Fig 1.
Front-of-package labels used in experiment.
Note. Labels listed above represent the version of each label used on the yogurt (excess sugar).
Table 1.
Product nutrition details and label applied to each product*.
Fig 2.
CONSORT flow diagram.
Table 2.
Socio-demographic characteristics of the sample (n = 8,004).
Fig 3.
Predicted percent, by label condition.
Note. * p<0.001 compared to nutrient warning (adjusted for 19 comparisons using Holm’s sequentially rejective method). Data were analyzed by logistic regressions of the outcomes on indicator variables for the arm. Inference on the predicted percent is based on the delta method. Missing data at the participant level were as follows: 40 (0.5%) for ‘Wanted to purchase the less healthy fruit drink’ (11 in nutrient warning, 12 in no label, 9 in Nutri-score, and 8 in GDA), 16 (0.2%) for ‘Correctly identified the less healthy fruit drink as higher in sugar’ (4 in nutrient warning, 3 in no label, 8 in Nutri-score, and 1 in GDA), and 26 (0.3%) for ‘Correctly identified the less healthy fruit drink as less healthy’ (4 in nutrient warning, 15 in no label, 4 in Nutri-score, and 3 in GDA).
Table 3.
Predicted percent and predicted means of secondary outcomes by label type.
Fig 4.
Percent of participants selecting each label type as the most discouraging.
Note. * p<0.001 compared to nutrient warning (adjusted for 19 comparisons using Holm’s sequentially rejective method). Data are estimated proportions and their logit-transformed 95% confidence intervals. Missing data at the participant level were as follows: 131 (1.6%) (22 in nutrient warning, 27 in no label, 39 in Nutri-score, and 43 in GDA).