Experimental design \& procedures
The aim of this study is to use a 3-arm randomized controlled trial with an online grocery store (NUSMart), to rigorously evaluate the effect of the two FOP labels, Chilean warning labels and Nutri-Score, on diet quality.
1. Food labels
The investigators tested 2 different types of food labels in this study. They are as follows:
1. Chilean Warning Labels: These labels were implemented in Chile in 2016. They are shown on those less-healthy food and beverage products that don't fall within thresholds that are defined for the following nutrients: sugar, sodium, calories and saturated fats. The original labels are in Spanish but the ones that the investigators displayed were in Arabic.
2. Nutri-Score Labels: These labels were implemented by the French government in 2017. They are shown on all food and beverage products and range from A to E with the corresponding colours moving from Green to Red (symbolizing best to worst). The scores and grades are calculated based on 7 different nutrients grouped into 2 categories i.e. nutrients to consume (such as fibre) and nutrients to avoid (such as sugar). The investigators modified the labels by adding % sugar per serving on the right side of the labels.
2. Overview of RCT Design
To test these labels, the investigators used three different versions of NUSMart. Each participant was randomly assigned to one of the following arms and asked to complete a one-time hypothetical shop.
Arm 1: Control arm. Participants experienced an unmodified version of NUSMart, which replicates the traditional shopping experience of online grocery stores.
Arm 2: Similar to Arm 1 except that products were labeled with Chilean warning FOP labels.
Arm 3: Similar to Arm 1 except that products were labeled with NutriScore FOP labels.
To minimize hypothetical shopping biases, the investigators set a minimum spending value per person per week and adjusted the total minimum spending value depending on household size.
3. Subject related procedures
As the investigators have done in prior studies, participants were recruited if they were KSA residents aged 18 years or older, could speak Arabic and were primary weekly shoppers for their households. Recruitment was done using a Saudi Arabia based web panelist.
On recruitment, they were asked to fill in a baseline questionnaire to collect demographic and health characteristics, which took approximately 10-15 minutes to complete.
The baseline questionnaire includes a question as to whether any household members have a medical condition, such as diabetes or hypertension, which requires limiting the types of foods they eat. The investigators asked this question to allow for testing whether the interventions differentially influence these households, with the expectation that households with less healthy patients may obtain greater benefits from purchasing baskets with higher mean Nutri-Scores. Since the objective of this RCT is to quantify the effectiveness of the interventions on diet quality, a precursor to NCDs, the collection of household health indicators is reasonable.
Upon completion of the baseline survey, participants were randomly assigned to one of the 3 shopping conditions (Arm 1 \[Control\], Arm 2, or Arm 3). They logged into the NUSMart website and were asked to purchase their weekly groceries with a minimum spend per person per week (the total minimum spending value depends on the household size). Each participant therefore shopped only once during the study in one of the arms (1,2 or 3) which was randomly assigned to them.
Participants were given explicit instructions to purchase groceries for the purposes of breakfast, lunch, dinner, snacks and beverage consumption for one week keeping in mind that they would not be consuming any meals outside. The stores were set up such that participants were not allowed to check out with their cart unless they selected items from at least 4 different NUSMart categories. This ensures that participants shop as though they are shopping for their own consumption i.e. to minimize hypothetical biases.
A post-study survey was conducted after the last shop to collect participant opinions on the labels and understand their impact better.
