Research Review

The In!uence of Sugar-Sweetened Beverage Warnings

A Randomized Trial of Adolescents’ Choices and Beliefs Eric M. VanEpps, PhD,1,2 Christina A. Roberto, PhD2

This activity is available for CME credit. See page A3 for information.

Introduction: California, New York, and the cities of San Francisco and Baltimore have introduced bills requiring health-related warning labels for sugar-sweetened beverages. This study measures the extent to which these warning labels in!uence adolescents’ beliefs and hypothetical choices.

Design: Participants completed an online survey in which they chose a beverage in a hypothetical vending machine task, rated perceptions of different beverages, and indicated interest in coupons for beverages. Data were collected and analyzed in 2015.

Setting/participants: A total of 2,202 demographically diverse adolescents aged 12–18 years completed the online survey.

Intervention: Participants were randomly assigned to one of six conditions: (1) no warning label; (2) calorie label; (3–6) one of four text versions of a warning label (e.g., SAFETY WARNING: Drinking beverages with added sugar(s) contributes to obesity, diabetes, and tooth decay).

Main outcome measures: Hypothetical choices, perceptions of beverages, interest in coupons, and endorsement of warning label policies were assessed.

Results: Controlling for frequency of beverage purchases, signi"cantly fewer adolescents chose a sugar-sweetened beverage in three of the four warning label conditions (65%, 63%, and 61%) than in the no label (77%) condition. Adolescents in the four warning label conditions chose fewer sugar- sweetened beverage coupons and believed that sugar-sweetened beverages were less likely to help them lead a healthy life and had more added sugar compared with the no label condition.

Conclusions: Health-related warning labels on sugar-sweetened beverages improved adolescents’ recognition of the sugar content of such beverages and reduced hypothetical choices to buy sugar- sweetened beverages. (Am J Prev Med 2016;51(5):664–672) Published by Elsevier Inc. on behalf of American Journal of Preventive Medicine

Introduction

Recent surveys have found that 77% of American adolescents drink sugar-sweetened beverages (SSBs) daily.1 Although soda consumption

among adolescents has decreased over the last 15 years,

the consumption of other SSBs like sports/energy drinks has simultaneously increased.1 According to one study, SSBs contribute approximately 225 kilocalories daily to the diets of adolescents aged 12–19 years, comprising approximately 10% of calorie intake.2 Research has linked children’s SSB consumption with weight gain and risk of obesity in adulthood, as well as dental caries.3–5

In response to high levels of SSB consumption and the health concerns associated with overconsumption, legis- lative bills introduced in California, New York, Vermont, Hawaii, and Washington would require health-related warning labels to be displayed on individual beverage packaging.6–11 Similarly, San Francisco passed a law in

From the 1VA Center for Health Equity Research and Promotion, Corporal Michael J. Crescenz VA Medical Center, Philadelphia, Pennsylvania; and 2Perelman School of Medicine, Department of Medical Ethics and Health Policy, University of Pennsylvania, Philadelphia, Pennsylvania

Address correspondence to: Eric M. VanEpps, PhD, 1127 Blockley Hall, 423 Guardian Drive, Philadelphia PA 19104. E-mail: [email protected]. upenn.edu.

0749-3797/$36.00 http://dx.doi.org/10.1016/j.amepre.2016.07.010

664 Am J Prev Med 2016;51(5):664–672 Published by Elsevier Inc. on behalf of American Journal of Preventive Medicine

2015 (still yet to be implemented at the time of this writing) requiring SSB advertisements to include a warning label that informs consumers of the potential health harms associated with drinking SSBs,12 and an ordinance introduced in the city of Baltimore would require health-related warnings on certain SSB advertise- ments, menus, and signs in locations where SSBs are sold.13 Studies of text warnings for tobacco products show improved consumer education, greater knowledge of health harms, and increased perceived risk of tobacco use.14 Combined with a recent study of the impact of SSB warning labels on parents’ decisions and judgments,15

this suggests SSB warning labels would similarly educate adult consumers, but there is a lack of research on how these labels would in!uence adolescents. Therefore, the present study aimed to answer the following questions:

1. Do warning labels educate adolescents about the health concerns associated with SSB intake, and do these warning labels outperform front-of-package calorie labels?

2. Do warning and calorie labels in!uence adolescents’ intentions to buy SSBs?

3. Do warning and calorie labels change adolescents’ percep- tions and purchase intentions of non-labeled beverages?

4. Do the effects of warning labels differ depending on label phrasing?

5. Are the effects of warning labels moderated by parent education, age, or whether one is overweight?

6. What are adolescents’ beliefs regarding policies requiring warning labels on SSBs?

It was hypothesized that exposure to a warning label would increase adolescents’ perceptions of SSB-related health problems and reduce purchase intentions for SSBs relative to exposure to calorie labels or no labels. The authors additionally generated different warning label phrasings, and hypothesized that two of these phrasings would have a greater impact on adolescents’ perceptions and intentions than the label phrasing recently proposed in California. Finally, the authors hypothesized that the effect of warning labels would be moderated by the education level of adolescents’ parents. This research provides the "rst evidence regarding

how adolescent beliefs and intentions can be in!uenced by SSB warning labels, and can inform regulatory efforts at both the local and state level where such labeling policies are being considered.

Methods Participants

The authors recruited 2,495 adolescents aged 12–18 years to complete an online survey. Participants were recruited through

Survey Sampling International, which maintains online panels and recruits from other online networks and websites. Recruitment materials invited individuals to “take a survey” without including additional details, thus minimizing selection bias. Once recruited, potential participants passed quality control questions and were then randomly assigned to surveys for which they likely qualify. Based on the recruited population, Survey Sampling International offers a variety of incentives, including cash, lotteries, and donations to charity. For this study, the authors worked with Survey Sampling

International to recruit families with an adolescent aged 12–18 years residing in the household, speci"cally seeking to oversample Hispanics and African Americans (Table 1) because they have higher obesity prevalence than other groups.16 Adolescents were recruited such that their parents’ education level was representa- tive of the U.S. population according to 2010 Census data. Overall, 2,495 participants started the survey, 2,282 completed

it, and 2,202 correctly answered the data integrity question (described below); these 2,202 adolescents made up the "nal sample.

Label Development and Randomization

Adolescent participants were randomized to one of six label conditions. The "rst condition was a control group in which participants were shown beverages without any health-related warning label (no label control). Participants in the second condition viewed a “calories per bottle” label on all beverages, not only SSBs (calorie label). These labels were identical to the American Beverage Association’s “Clear on Calories” labels.17

Although calorie labels may in!uence choices and perceptions, it was hypothesized that these labels would be less in!uential than labels that explicitly include a safety warning. Conditions 3–6 featured SSB warning labels. The "rst warning label read: SAFETY WARNING: Drinking beverages with added sugar(s) contributes to obesity, diabetes, and tooth decay (California label). The text in this warning is the same text proposed in the California bill6,7; the remaining three experimental conditions were mod- i"cations of that text. All labels were reviewed for accuracy by a scienti"c advisory board and a legal team to ensure legally permissible claims were tested. The "rst warning label modi"cation altered the original text by

changing “obesity” to “weight gain” (weight gain label) because obesity might seem like a distant, abstract problem, whereas weight gain might feel more immediate and tangible. For the second label, the phrase “preventable diseases like” was inserted before “obesity, diabetes, and tooth decay” (preventable label) to highlight that disease risk could be modi"ed by one’s behavior. Finally, a label that added the words “type 2” before “diabetes” (Type 2 diabetes label) was tested to address concerns that the California label is misleading because SSB intake does not impact development of Type 1 diabetes. Figure 1 displays all label images. The authors hypothesized that the “weight gain” and “preventable” labels would be more effective than the other warning labels, and that all four warning labels would be more in!uential than the no label and calorie label conditions because those groups did not view explicit safety warnings or descriptions of health problems. Study beverages quali"ed for a warning label based on

criteria set by the proposed California legislation, which

VanEpps and Roberto / Am J Prev Med 2016;51(5):664–672 665

November 2016

mandates a label for any nonalcoholic beverage with added sweeteners that contains Z75 calories per 12 !uid ounces.6,7

Survey Procedures

Upon meeting recruitment criteria and providing informed con- sent, adolescents completed the survey online (median completion time was 16 minutes). The survey was administered via Qualtrics and could only be completed on a computer-size screen; mobile devices or tablets were not permitted because their small screens would make it dif"cult to read labels and complete the survey. All data were collected and analyzed in 2015. The Harvard T. H. Chan School of Public Health IRB approved this study (Figure 2).

Primary Outcomes

Adolescents were asked to imagine they wanted to purchase a beverage from a vending machine, viewed 20 popular 20-ounce beverages (12 SSBs) presented in random order in two columns on the computer screen, and selected one for hypothetical purchase. Beverages that spanned a range of added sugar content were included. An effort was made to include beverages that consumers likely know are high in added sugar content (e.g., Coca Cola) or low in added sugar content (e.g., Dasani bottled water) as well as drinks that many consumers likely do not realize are high in added sugar (e.g., Arizona Green Iced Tea, Powerade). Study beverages consisted of sodas, juices, iced teas, still and seltzer waters, lemonade, and sports drinks. Energy drinks, such as Red Bull or Monster, were not included, nor were coffee or iced coffee beverages. Participants were required to view all beverages and were told to select the product they wanted even if they typically

Table 1. Socio-demographic Characteristics of Sample

Demographic characteristic Sample

N 2,202

Female, % 50.3

Average age (years) 15.0

Median BMI 22.1

Hispanic, % 31.6

Race, %

White 62.9

Black 33.6

Asian 1.8

Native American 2.1

Hawaiian 0.3

Other 4.5

Grade in school, %

5th 1.5

6th 6.3

7th 10.4

8th 13.9

9th 17.8

10th 18.4

11th 16.9

12th 14.9

Mother’s education, %

Less than high school 9.6

High school degree 18.6

Associate’s degree 10.6

Some college 20.6

College degree 25.2

At least some graduate school 14.0

Don’t know 1.4

Father’s education, %

Less than high school 6.8

High school degree 24.5

Associate’s degree 7.7

Some college 16.1

College degree 23.3

At least some graduate school 16.0

(continued)

Table 1. (continued)

Demographic characteristic Sample

Don’t know 5.6

Relationship with weight, %

Trying to lose weight 30.0

Trying to maintain weight 33.7

Trying to gain weight 4.9

Not trying to gain or lose weight 31.4

Has a doctor ever said you are overweight?, %

No 74.6

Currently 21.2

Not currently, but in the past 4.2

Has a doctor ever said you have type 2 diabetes?, %

No 94.4

Currently 5.3

Not currently, but in the past 0.3

VanEpps and Roberto / Am J Prev Med 2016;51(5):664–672666

www.ajpmonline.org

buy a different !avor (e.g., they should choose Fruit Punch Powerade even if they normally purchase Lemon Lime Powerade). Because consumers would have never previously encountered warning labels and might not realize why some beverages had them whereas others did not, the authors included a sentence in the instructional paragraph for the vending machine task that read: Drinks with a lot of added sugar have a safety warning label on them. Beverages displayed labels based on study condition and

eligibility for SSB warning labels. Calorie or warning labels were enlarged and displayed above beverage images to make these labels legible on a computer screen (Appendix Figure 1, available online). Next, participants answered questions about their perceptions

of and intentions regarding ten randomly ordered 20-ounce beverages, of which six were SSBs. Responses to these questions were averaged across beverages for analyses. Appendix Table 1 (available online) lists the survey questions. Participants viewed the same 20 randomly ordered beverages as

in the vending machine task and selected each beverage for which they would be interested in receiving a discount coupon.

Secondary Outcomes

At the end of the survey, participants in the control, calorie label, or California warning label conditions were presented with the California warning label, whereas those in the other conditions saw their assigned warning label. All participants were asked whether such a label would change their health beliefs about SSBs, willingness to buy SSBs, and whether they would favor a government policy requiring this label on SSBs (Appendix Table 1, available online).

Additional Measures

Participants indicated whether their doctor has told them they are currently overweight or have Type 2 diabetes, and whether they are currently trying to lose, gain, or maintain weight. Participants also stated their age, gender, height, weight, ethnicity, race, grade in school, parents’ educational level, and their U.S. state or territory of residence. As a manipulation check, participants indicated whether they

saw a warning label on any beverage (choosing among yes, no, and I don’t know). As expected, those in each of the warning label conditions were signi"cantly more likely to report seeing a warning label (range, 70%–75%) than those in the calorie label condition (6%) or those in the no label condition (6%), (!2[10]!895.14, po0.001). The "nal question asked participants the number of days in a

week (multiple choice, options ranged from 5 to 25 days). Those answering incorrectly were excluded from analyses (n!80).

Statistical Analysis

There were no observed differences across conditions with regard to gender, age, respondent BMI, respondent education, parents’ education, relationship with weight, or overweight status (all p40.15), suggesting that randomization to condition successfully balanced the groups. To assess differences in the outcome variables across conditions,

the authors regressed each dependent variable on label condition, controlling for self-reported frequency of both SSB and non- SSB purchases over the past month, and using a po0.05 signi"cance threshold. ANCOVAs were used to analyze continu- ous outcomes and logistic regressions to analyze categorical outcomes. For the ANCOVAs, pairwise comparisons were conducted with Tukey post hoc corrections; for the logistic regressions, the authors varied which condition was the reference group and used the Bonferroni–Holm procedure to correct for multiple comparisons.18

To determine whether the effects of calorie and warning labels differed by parent education level, each outcome variable was regressed on

1. a categorical variable for experimental condition, with no label as the reference group;

2. a binary variable for whether at least one parent received education beyond high school;

3. the interaction between experimental condition and parent education; and

4. the self-reported average frequency of SSB and non-SSB purchases.

The authors hypothesized that warning labels would be more in!uential for children of parents with a higher education level. As exploratory analyses, the authors also examined whether

participant age or being overweight moderated the effects of warning labels, replacing the binary variable for parent education with either mean-centered participant age or a binary variable for doctor-diagnosed current overweight status (as reported by participants; combining no and in the past responses). For all interaction analyses, the Bonferroni–Holm procedure was used to correct for multiple comparisons.

Figure 1. Different label conditions.

VanEpps and Roberto / Am J Prev Med 2016;51(5):664–672 667

November 2016

Results Among the six label conditions, there were differences in 13 of 15 analyzed measures (Table 2). Compared with the control condition, three of four warning labels reduced vending machine selection of SSBs. All four warning labels reduced the number of selected SSB coupons. They also increased subjective perceptions of added sugar in SSBs, and led to lower subjective perceptions of SSBs promoting a healthy life, helping one feel energized, or helping one focus at school. Calorie labels did not reduce selection of SSBs or SSB coupons compared to the control condition, and led to signi"cantly more SSB coupons being selected than in any warning label condition. Relative to the control condition, calorie labels increased perceptions of added sugar in SSBs and reduced percep- tions of SSBs promoting a healthy life or helping one focus at school, though not to the same extent as warning labels (Table 2). Finally, compared with all other con- ditions, calorie labels signi"cantly increased the esti- mated calories in SSBs, which were underestimated on average in all conditions. Among the warning label conditions, there were

signi"cant differences on three of 15 measures, although none of these differences followed ex ante predictions regarding the intended effects of these labels’ phrasings. Those exposed to the California label were more likely to select an SSB in the vending machine task than those

shown the Type 2 diabetes label, and, curiously, rated the risk of diabetes as lower than in any other condition. Those who saw type 2 diabetes labels rated SSBs as less delicious than those shown the weight gain label. Analyses of non-labeled beverages revealed signi"cant

effects of label condition on seven of eight perceptions and intentions measures, but only one of four disease risk measures (risk of weight gain). Most of these differences appear to be driven by the calorie label condition, as participants exposed to calorie labels rated these bever- ages as less healthy and having more added sugar than participants in other conditions (Appendix Table 2, available online). There were signi"cant interactions of the weight gain

warning label and overweight status (p!0.004) and the “preventable” warning label and overweight status (p!0.001) on vending machine choice. In both cases, overweight adolescents exposed to the warning label were less likely to choose SSBs than non-overweight adoles- cents shown the same labels. No other interactions (e.g., by parent education or age) emerged as statistically signi"cant. On average, participants reported that a warning label

would change their beliefs about a beverage’s healthful- ness (M!3.51 [SD!1.34] on a 5-point scale) and that a label would encourage them to purchase fewer SSBs (M!3.65 [SD!1.25] on a 5-point scale). Additionally,

Figure 2. CONSORT !ow diagram.

VanEpps and Roberto / Am J Prev Med 2016;51(5):664–672668

www.ajpmonline.org

Table 2. Sugar-Sweetened Beverage (SSB) Outcomes by Study Condition, Percentages and Means (and SEs)

Control Calorie label California warning

Weight gain warning

Preventable warning

Type 2 diabetes warning

Vending machine choice

% Choosing an SSB 77.2%d,e,f

(2.2%) 72.5%f (2.4%) 69.1%

(2.4%) 64.5%a (2.5%) 63.0%a (2.6%) 60.8%a,b (2.5%)

SSB perceptions and intentions

Delicious (1–7) 4.87 (0.06) 4.97f (0.07) 4.90 (0.06) 4.97f (0.07) 4.82 (0.06) 4.69b,d (0.07)

Healthy (1–7) 3.90e,f (0.06)

3.95e,f (0.07) 3.77 (0.08) 3.79 (0.08) 3.61a,b (0.08) 3.61a,b (0.08)

Purchase intention (1–7)

4.13 (0.07) 4.21e,f (0.08) 4.08 (0.08) 4.07 (0.08) 3.93b (0.08) 3.92b (0.08)

Likely to drink (1–7) 4.15 (0.07) 4.25e,f (0.08) 4.08 (0.08) 4.15 (0.08) 3.96b (0.08) 3.95b (0.08)

Energized (1–7) 5.02c,d,e,f (0.06)

4.80f (0.06) 4.72a (0.06) 4.70a (0.07) 4.69a (0.07) 4.52a,b (0.07)

Focus (1–7) 4.51b,c,d,e,f (0.06)

4.09a (0.08) 4.07a (0.07) 4.07a (0.08) 3.93a (0.08) 3.90a (0.08)

Amount of added sugar (1–4)

2.88b,c,d,e,f (0.03)

3.01a (0.03) 3.05a (0.03) 3.05a (0.03) 3.06a (0.03) 3.07a (0.03)

Estimated calories 91.22b,e,f (4.96)

180.62a,c,d,e,f (6.35)

102.61b (5.33)

104.30b (5.86)

114.61a,b (5.88)

111.53a,b (5.46)

SSB disease risk

Weight gain (1–7) 4.65 (0.06) 4.57 (0.06) 4.51 (0.07) 4.66 (0.06) 4.76 (0.06) 4.56 (0.07)

Heart disease (1–7) 4.23 (0.06) 4.22 (0.06) 4.31 (0.07) 4.28 (0.07) 4.36 (0.07) 4.28 (0.07)

Diabetes (1–7) 4.44c,e (0.06)

4.43c,e (0.06) 4.05a,b,d,e,f (0.07)

4.57c (0.07) 4.71a,b,c (0.07)

4.58c (0.07)

Healthy life (1–7) 4.57b,c,d,e,f (0.05)

3.90a,f (0.07) 3.83a (0.07) 3.76a (0.07) 3.70a (0.07) 3.67a,b (0.07)

Coupon choice

Number of SSB coupons (0–12)

3.64c,d,e,f (0.14)

3.66c,d,e,f (0.13)

3.00a,b (0.13)

2.92a,b (0.13) 2.85a,b (0.14) 2.70a,b (0.13)

Number of non-SSB coupons (0–8)

2.70 (0.10) 2.88d (0.10) 2.57 (0.10) 2.47b (0.10) 2.58 (0.10) 2.65 (0.10)

Note: N ! 2,202. Raw statistics are displayed. The “Perceptions and Intentions” means are averages across beverages. Within each row, boldface percentages or means with different superscripts differ at po0.05 (after correcting for multiple comparisons using Tukey post hoc tests), compared to the number of the corresponding column: aSigni"cantly different from control condition. bSigni"cantly different from calorie label condition. cSigni"cantly different from California warning. dSigni"cantly different from weight gain warning. eSigni"cantly different from preventable warning. fSigni"cantly different from Type 2 diabetes warning. For the vending machine choice, a Bonferroni-Holm correction rather than the Tukey post hoc test was applied because data were analyzed using logistic regression. All statistical tests controlled for the self-reported frequency of purchasing beverages that quali"ed for a warning label and ones that did not. Analyses of “Estimated Calories” were conducted on log-transformed estimates (i.e., log10[Calories"1]); the table converts the log means and SEs into calories (i.e., using 10log to calculate the mean). Non-SSB refers to those beverages that did not qualify for an SSB warning label. For analyses related to vending machine choice and coupon choice, the 12 SSBs were Pom Coconut, Nestea, 7Up, Canada Dry Ginger Ale, Tropicana Lemonade, Coca Cola, Arizona Green Tea, Mountain Dew, Purity Organic: Peach Paradise, Minute Maid Lemonade, Old Orchard Ruby Red Grapefruit Juice, and Mountain Berry Blast Powerade. The eight non-SSBs (those beverages that did not qualify for an SSB warning label) were Dasani Water, Simply Orange, Schweppes Seltzer Water, Diet Coca Cola, Honest Green Tea, Tropicana Orange Juice, Polar Seltzer Water, and Power-C Dragonfruit Vitamin Water. For analyses related to SSB perceptions and intentions and disease risk, the six SSBs were Coca Cola, Arizona Green Tea, Mountain Dew, Minute Maid Lemonade, Mountain Berry Blast Powerade, and Purity Organic: Peach Paradise. The four non-SSBs were Tropicana Orange Juice, Diet Coca Cola, Dasani Water, and Power-C Dragonfruit Vitamin Water.

VanEpps and Roberto / Am J Prev Med 2016;51(5):664–672 669

November 2016

62.7% of participants favored an SSB warning label policy, whereas only 7.8% were opposed (the average support was"0.85 [SD!1.06] on a scale from –2 to "2). These beliefs did not differ across experimental condi- tions.

Discussion Warning labels reduced adolescents’ perceptions that SSBs contribute to a healthy life and reduced beliefs that SSBs can increase their energy and help them focus. Calorie labels increased adolescents’ estimates of the calories in SSBs, as did two of four warning labels. Both calorie and warning labels led participants to subjectively evaluate SSBs to have more added sugar. Although shifts in perceptions are important, this

study also provides preliminary evidence that SSB warn- ing labels may affect behavior. In the vending machine task, participants who saw SSBs with warning labels were less likely to hypothetically purchase an SSB relative to those who saw no labels, an effect that was statistically signi"cant for three of four warning label conditions. When selecting hypothetical beverage coupons, adoles- cents who saw warning labels chose signi"cantly fewer SSB coupons than the control and calorie label con- ditions, suggesting that the warning labels reduced desire for a range of SSB options. Additionally, when stating perceptions and intentions, two of the warning label conditions led participants to report lower intentions to purchase SSBs in the future than those exposed to calorie labels. Overall, it appears that warning labels may encourage adolescents to purchase healthier beverages, whereas “calories per bottle” labels had no such impact on behavioral intentions. Whether this diminished impact of calorie labels is due to the safety and health information included in the health-related warning labels, greater novelty of warning labels, or consumer dif"culty in interpreting calorie labels is a question for future research. It is possible that additional interpreta- tive information, such as a clearly communicated thresh- old for high calorie content, could facilitate greater impact of calorie labels on behavioral intentions. By analyzing responses to both SSBs and beverages

that did not qualify for a warning label, the authors were able to assess potential spillover effects of SSB warning labels. The results suggest that SSB warning labels have little impact, either positively or negatively, on judgments of non-labeled beverages. This stands in contrast to the in!uence of calorie labels, which decreased perceptions of the healthfulness of non-labeled beverages and increased their perceived sugar content. These differ- ences between calorie labels and warning labels may exist because calorie labels were posted on all beverages

(even ones that contain some sugar, but would not qualify for a warning label), whereas warning labels only appeared on SSBs meeting a certain added sugar threshold. The in!uence of warning labels on measured out-

comes did not vary based on parent education, suggest- ing SSB warning labels may be helpful for adolescents regardless of their parents’ education level. This suggests that factors related to SES may have less impact on adolescents’ ability to use these labels relative to other types of nutrition labels, whereas research on restaurant menu calorie labeling has found that higher-income and higher-educated individuals are more likely to use menu calorie labels to make purchasing decisions.19,20 Future research could investigate whether SSB warning labels are categorically different from (and perhaps more universally understood than) other labels, or whether the hypothetical setting contributed to the uniqueness of the present results. Overall, there was little support for hypotheses that

modi"ed label phrasings would differentially impact outcomes of interest, although alternative phrasings were directionally more effective than the California label on several outcomes. Combined with past research regard- ing parents’ decisions,15 these data suggest the California text is a reasonable baseline standard for improving consumer knowledge of SSBs, but future research could further examine whether alternative phrasing, design, or placement of warning labels can improve their impact. Finally, adolescents expressed that government-

sponsored SSB warning labels would shift their beliefs about a beverage’s healthfulness and would motivate them to consume fewer SSBs. In addition, the majority of respondents favored a policy to place warning labels on SSBs.

Limitations This study has several limitations concerning general- izability. First, decisions were hypothetical. However, given that SSB labels do not yet exist in stores or other actual decision contexts, this hypothetical context enabled the authors to investigate the potential effect of a labeling policy. In line with research suggesting that tobacco labels have the largest impact when clearly legible and prominently displayed,14 the present work studied warning labels under conditions where labels were highly visible to identify how they could impact consumers who see them. Although such a design may overestimate the effect of a warning label, such research is important because the absence of an effect would suggest that labels would not in!uence actual decisions. It is also possible that the inclusion of an instructional sentence informing participants that warning labels appeared on

VanEpps and Roberto / Am J Prev Med 2016;51(5):664–672670

www.ajpmonline.org

drinks with a lot of added sugar could have increased the labels’ impact. Additionally, although this study had a large sample size, it may have lacked adequate statistical power to detect signi"cant differences between different label phrasings. The survey is also limited because of potential desirability bias. Consum- ers may realize that the authors are testing responses to labels and infer that they should indicate that they would not purchase an SSB. However, survey responses were anonymous and respondents had little incentive to please the researcher. Further, a strong social desirability bias would predict more signi"cant effects from exposure to the equally salient calorie labels, but such effects were not detected. The authors also did not recruit a nationally representative sample. However, the sample was large, racially and ethnically diverse, and recruited so that parent education level re!ected the educational makeup of the U.S. Finally, the warning label was tested using the inclusion criteria set by California legislation so that this study might inform current policy debates. However, different labeling requirements could include additional beverages, such as 100% fruit juices, and the present study does not address how warning labels would affect perceptions of those beverages. This study has a number of strengths, including a

large sample size, a randomized controlled design with both no label control and calorie label groups, a sample of adolescents ranging in age from 12 to 18 years, and a large proportion of racial and ethnic minority participants with a range of parental educa- tion levels. The present research is among the "rst to examine the in!uence of SSB warning labels and provides timely data on the potential for such labels to educate adolescents and reduce SSB intake. This study provides preliminary support for placing warn- ing labels on SSBs, setting the stage for future research to identify their impact on overall dietary choices in different settings.

Conclusions These results suggest that SSB warning labels are a promising strategy to reduce adolescents’ perceptions of SSBs’ healthfulness and decrease adolescents’ likelihood of buying SSBs.

This work was commissioned by the Healthy Eating Research Program of the Robert Wood Johnson Foundation. Dr. Roberto is supported by the National Institute On Aging of NIH under Award Number P30AG034546. The content of this article is solely the responsibility of the authors and does not necessarily represent the of"cial views of NIH. Dr. VanEpps

analyzed the data and drafted the initial manuscript. Dr. Roberto conceptualized and designed the study, critically reviewed the manuscript, and approved the "nal manuscript as submitted. No "nancial disclosures were reported by the authors of

this paper.

References 1. Han E, Powell LM. Consumption patterns of sugar-sweetened bev-

erages in the United States. J Acad Nutr Diet. 2013;113(1):43–53. http://dx. doi.org/10.1016/j.jand.2012.09.016.

2. Kit B, Fakhouri T, Park S, Nielsen S, Ogden C. Trends in sugar- sweetened beverage consumption among youth and adults in the United States: 1999–2010. Am J Clin Nutr. 2013;98(1):180–188. http: //dx.doi.org/10.3945/ajcn.112.057943.

3. Te Morenga L, Mallard S, Mann J. Dietary sugars and body weight: systematic review and meta analyses of randomized controlled trials and cohort studies. BMJ. 2013:346. http://dx.doi.org/10.1136/bmj.e7492.

4. Ludwig DS, Peterson KE, Gortmaker SL. Relation between consump- tion of sugar-sweetened drinks and childhood obesity: a prospective, observational analysis. Lancet. 2001;357(9255):505–508. http://dx.doi. org/10.1016/S0140-6736(00)04041-1.

5. Sohn W, Burt BA, Sowers MR. Carbonated soft drinks and dental caries in the primary dentition. J Dent Res. 2006;85(3):262–266. http: //dx.doi.org/10.1177/154405910608500311.

6. Senate Bill-1000. Public Health: sugar-sweetened beverages: safety warnings. http://leginfo.legislature.ca.gov/faces/billNavClient.xhtml? bill_id=201320140SB1000. Accessed April 2016.

7. Senate Bill-203. Sugar-sweetened beverages: safety warnings. https://leginfo. legislature.ca.gov/faces/billNavClient.xhtml?bill_id=201520160SB203. Accessed April 2016.

8. New York State Assembly Bill 2320-B. Requires sugar-sweetened bev- erages to be labeled with a safety warning. http://assembly.state.ny.us/leg/? default_!d=&bn=A02320&term=2015&Summary=Y&Actions=Y& Text=Y&Votes=Y#A02320. Accessed April 2016.

9. Vermont House Bill H0089. An act relating to health and safety warnings on sugar-sweetened beverages. https://legiscan.com/VT/text/ H0089/2015. Accessed April 2016.

10. State of Hawaii Senate Bill No. 1270. Relating to beverages. www.capitol. hawaii.gov/session2015/bills/SB1270_.htm. Accessed April 2016.

11. State of Washington House Bill 2798. An act relating to mitigating the adverse impacts of sugar-sweetened beverages; adding a new chapter to Title 70 RCW; prescribing penalties, and providing an effective date. http://law"lesext.leg.wa.gov/biennium/2015-16/Pdf/Bills/House%20Bills/ 2798.pdf. Accessed April 2016.

12. Sugar-sweetened beverage warning for advertisements. Ordinance No. 100-15. Article 42, Division 1 Sections 4200-06: Sugar sweetened beverage warning ordinance. www.sfbos.org/ftp/uploaded"les/bdsupvrs/ordinan ces15/o0100-15.pdf. Accessed April 2016.

13. City of Baltimore Ordinance 16-0617. Sugar-Sweetened Beverages— Warning Labels. https://baltimore.legistar.com/LegislationDetail.aspx? ID=2547410&GUID=BF49C0ED-0647-4625-B7AE-C2592FCAFD7C &Options=ID%7CText%7C&Search=sugar&FullText=1. Accessed April 2016.

14. Hammond, D. Health warning messages on tobacco products: a review. Tob Control. 2011;20(5):327–337. http://dx.doi.org/10.1136/ tc.2010.037630.

15. Roberto CA, Wong D, Musicus A, Hammond D. The in!uence of sugar-sweetened beverage health warning labels on parents’ choices. Pediatrics. 2016;137(2). http://dx.doi.org/10.1542/peds.2015-3185.

16. Division of Nutrition, Physical Activity, and Obesity: Adult Obesity Facts. Centers for Disease Control and Prevention website. www.cdc.

VanEpps and Roberto / Am J Prev Med 2016;51(5):664–672 671

November 2016

gov/obesity/data/adult.html. Updated June 16, 2015. Accessed July 30, 2015.

17. Clear on Calories. American Beverage Association website. www.amer ibev.org/nutrition-science/clear-on-calories/. Accessed July 30, 2015.

18. Holm S. A simple sequentially rejective multiple test procedure. Scandinavian J Stat. 1979;6(2):65–70.

19. Breck A, Cantor J, Martinez O, Elbel B. Who reports noticing and using calorie information posted on fast food restaurant menus? Appetite. 2014;81:30–36. http://dx.doi.org/10.1016/j.appet.2014.05.027.

20. Chen R, Smyser M, Chan N, Ta M, Saelens BE, Krieger J. Changes in awareness and use of calorie information after mandatory menu

labeling in restaurants in King County, Washington. Am J Public Health. 2015;105(3):546–553. http://dx.doi.org/10.2105/AJPH.2014. 302262.

Appendix

Supplementary data

Supplementary data associated with this article can be found at http://dx.doi.org/10.1016/j.amepre.2016.07.010.

VanEpps and Roberto / Am J Prev Med 2016;51(5):664–672672

www.ajpmonline.org