderation. Additionally, as our analysis indicated a potential effect modification of OJ intake and BMI change by weight status, further analyses are needed to elucidate possible mechanistic differences in BMI change in children with overweight/obesity.

Longitudinal studies investigating the relationship between total milk intake and excess weight gain have similarly found a lack of an association^24,37–39. However, while there was no significant difference in 2-year BMI change by milk fat content in the current study, others have observed that low fat milk intake is associated with weight gain while high fat milk is not^40–43. Although, in GUTS I participants, both skim and 1% milks were associated with an increase in BMI compared to other types of milk in boys but only skim milk was associated with BMI change in girls despite a similar prevalence (25–26%) of 1% milk consumption⁴⁰. This suggests that excess weight gain is not universally attributed to low fat milk consumption and that gender may be an important mediator. Excess weight gain due to low fat milk intake may be partly due to reverse causation wherein children at risk of being overweight or obese consume low fat dairy in an effort to reduce caloric intake. Another possible explanation is that the fat content in high fat milk increases satiety by stimulating the release of cholecystokinin, leading to a reduction in overall food/beverage consumption⁴⁴. Intervention trials comparing the effects of low fat and high fat milk on adiposity in a free-living and calorically unrestricted setting are needed to clarify whether differences based on milk fat content exist.

There are two major implications of this study. First, in instances where children are failing to consume adequate amounts of whole fruit and certain micronutrients such as vitamin C, potassium and folate, consuming a moderate amount of fruit juice may help to improve intakes of these key nutrients while allaying concerns of weight gain. However, whole fruit should still be the preferred choice and fruit juice intake should be kept within recommended limits. Second, high fat milk may be a useful alternative to low fat milk when dairy intake and the consumption of calcium, vitamin D and potassium are inadequate. Although, it should be noted that high fat milk still contains substantially more calories and saturated fat than low fat milk and thus moderation should be encouraged.

This study has several strengths, including the use of both juice and milk together in the multiple regression analysis which reduces the effect of confounding by one another and therefore providing a more accurate assessment. Additionally, imputing missing BMI values reduces potential bias by including in the analysis those that may have different characteristics that are typically associated with omitting self-reported height/weight. Finally, the sample size of the cohort was large and further increased by using multiple observations of 2-year intervals for each participant, increasing the confidence of estimates. However, there are also limitations worth noting. First, the milk type question in the FFQ only allowed for one milk type to be selected and therefore does not allow for the possibility for a participant to indicate whether he/she consumes multiple types. Consequently, it had to be assumed that each participant exclusively consumed one type of milk during the selected survey interval. Second, fruit juice was not explicitly defined, giving the possibility of misclassification bias. However, participants were given the opportunity to indicate consumption of other sugar-sweetened juice drinks (i.e., fruit-flavored drinks) in the FFQ, thus our analysis is likely to have only included 100% fruit juice. Third, we were unable to control for socioeconomic variables since the questionnaires were targeted for the young participants. Since socioeconomic status may potentially influence beverage intake as well as BMI status it is uncertain whether fruit juice was independently associated with 2-year BMI change. Fourth, self-reports of height and weight are potentially subject to certain biases. However, they have previously been validated in the adolescent/pre-adolescent population^45–47. Furthermore, since the primary outcome was change in BMI, any instances of under- or over-reporting would likely be non-differential to time and would therefore be more valid than if the outcome were a single measurement of BMI. Finally, the GUTS II cohort was not selected to be nationally representative of the US population and therefore potentially limits generalizability, although attempts were made to increase generalizability by purposefully enrolling participants from all 50 US states.

In conclusion, baseline OJ consumption was inversely associated with 2-year change in BMI among girls. There were no significant associations between fruit juice or milk and BMI change in boys. These results suggest that OJ is unlikely to contribute to childhood obesity but future studies assessing intake and BMI over the same time period and controlling for additional potential confounders are needed to confirm these findings.

Abbreviations:

OJ

orange juice

BMI

body mass index

GUTS

Growing Up Today Study

NHS II

Nurses’ Health Study II

FFQ

food frequency questionnaire

MET

metabolic equivalent of task

SE

standard error