Restaurant Menu Labeling: Added-sugar Labels and Labeling Thresholds

The main goal of this study is to compare two added-sugar menu labels--(1) red, icon-only, 150% of the menu text height and (2) black, icon-plus-text, 100% of the menu text height--to a no-label control on grams of added sugar ordered using two different added-sugar labeling thresholds: (1) ≥50% of Daily Value for added sugar (equivalent to ≥25g) and (2) ≥100% of Daily Value for added sugar (equivalent to ≥50g). This results in 5 conditions: (1) icon-only at 50% threshold, (2) icon-plus-text at 50% threshold, (3) icon-only at 100% threshold, (4) icon-plus-text at 100% threshold, and (5) no-label control.

The two primary outcomes are (1) grams of added sugar ordered averaged across a fast-food and full-service restaurant menus and (2) grams of added sugar ordered from a café menu. The secondary outcomes are (1) noticing a high added sugar nutrition label and (2) knowledge of items' added sugar content.

For continuous outcomes (e.g., grams of added sugar, knowledge), linear regression models regressing the outcome on an indicator for experimental condition will be used. If the outcomes are not normally distributed, alternative modeling approaches or data transformations will be explored. For dichotomous outcomes (e.g., noticing the added-sugar label), the probability ratio will be modeled using Poisson regression with a robust error variance, regressing the outcome on an indicator for experimental condition.

Exploration of effect modification of the labels on primary outcomes by type 2 diabetes status and sociodemographic variables important for health equity will be conducted using interaction terms and stratified models.

Exploratory process outcomes include (1) selection of at least one item that contains ≥50% of the Daily Value of added sugar and (2) selection of at least one item that contains ≥100% of Daily Value of added sugar. Other exploratory outcomes include the perception that the labels grabbed attention and support for an added sugars warning label policy. Additionally, differences in the primary outcomes between each of the 5 unique conditions will be examined. If there is evidence of interactions, the modeling approach will be modified.