AJCN North Carolina Research Campus
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Data
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Google Scholar
Right arrow Articles by Forshee, R. A
Right arrow Articles by Storey, M. L
PubMed
Right arrow PubMed Citation
Right arrow Articles by Forshee, R. A
Right arrow Articles by Storey, M. L
Agricola
Right arrow Articles by Forshee, R. A
Right arrow Articles by Storey, M. L
American Journal of Clinical Nutrition, Vol. 87, No. 6, 1662-1671, June 2008
© 2008 American Society for Nutrition

ORIGINAL RESEARCH COMMUNICATION

Sugar-sweetened beverages and body mass index in children and adolescents: a meta-analysis1,2,3,4

Richard A Forshee, Patricia A Anderson and Maureen L Storey

1 From the Center for Food, Nutrition, and Agriculture Policy, University of Maryland, College Park, MD

2 The authors retained complete control of the study design, collection of data, analysis of data, and interpretation of results. The research proposal to the sponsor was approved as submitted, but the sponsor requested that an independent expert on meta-analysis—to be chosen by the authors—review the manuscript. Courtesy copies of the manuscript were provided to the sponsor at the time of submission, but the sponsor had no editorial control. One author (MLS) accepted a position with the sponsor after the first decision letter regarding the manuscript was received. The data used in the analysis are available to other researchers for replication and extension of the analysis. The views expressed by the authors are their own and may not represent the views of the University of Maryland.

3 Supported by a grant from the American Beverage Association.

4 Reprints not available. Address correspondence to ML Storey, American Beverage Association, 1101 16th Street, NW, Washington, DC 20036. E-mail: mstorey{at}ameribev.org.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Background: Rates of overweight and obesity have increased. Consumption of sugar-sweetened beverages (SBs) may play a role.

Objective: The purpose of this meta-analysis was to determine whether the results of original research with the use of longitudinal and randomized controlled trials (RCTs) support the hypothesis that SB consumption is associated with weight gain among children and adolescents.

Design: The MEDLINE database was used to retrieve all original studies of SBs and weight gain involving children and adolescents. Twelve (10 longitudinal and 2 RCT) studies were reviewed. Eight of the longitudinal studies and both RCT studies were incorporated into a quantitative meta-analysis. Forest plots and overall estimates and CIs for the association of the difference ({Delta}) in SB consumption with {Delta}body mass index (BMI; in kg/m2) were produced. Funnel plots were examined as a diagnostic test for publication bias. Databases of unpublished scientific studies were searched. Sensitivity tests were conducted to examine the robustness of the meta-analysis results.

Results: The overall estimate of the association was a 0.004 (95% CI: –0.006, 0.014) change in BMI during the time period defined by the study for each serving per day change in SB consumption with the fixed-effects model and 0.017 (95% CI: –0.009, 0.044) with the random-effects model. The funnel plot is consistent with publication bias against studies that do not report statistically significant findings. The sensitivity tests suggest that the results are robust to alternative assumptions and new studies.

Conclusion: The quantitative meta-analysis and qualitative review found that the association between SB consumption and BMI was near zero, based on the current body of scientific evidence.


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Rates of overweight and obesity have increased nationally and globally (1). The causes of obesity are numerous and complex; involving genetic, psychological, social, and environmental components (2). The role of sugar-sweetened beverages (SBs) in the development of overweight and obesity has received scientific and policy attention. Several researchers, advocacy groups, and professional groups opine that SB consumption plays an important role in the development of overweight and obesity and warrants special attention as policies are developed to reduce the overweight and obesity rates (3, 4).

The overall effect of SBs on overweight and obesity depends primarily on 2 factors: the current distribution of consumption of SBs and the magnitude of the effect (if any) of SB consumption on body mass index (BMI; in kg/m2) or other measures of weight status.

In the United States, current average consumption of SBs for adolescents 12-19 y of age are 630 g/d and 409 g/d for males and females, respectively. Average consumption of fruit drinks and ades is 105 g/d and 115 g/d for males and females, respectively. For children 6-11 y of age, average consumption of SBs is 284 g/d for boys and 213 g/d for girls, and the average consumption of fruit drinks and ades is 102 g/d for boys and 96 g/d for girls (5).

The purpose of this meta-analysis was to determine whether the results of original research that used longitudinal and randomized controlled trials (RCTs) support the hypothesis that SB consumption is associated with increased BMI among children and adolescents and, if so, to determine the magnitude of the effect.


   SUBJECTS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
The MEDLINE database (National Library of Medicine, Bethesda, MD) was used to identify relevant English-language articles published between 1966 and October 2006 about studies of the intake of SBs and BMI, obesity, or both. Sugar-sweetened drinks included soft drinks and fruit drinks or ades. On the basis of guidance from the medical subheading (MeSH) terms database, the following search term combinations were used: "beverages obesity," "beverages BMI," "soft drinks obesity," "soft drinks BMI," "sweetened drinks," and "sugar drink fat mass." The searches yielded 127, 70, 22, 67, 112, and 3 articles, respectively.

Selection of core articles was restricted to original research in the English language with human subjects < 19 y of age that examined the association between SBs and weight gain, obesity, or both. Ecologic and cross-sectional designs were excluded. The bibliographies from 5 review articles (6-10) were also checked to ensure that all relevant studies had been captured. A total of 12 (10 longitudinal and 2 RCT) studies were reviewed. Eight of the longitudinal studies and both RCT studies were incorporated into a quantitative meta-analysis. The remaining studies did not provide the necessary information for inclusion in the quantitative meta-analysis, so these studies were reviewed individually.

To minimize possible publication bias, searches were also conducted on databases that contain unpublished scientific studies. Six databases were searched: Computer Retrieval of Information on Scientific Projects, Current Research Information System, Web of Science, ClinicalTrials.gov, BIOSIS Previews, and Proquest Interdisciplinary Dissertations and Theses. E-mail requests for more information were sent to the corresponding author for any eligible study identified in our search of unpublished literature and to any corresponding author whose published study did not provide enough information to be included in the quantitative analysis. An independent expert reviewed an earlier draft of the manuscript and provided constructive criticism and useful advice.

Eight longitudinal (Table 1Go) and 2 RCT studies (Table 2Go) were included in the meta-analysis. Coefficients and standard errors were extracted from the articles and compiled into a statistical database in STATA software, version 9.2 (11). When standard errors were not reported, they were calculated by us, based on reported CIs or P values. The results were analyzed with the use of METAN: STATA module for fixed- and random-effects meta-analysis program (12). METAFUNNEL: Stata module to produce funnel plots for meta-analysis (13) was used to produce funnel plots to assess the potential for publication bias. METANINF: Stata module to evaluate influence of a single study in meta-analysis estimation (14) was used to test for particularly influential studies. Sensitivity tests were conducted to assess the extent to which assumptions used in the meta-analysis or future research findings could affect the results. The databases and Stata command files used for the analysis are available in a digital repository for other researchers to review and replicate (see Supplemental Data in the current online issue).


View this table:
[in this window]
[in a new window]

  		TABLE 1 Review of longitudinal epidemiologic studies on the association between sugar-sweetened beverage (SB) consumption and weight gain

 

View this table:
[in this window]
[in a new window]

  		TABLE 2 Review of randomized controlled trials (RCTs) of the association between sugar-sweetened beverage (SB) consumption and weight gain

 
For some studies it was necessary to apply scaling factors to express the results in terms of the change ({Delta}) in BMI units per 12-oz serving {Delta}SB. One study defined a serving size as 100 g instead of the more typical 370-g (12-oz) serving size (15), and another used 1-oz units in the analysis (16). For consistency, the coefficient and SE for those studies were scaled by factors of 3.7 and 12, respectively, to make them consistent in serving size with other studies. When results were presented in weight change, the effect size was divided by the square of the average height in meters for respondents in the study.

Forest plots show the estimated effect size, CI, and the precision of each study in the meta-analysis. Forest plots were generated, and overall estimates of the pooled relation and SE were calculated with the use of both fixed-effects and random-effects models. Fixed-effects models assume that a single common effect underlies all of the studies in the meta-analysis. Random-effects models assume that not all studies in the analysis are estimating the same underlying common effect. Test statistics indicated heterogeneity in the results from the studies included in the meta-analysis, so the random-effects estimates are more appropriate.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Quantitative meta-analysis of longitudinal studies
Ten longitudinal studies were identified for the review (15-24), and 8 studies were included in the quantitative analysis (15-18, 20-23). Four of those studies used {Delta}BMI as the outcome variable and {Delta}SB consumption as the key independent variable. Two of those studies reported separate, independent models for males and females and are consequently included twice in the meta-analysis forest plot (16, 17). One longitudinal study used the change in fat mass as the outcome variable and SB consumption as an independent variable in a multilevel random-effects model (16). The study reported no statistically significant association between SB consumption and fat mass development. The effect size and SE for fat mass development was converted to {Delta}BMI units with the use of an average height of 1.6 m for females and 1.7 m for males. Blum et al (18) reported that they did not find a statistically significant association between SB and BMI z score in a 2-y study of 166 school-age children. The coefficient was not reported in the original article, but a review reported that the coefficient was –0.003 with an SE of 0.004 (9). These statistics were converted to {Delta}BMI units with the use of the L-M-S method. A longitudinal study used BMI z score as the outcome variable and SB consumption as the independent variable in a linear mixed-effects model (23). That study found a positive association between SB consumption and BMI z score. BMI z score was converted to {Delta}BMI with the use of the L-M-S method. Finally, a longitudinal study of 21 subjects used {Delta}kilogram as the outcome variable (21). The {Delta}kilogram was converted to {Delta}BMI by dividing by the average height in meters squared of the subjects.

Randomized controlled trial studies
We identified 2 RCT studies that met our criteria (25, 26). For the purpose of the quantitative meta-analysis we extracted the estimated difference and SE between the intervention group and the control group. None of the RCT studies found a statistically significant difference between the treatment and control groups. The estimated differences in BMI ranged from 0.1 to 0.14.

Viewed graphically, the studies with the most weight showed remarkably similar results (Figure 1Go). All of the studies with >5% weight had effect sizes near zero, and all had relatively precise estimates because of the large sample sizes used in the studies. The estimated associations between {Delta}SB consumption and {Delta}BMI ranged from –0.02 to 0.04 {Delta}BMI per serving per day among the studies with >5% weight. The results from Ludwig et al (20) and Phillips et al (23) stand out from the others. Ludwig et al (20) had the highest estimated association (0.24 {Delta}BMI/serving per day) and the largest 95% CI of the longitudinal studies. Phillips et al (23) also had a positive, statistically significant estimated association with a relatively large CI.


Figure 1
View larger version (15K):
[in this window]
[in a new window]

  		FIGURE 1. Forest plot of studies of sugar-sweetened beverage (SB) consumption and BMI (in kg/m2) in children and adolescents. I–V, fixed-effects estimate (inverse variance method); D+L, random-effects estimate (Der Simonian and Laird method). The points for each study indicate the reported estimate for the predicted change ({Delta}) in BMI per serving {Delta}SB consumption. The size of the gray boxes indicates the weight the study received in the calculation of the overall association and CI. The weight given each study is based on the precision of the estimate. The horizontal lines indicate the 95% CI for each study. The 2 diamonds represent the overall estimate of the predicted {Delta}BMI per serving {Delta}SB consumption, based on the pooled results from all studies. The upper open diamond represents the fixed-effects estimation that assumes that the "true" treatment effects in all of the studies are the same. The lower open diamond represents the random-effects estimation that assumes that the true treatment effect may vary across studies. The study by Mrdjenovic and Levitsky (21) was omitted from this forest plot because it had a much larger CI than any other study. The omission of that study does not affect either the fixed-effects or random-effects estimates.

 
The overall estimate of the association was a 0.004 change in BMI during the time period defined by the study for each serving per day change in SB consumption with a 95% CIs of –0.006 and 0.014 with the use of the fixed-effects model and 0.017 with a 95% CIs of –0.009 and 0.044 with the use of the random-effects model. The study by Mrdjenovic and Levitsky (21) was excluded from the forest plot because it had a much larger CI than any other study, but adding the study by Mrdjenovic and Levitsky (21) to the meta-analysis produces exactly the same results.

Qualitative analysis of longitudinal studies
The results of the other longitudinal studies are consistent with the results of the quantitative meta-analysis. They show that, although the association between {Delta}SB consumption and {Delta}BMI may be statistically significant in some studies, the magnitude of the association is not large.

Field et al (19) analyzed the Growing Up Today Survey and reported no association between snack food consumption (including SBs) and {Delta}BMI. The study was not included in the meta-analysis because the study did not estimate the independent association of SBs with {Delta}BMI. In addition, another study included in the meta-analysis used the same Growing Up Today Survey data set (17). Including 2 studies that used the same data set would be unnecessary and inappropriate.

Welsh et al (24) examined the association between baseline consumption of all sweet drinks (soda, fruit drinks, vitamin C–containing juice, and other juice) and development of overweight among 10 904 low-income children 2-3 y of age. Consumption of sweet drinks was not associated with the risk of overweight for children who were normal or underweight at baseline. A statistically significant association was observed between sweet drink consumption and the risk of overweight for those children who were overweight or at risk of overweight at baseline. The study was not included in the meta-analysis because it did not report the association between SBs and BMI. All of the reported estimates combined SBs with other beverages. The results of these longitudinal studies are consistent with the findings from the meta-analysis.

Publication bias and sensitivity tests
We examined the potential for publication bias and the overall robustness of the meta-analysis with the use of 4 approaches. We analyzed the influence of each individual study on the overall results, performed a funnel plot analysis, searched for relevant studies in databases of unpublished research, and conducted sensitivity tests.

No single study had a large influence on the overall results. The meta-analysis was reestimated, removing one study at a time. The average effect size ranged from a maximum of 0.020 (95% CI: –0.004, 0.043) when the study by Mundt et al (26) (males) was excluded to a minimum of 0.005 (95% CI: –0.012, 0.022) when the study by Phillips et al (23) was excluded in the random-effects model.

One general concern with meta-analysis is publication bias—the set of published studies may not represent the full spectrum of results from published and unpublished studies. For example, published studies may not include some studies that did not report statistically significant findings because there is a tendency to reject studies with results that are not significant. If unpublished studies finding no relation exist, the overall effect size will be closer to zero or have a wider variance than reported in this analysis. Recent articles have suggested that research supported by the food industry may not publish results that show a positive association between SB consumption and BMI (27, 28). To the best of our knowledge, none of the studies included in this meta-analysis received funding from the food industry. If statistically significant results were not published, the overall effect size will be larger than the pooled estimate reported in this analysis.

The funnel plot is a common diagnostic tool to assess publication bias. The studies are plotted with the precision of the study on the vertical axis and the effect size on the horizontal axis. In the absence of publication bias, the funnel plot is expected to show greater dispersion among the less-precise studies at the bottom of the plot, but studies should be distributed in a roughly symmetrical pattern around the average effect size. If the funnel plot is asymmetrical, it suggests that publication bias may be present.

The funnel plot for these studies shows that the studies with precise estimates are tightly and symmetrically grouped around the average effect size (Figure 2Go). The less-precise studies are not symmetrically grouped around the average effect size. None of the less-precise studies showed an effect size less than the average effect size, and many were above the pseudo-95% CI. This is consistent with a bias against publishing studies that do not show statistically significant results. It is not consistent with a bias against publishing studies that show large effect sizes. If studies with nonsignificant or negative findings exist but were not published, the pooled estimates from the meta-analysis will be larger than the true effect.


Figure 2
View larger version (7K):
[in this window]
[in a new window]

  		FIGURE 2. Funnel plot of studies of sugar-sweetened beverage (SB) consumption and BMI (in kg/m2) in children and adolescents. Studies are plotted with the estimated coefficient along the horizontal axis and the SE of the coefficient along the vertical access. In the absence of publication bias, the plot should be symmetric around the average coefficient with greater dispersion among the less-precise studies at the bottom of the plot. The more-precise studies are symmetrically distributed and generally fall within the pseudo-95% CI. The less-precise studies are not symmetrically distributed around the average coefficient. This is consistent with publication bias against studies that did not show statistically significant results.

 
To further examine the possibility of publication bias, we conducted a search of databases that contain unpublished scientific studies. The databases searched were Computer Retrieval of Information on Scientific Projects, Current Research Information System, Web of Science, ClinicalTrials.gov, BIOSIS Previews, and Proquest Interdisciplinary Dissertations and Theses. We sent an e-mail request to the authors or principal investigators of all relevant studies to ascertain whether any results had been published. No authors identified published studies in response to the e-mail requests.

From this search, we identified 2 abstracts that were eligible for inclusion in the meta-analysis. Jimenez et al (29) reported on a prospective cohort of 3070 Mexican adolescents 11–19 y of age. They found a positive association between the percentage of energy from SBs and BMI over time. Respondents who increased the percentage of energy from SBs had greater weight gain (β = 0.4, 95% CI: 0.19, 0.57) than did those who maintained or reduced the percentage of energy from SBs. The study was supported by Bristol-Myers Squibb Foundation, NY, and CONACyT, Mexico. When the study by Jimenez et al (29) is included in a sensitivity test, the fixed-effects pooled estimate is 0.00 (95% CI: –0.01, 0.01) and the random-effects pooled estimate is 0.03 (95% CI: 0.00, 0.06).

Gropper et al (30) studied 109 low-income African American children in rural Alabama. When the analysis was restricted to the 80 respondents who reported some SB consumption, they found that the 30% of respondents with the greatest weight gain reported more consumption of SBs (208 ± 22 kcal) than did the 30% of respondents with the lowest weight gain (142 ± 22 kcal). The study was funded by a grant from the US Department of Agriculture. The study by Gropper et al (30) was not included in a sensitivity test because the researchers did not report the association between SB consumption and BMI.

The search did show several important ongoing studies that will eventually contribute new data for any future meta-analysis. We performed 4 sensitivity tests to determine the extent to which assumptions used for the analysis or results from new studies could affect the findings of the meta-analysis.

Sensitivity test 1: effect sizes that are not adjusted for total energy
Berkey et al (17) and Ludwig et al (20) reported the results of models that controlled for total energy as well as models that did not adjust for total energy, and the meta-analysis presented here uses the results from the energy-adjusted models. Using the results from the unadjusted models has a small effect on the results of the meta-analysis. The fixed-effect estimate is 0.008 (95% CI: –0.002, 0.018) and the random-effects estimate is 0.023 (95% CI: –0.004, 0.050).

Sensitivity test 2: 5 more studies comparable to Berkey et al
Berkey et al (17) is a high-quality study showing a positive, nonsignificant relation between {Delta}SB and {Delta}BMI. Sensitivity test 2 imagines that 10 more studies (5 for girls and 5 for boys) are reported with the exact same effect sizes and precision as the study by Berkey et al (17). The results from this sensitivity test have nearly the same random-effects estimate (0.015) but a smaller 95% CI (0.004, 0.026) than did the original meta-analysis results. The results of the sensitivity test are statistically significant but close to zero.

Sensitivity test 3: high BMI results of Ebbeling et al
Ebbeling et al (25) reported on the interaction effect between BMI at baseline and the treatment effect. Subjects in the highest tertile of subjects showed a larger treatment effect than did subjects in the first and second tertile. Sensitivity test 3 uses the results from the highest tertile in place of the results for the entire sample and reestimates the model for RCT studies only. Limiting the sensitivity test to RCT studies ensures that the result from Ebbeling et al (25) receives the highest possible weight. The meta-analysis estimates a larger effect size but also larger CIs. The fixed-effects estimates are 0.004 (95% CI: –0.006, 0.014) and the random-effects estimates are 0.019 (95% CI: –0.009, 0.047). Neither estimate is distinguishable from zero. This estimate excludes the subjects in the lower 2 tertiles of baseline BMI, so it may not be generalizable to the entire population.

Sensitivity test 4: blockbuster studies analysis
Fail-safe studies analyze how many studies with null results would be required before the meta-analysis results would not be statistically significant. In this case, the relevant question is whether new studies with large effect sizes and high precision would produce a statistically significant pooled estimate that is large enough to be substantively important.

Sensitivity test 4 adds a "blockbuster" study that reports an effect size ≥2 times as large as any other longitudinal study examined (0.50) and with a precision equal to the most precise estimate of any longitudinal study examined (0.01). With the hypothetical blockbuster study, the fixed-effects estimate is 0.110 (95% CI: 0.101, 0.119) and the random-effects estimate is 0.091 (95% CI: –0.041, 0.224). The fixed-effects estimate is statistically significant, but the random-effects estimate is not statistically significant. A test for heterogeneity shows that the assumptions needed for the fixed-effects estimate are not supported, so the random-effects estimate is the most appropriate one to use.

Limitations
The limitations of each individual study are relevant for the meta-analysis. In particular, measurement error in the instruments used to measure beverage consumption in the longitudinal studies may have affected their results. If the measurement error is random, the standard errors will be larger than they would be in the absence of measurement error. If there is systematic measurement error, the reported coefficients may be biased. However, the longitudinal studies reported used validated dietary instruments that should minimize measurement error as much as possible in self-reported surveys.

The studies used different instruments to measure SB consumption, different measures of weight gain, different statistical models to estimate the effect sizes, and different units of time. Every effort was made to scale the effect sizes to comparable units, but these differences raise the issue of comparability between the studies. The sensitivity tests and tests for the influence of individual studies partially address this limitation.

Most of the studies covered a relatively short time period, typically 1 or 2 y. This limits the ability to assess any long-term effects from SB consumption. Restricting the meta-analysis to children and adolescents limits the ability to extrapolate these results to adults and understand how these dietary patterns carry over into adulthood (31). It is important to note that all of these limitations apply to purely narrative reviews of the literature as well as to meta-analysis.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
The results of the meta-analysis show that the current science base finds that the relation between SB consumption and BMI among children and adolescents is near zero. The best current scientific evidence shows that the relation between {Delta}SB consumption and {Delta}BMI is probably {approx}0.02 with an upper confidence limit of 0.04. The overall estimate is not statistically significant. The time period used to assess this relation varied across studies, but the most common time period was 1 or 2 y.

Neither of the RCT studies for children or adolescents found a statistically significant difference in either BMI or weight change (in kg) between the treatment and control groups. Ebbeling et al (25) did find a statistically significant association in the subset of the sample that was in the upper tertile of BMI at baseline, and this possible interaction between BMI and soft drink consumption deserves further study.

A statistical analysis can never definitely state that an association or effect is exactly zero. Given a large enough sample size or enough additional studies, even a small association may be distinguished from zero. However, these results of the quantitative meta-analysis provide a precise estimate of the association of the relation between SB consumption and BMI that is close to zero. Several independent longitudinal studies have consistently found the association to be close to zero with a small CI. Taken as a set, the studies suggest that the association between {Delta}SB consumption and {Delta}BMI is {approx}0.02, and we can probably rule out any effect size greater than {approx}0.04/study period for each serving per day {Delta}SB unless significant new evidence is presented. For context, the BMI-for-age data from the Centers for Disease Control and Prevention show that during adolescence BMI at the median increases by {approx}0.5 unit/y. The average consumption of SB is <2 (12-oz) servings/d even in the age-sex group with the highest consumption of SBs. Completely eliminating two 12-oz servings would reduce BMI by only {approx}0.04 with the overall estimate or by {approx}0.08 with the upper end of the 95% CI.

Several investigators have recently published critical reviews of the literature on SB consumption and weight gain and have come to different conclusions. A recent meta-analysis that used different methods found a small effect size of SB consumption on weight gain for children. The estimated effect size was r = 0.03 with a CI of 0.02 and 0.04 (28). Malik et al concluded that "[t]he weight of epidemiologic and experimental evidence indicates that a greater consumption of SBs is associated with weight gain and obesity. Although more research is needed, sufficient evidence exists for public health strategies to discourage consumption of sugary drinks as part of a healthy lifestyle" (9; p 274). However, 3 other recently published review articles have reviewed much of the same literature and concluded that the evidence for a relation is weak or equivocal (6, 32, 33). The assessment of the literature in those 3 articles found less consistent scientific evidence for the conclusion that SB consumption makes a unique contribution to the risk of weight gain and obesity.

There are 2 primary reasons that these separate reviews reached different conclusions than that of Malik et al (9). First, the other review articles considered the magnitude of the reported associations between SB consumption and BMI in the studies. Even when statistically significant, these associations were generally small. Second, many of the articles cited by Malik et al (9) as supportive of a link between SB consumption and weight gain or obesity contain other findings that contradict any link between SB consumption and obesity other than that which may be associated with its energy content. For example, DiMeglio and Mattes (34), James et al (26), and Ebbeling et al (25) each report no statistically significant difference in BMI or weight gain between the control and treatment groups.

SBs are a source of energy, and excess energy consumption will lead to weight gain. Dietary advice and education for children and adolescents should clearly communicate that SBs should only be consumed in moderation as part of a balanced diet. Children and adolescents who are overweight or at risk of becoming overweight should identify all sources of excess calories and work to modify their diet and increase their physical activity. These results suggest that some form of compensation is going on, and, depending on the form of compensation, it may have beneficial or detrimental health effects. Reducing SB consumption may or may not have other benefits, but these results suggest that the effect on weight will be effectively zero when considering the entire subpopulation of children and adolescents. More research is needed to determine whether certain subpopulations, such as those who are already overweight, may see a weight loss benefit from reducing SB consumption.

Obesity and overweight among children and adolescents are serious public health problems. More studies, particularly RCTs, are needed to investigate proposed processes for reducing obesity. The strongest current evidence is that reducing or eliminating SB consumption would not have a large effect on the BMI distribution of children or adolescents.


   ACKNOWLEDGMENTS
 
We thank Dr David Allison of the University of Alabama at Birmingham for his review of an earlier version of the manuscript and for providing very useful comments and suggestions. We thank Dr Clark Mundt and Dr Adam Baxter-Jones of the University of Saskatchewan for providing additional information about their analysis. Any remaining errors are the responsibility of the authors.

The author's responsibilities were as follows—RAF: designed the study, analyzed the data, and contributed to writing the manuscript; PAA: conducted the MEDLINE searches, acquired the articles, and contributed to writing the manuscript; MLS: contributed to writing the manuscript.

The research center with which the authors are affiliated has received financial support from the Coca-Cola Company and PepsiCo Inc that was unrelated to this project. MLS was affiliated with the University of Maryland when the manuscript was written and later accepted a position with the American Beverage Association. RAF and PAA declare that they have no personal conflicts of interest, such as stock ownership, employment, or consulting arrangements.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. World Health Organization. Obesity and overweight fact sheet. Geneva, Switzerland: World Health Organization, 2003. Internet: http://www.who.int/dietphysicalactivity/media/en/gsfs_obesity.pdf (accessed 12 November 2006).
  2. Keith SW, Redden DT, Katzmarzyk PT, et al. Putative contributors to the secular increase in obesity: exploring the roads less traveled. Int J Obes (Lond) 2006;30:1585–94.[Medline]
  3. Ebbeling CB, Pawlak DB, Ludwig DS. Childhood obesity: public-health crisis, common sense cure. Lancet 2002;360:473–82.[Medline]
  4. Mello MM, Studdert DM, Brennan TA. Obesity–the new frontier of public health law. N Engl J Med 2006;354:2601–10.[Free Full Text]
  5. Storey ML, Forshee RA, Anderson PA. Beverage consumption in the US population. J Am Diet Assoc 2006;106:1992–2000.[Medline]
  6. Bachman CM, Baranowski T, Nicklas TA. Is there an association between sweetened beverages and adiposity? Nutr Rev 2006;64:153–74.[Medline]
  7. Bawa S. The role of the consumption of beverages in the obesity epidemic. J R Soc Health 2005;125:124–8.[Medline]
  8. Gill TP, Rangan AM, Webb KL. The weight of evidence suggests that soft drinks are a major issue in childhood and adolescent obesity. Med J Aust 2006;184:263–4.[Medline]
  9. Malik VS, Schulze MB, Hu FB. Intake of sugar-sweetened beverages and weight gain: a systematic review. Am J Clin Nutr 2006;84:274–88.[Abstract/Free Full Text]
  10. Wiecha JL, Finkelstein D, Troped PJ, Fragala M, Peterson KE. School vending machine use and fast-food restaurant use are associated with sugar-sweetened beverage intake in youth. J Am Diet Assoc 2006;106:1624–30.[Medline]
  11. STATA software, version 9.2 Stata Corp. College Station, TX.
  12. Bradburn M, Deeks J, Altman D, et al. METAN: Stata module for fixed and random effects meta-analysis. Version 2.34 (Boston College Department of Economics, Boston, MA, 2007). Internet: http://ideas.repec.org/c/boc/bocode/s456798.html (accessed 7 January 2008).
  13. Stern J. METAFUNNEL: Stata module to produce funnel plots for meta-analysis. Version 1.0.2 (Boston College Department of Economics, Boston, MA, 2003). Internet: http://ideas.repec.org/c/boc/bocode/s434101.html (accessed 7 January 2008).
  14. Steichen T. METANINF: Stata module to evaluate influence of a single study in meta-analysis estimation. Version 1.0.2 (Boston College Department of Economics, Boston, MA, 2004). Internet: http://ideas.repec.org/c/boc/bocode/s419201.html (accessed 7 January 2008).
  15. Striegel-Moore RH, Thompson D, Affenito SG, et al. Correlates of beverage intake in adolescent girls: the National Heart, Lung, and Blood Institute Growth and Health Study. J Pediatr 2006;148:183–7.[Medline]
  16. Mundt CA, Baxter-Jones AD, Whiting SJ, Bailey DA, Faulkner RA, Mirwald RL. Relationships of activity and sugar drink intake on fat mass development in youths. Med Sci Sports Exerc 2006;38:1245–54.[Medline]
  17. Berkey CS, Rockett HR, Field AE, Gillman MW, Colditz GA. Sugar-added beverages and adolescent weight change. Obes Res 2004;12:778–88.[Medline]
  18. Blum JW, Jacobsen DJ, Donnelly JE. Beverage consumption patterns in elementary school aged children across a two-year period. J Am Coll Nutr 2005;24:93–8.[Abstract/Free Full Text]
  19. Field AE, Austin SB, Gillman MW, Rosner B, Rockett HR, Colditz GA. Snack food intake does not predict weight change among children and adolescents. Int J Obes Relat Metab Disord 2004;28:1210–6.[Medline]
  20. Ludwig DS, Peterson KE, Gortmaker SL. Relation between consumption of sugar-sweetened drinks and childhood obesity: a prospective, observational analysis. Lance. 2001;357:505–8.[Medline]
  21. Mrdjenovic G, Levitsky DA. Nutritional and energetic consequences of sweetened drink consumption in 6- to 13-year-old children. J Pediatr 2003;142:604–10.[Medline]
  22. Newby PK, Peterson KE, Berkey CS, Leppert J, Willett WC, Colditz GA. Beverage consumption is not associated with changes in weight and body mass index among low-income preschool children in North Dakota. J Am Diet Assoc 2004;104:1086–94.[Medline]
  23. Phillips SM, Bandini LG, Naumova EN, et al. Energy-dense snack food intake in adolescence: longitudinal relationship to weight and fatness. Obes Res 2004;12:461–72.[Medline]
  24. Welsh JA, Cogswell ME, Rogers S, Rockett H, Mei Z, Grummer-Strawn LM. Overweight among low-income preschool children associated with the consumption of sweet drinks: Missouri, 1999–2002. Pediatrics 2005;115:e223–9.[Abstract/Free Full Text]
  25. Ebbeling CB, Feldman HA, Osganian SK, Chomitz VR, Ellenbogen SJ, Ludwig DS. Effects of decreasing sugar-sweetened beverage consumption on body weight in adolescents: a randomized, controlled pilot study. Pediatrics 2006;117:673–80.[Abstract/Free Full Text]
  26. James J, Thomas P, Cavan D, Kerr D. Preventing childhood obesity by reducing consumption of carbonated drinks: cluster randomised controlled trial. BMJ. 2004;328:1237.[Abstract/Free Full Text]
  27. Lesser LI, Ebbeling CB, Goozner M, Wypij D, Ludwig DS. Relationship between funding source and conclusion among nutrition-related scientific articles. PLoS Med 2007;4:e5.[Medline]
  28. Vartanian LR, Schwartz MB, Brownell KD. Effects of soft drink consumption on nutrition and health: a systematic review and meta-analysis. Am J Public Health 2007;97:667–75.[Abstract/Free Full Text]
  29. Jimenez A, Flores M, Rodriguez S, Allen B, Burguete A, Lazcano E. Consumption of sugar-sweetened beverages is associated with an increase in body mass index in a cohort of Mexican adolescents. FASEB J 2006;20:A987.[Free Full Text]
  30. Gropper SS, Gilmore E, Galloway P, et al. Sweetened beverage consumption and weight gain among low income African American children in rural Alabama. FASEB J 2005;19:A1021.
  31. Tordoff MG, Alleva AM. Effect of drinking soda sweetened with aspartame or high-fructose corn syrup on food intake and body weight. Am J Clin Nutr 1990;51:963–9.[Abstract/Free Full Text]
  32. Forshee RA, Storey ML, Allison DB, et al. A critical examination of the evidence relating high fructose corn syrup and weight gain. Crit Rev Food Sci Nutr 2007;47:561–82.[Medline]
  33. Pereira MA. The possible role of sugar-sweetened beverages in obesity etiology: a review of the evidence. Int J Obes (Lond) 2006;30(suppl 3):S28–36.
  34. DiMeglio DP, Mattes RD. Liquid versus solid carbohydrate: effects on food intake and body weight. Int J Obes Relat Metab Disord 2000;24:794–800.[Medline]
Received for publication July 3, 2007. Accepted for publication February 13, 2008.




This article has been cited by other articles:


Home page
 Am. J. Clin. Nutr.Home page
M. Bes-Rastrollo and M. A. Martinez-Gonzalez
Differential underreporting and other caveats about sugar-sweetened beverages and weight gain
Am. J. Clinical Nutrition, November 1, 2008; 88(5): 1450 - 1451.
[Full Text] [PDF]

Home page
 Am. J. Clin. Nutr.Home page
M. L Storey
Reply to M Bes-Rastrollo and MA Martinez-Gonzalez
Am. J. Clinical Nutrition, November 1, 2008; 88(5): 1451 - 1452.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Data
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Google Scholar
Right arrow Articles by Forshee, R. A
Right arrow Articles by Storey, M. L
PubMed
Right arrow PubMed Citation
Right arrow Articles by Forshee, R. A
Right arrow Articles by Storey, M. L
Agricola
Right arrow Articles by Forshee, R. A
Right arrow Articles by Storey, M. L

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS