Bmj Usa

School soft drink intervention study

BMJ 2004; 329 doi: (Published 12 August 2004) Cite this as: BMJ 2004;329:E315
  1. Simone A French (french{at}, professor,
  2. Peter J Hannan, senior research fellow,
  3. Mary Story, professor
  1. Division of Epidemiology and Community Health University of Minnesota Minneapolis, MN
  2. Division of Epidemiology and Community Health University of Minnesota
  3. Division of Epidemiology and Community Health University of Minnesota

    Too good to be true?

    The paper by James et al in this issue of BMJ USA (p 410) describes the results of a five-session, classroom-based educational curriculum on changes in body mass index (BMI), over-weight prevalence, and carbonated beverage consumption among 7- to 11-year-old children in six schools. After one year, a significant difference between intervention and control groups was observed for change in prevalence of overweight. Total carbonated beverage intake did not significantly change among children in the control classrooms, but a significant decrease was observed among children in the intervention classrooms.

    No single study can provide irrefutable evidence linking soft drink consumption and obesity among children. While this paper provides some intriguing findings that warrant further exploration in follow-up studies, several notable methodological limitations cloud the interpretation of the findings and prevent firm conclusions from being drawn.

    Conceptually, perhaps the most important limitation is the absence of a mediational analysis that specifically links changes in soft drink consumption to changes in obesity prevalence.1 The study reports a significant decrease in the prevalence of overweight among children in intervention classrooms compared with children in control classrooms, and a significant decrease in total carbonated beverage consumption among children in intervention classrooms compared to those in control classrooms. However, it does not examine whether reductions in carbonated beverage consumption explain the reduction in obesity prevalence among children in intervention classrooms. Furthermore, it is unclear why the prevalence of overweight increased in the control classrooms, despite no change in carbonated beverage intake.

    At face value, the study does not convince the reader of the ability of a brief, five-session educational program to significantly change children's carbonated beverage intake on a scale large enough to impact the prevalence of overweight. If valid, these results present a striking contrast to previous school-based obesity prevention interventions that were much more intensive, involved both behavioral and environmental changes, and targeted both physical activity and eating behaviors.2,3

    At face value, the study does not convince the reader of the ability of a brief, five-session educational program to significantly… impact the prevalence of overweight.

    There are several limitations in the data for carbonated beverage intake. Only about half the children returned the beverage diaries that were used to assess changes in carbonated beverage intake. It is not clear whether the response rates for the beverage diaries differed among intervention and control classrooms. No data on the reliability or validity of these beverage measures are reported.4 It is not clear how the beverage data were collected and whether data on beverages other than carbonated beverages were collected. For example, no data are presented on the consumption of milk, fruit juices, fruit drinks, sports drinks or other beverages popular among children in this age range.

    Issues regarding the design and analysis of the study also warrant careful scrutiny.5 The analytic methods state that the classroom was the unit of analysis, but the presentation confuses individual information and cluster results. In table 1, for example, the change presented over 12 months implies a cohort analysis, but the text dealing with overweight (see full-text version on uses prevalence, a cross-sectional concept. No information is provided about how classrooms were randomized within schools, so the potential for contamination cannot be assessed, nor is it possible to assess how school-level effects should contribute to the analysis (they are not mentioned). The power calculation is based on change in soft drink consumption, a secondary outcome or behavioral mediator, not on change in overweight prevalence, the primary outcome. The power calculation is calculated based on 0.6 soft drinks per day but is presented in terms of detectable differences over three days. This makes it difficult to determine whether Cohen's effect size is good or very poor. To compound the problem, three detectable differences are given for four values of the intra-class correlation coefficient (ICC). Another point of confusion is the discrepancy between the data presented in the figure, in which all overweight prevalences exceed 20%, and the statement that “19.0% of the children were overweight at baseline.” Furthermore, missing data are dealt with in a rather cavalier manner.

    In summary, while the results of the study by James et al are intriguing and warrant further research, they are by no means conclusive. The study contributes to the growing scientific literature that is clarifying the role that sugar-sweetened beverages (primarily soft drinks) play in the rising epidemic of obesity among children and adolescents. It is clear that intake of these beverages has increased dramatically in the past two decades, and several studies suggest links with obesity in children. Interventions to reduce soft drink intake among children could be effective in preventing excess energy intake and over-weight. School settings are particularly important, given the widespread prevalence of soft drink vending machines in schools. Primary care physicians should advocate for school policies that ensure that only healthful beverages that contribute to nutrition and energy balance are available to children in schools.6


    • Paper p 410

    • Competing interests None declared.


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    View Abstract

    Sign in

    Log in through your institution

    Free trial

    Register for a free trial to to receive unlimited access to all content on for 14 days.
    Sign up for a free trial