Effects of 4 Interpretive Front-of-Package Labeling Systems on Hypothetical Beverage and Snack Selections: A Randomized Clinical Trial

Abstract

Importance: Policymakers and researchers have proposed a variety of interpretative front-of-package food labeling systems, but it remains unclear which is most effective at encouraging people to choose healthier foods and beverages, including among people with less education.

Objective: To test the effects of 4 interpretative front-of-package food labeling systems on the healthfulness of beverage and snack selections, overall and by education level.

Design, setting, and participants: This randomized clinical trial of a national sample of US adults 18 years and older was conducted online from November 16 to December 3, 2022.

Intervention: Participants were randomized to view products with 1 of 5 food labeling systems, including control (calorie labels only) or 1 of 4 interpretative labeling systems: green ("choose often") labels added to healthy foods; single traffic light labels added to healthy, moderately healthy, and unhealthy foods; physical activity calorie equivalent labels added to all products; and nutrient warning labels added to products high in calories, sugar, saturated fat, or sodium. All conditions had calorie labels on all products.

Main outcomes and measures: Participants selected 1 of 16 beverages and 1 of 16 snacks that they wanted to hypothetically purchase. The primary outcomes were calories selected from beverages and from snacks. Secondary outcomes included label reactions and perceptions.

Results: A total of 7945 participants completed the experiment and were included in analyses (4078 [51%] female, 3779 [48%] male, and 88 [1%] nonbinary or another gender; mean [SD] age, 47.5 [17.9 years]). Compared with the control arm, exposure to the green (average differential effect [ADE], -34.2; 95% CI, -42.2 to -26.1), traffic light (ADE, -31.5; 95% CI, -39.5 to -23.4), physical activity (ADE, -39.0; 95% CI, -47.0 to -31.1), or nutrient warning labels (ADE, -28.2; 95% CI, -36.2 to -20.2) led participants to select fewer calories from beverages (all P < .001). Similarly, compared with the control label, exposure to the green (ADE, -12.7; 95% CI, -17.3 to -8.2), traffic light (ADE, -13.7; 95% CI, -18.2 to -9.1), physical activity (ADE, -18.5; 95% CI, -23.1 to -13.9), or nutrient warning labels (ADE, -14.2; 95% CI, -18.8 to -9.6) led participants to select fewer calories from snacks (all P < .001). These effects did not differ by education level. The green labels were rated as less stigmatizing than the other interpretative systems but otherwise generally received the least favorable label reactions and perceptions (eg, elicited less attention, were perceived as less trustworthy), while the nutrient warnings and physical activity labels received the most favorable ratings.

Conclusions and relevance: In this randomized clinical trial of front-of-package food labeling systems, all 4 interpretative labeling systems reduced calories selected from beverages and from snacks compared with calorie labels, with no differences by education level.

Trial registration: ClinicalTrials.gov Identifier: NCT05432271.

Conflict of interest statement

Conflict of Interest Disclosures: Dr Stephens-Shields reported personal fees from the American College of Physicians and WilmerHale paid on behalf of Gilead Sciences outside the submitted work. No other disclosures were reported.

Figures

Figure 1.. Labeling Systems Tested and Example Product Stimulus Used

Participants were shown branded items with the labels shown in the figure.

Figure 2.. Participant Flow Diagram

aThe time of 191.2 seconds indicates one-third of the median completion time in a soft launch of the survey. Likelihood of exclusion due to completing too quickly or requesting data be excluded in debrief did not differ across trial arms.

Figure 3.. Effect of Interpretative Front-of-Package Labels on Beverage and Snack Selections

Error bars indicate 95% CIs. a-eMeans and percentages that do not share a superscript letter (including the absence of a letter) with one another are statistically different (Holm-Bonferroni–corrected P < .05) from one another in their effect on the outcomes (4 tests per outcome to compare each interpretative label to control and 6 tests per outcome to compare the 4 interpretative labels with one another).