Results

Sociodemographic characteristics of the study participants

Table 1 presents the sociodemographic characteristics of participants across the 5 experimental arms (control, HSR, WL, GDA, and MTL), each comprising ∼1230 to 1244 respondents. The study arms were largely comparable across key characteristics (age, sex, and area of residence). Adult educational attainment differed modestly across arms, whereas adolescent education levels were comparable. Baseline purchase behavior was comparable across arms for all products except chips among adults (Supplemental Table 2).

Correct identification of the nutrient of concern

In the control (barcode) arm, correct identification of nutrients of concern (HFSS) was low among both adolescents (14%; 95% CI: 11.4, 16.9) and adults (15%; 95% CI: 12.3, 17.9) (Figure 3). All FOPL formats significantly improved the accuracy of correct identification compared with the control (P < 0.001 for all comparisons). WL demonstrated the highest effectiveness, with 73% (95% CI: 69.0, 76.0) of adolescents and 68% (95% CI: 64.4, 71.7) of adults correctly identifying unhealthy products. MTL labels ranked second, achieving correct identification rates of 65% (95% CI: 61.4, 68.9) among adolescents and 62% (95% CI: 58.3, 65.9) among adults. GDA labels resulted in correct identification among 53% (95% CI: 49.4, 57.2) of adolescents and 49% (95% CI: 44.8, 52.7) of adults, whereas HSR labels showed more modest improvements at 29% (95% CI: 25.4, 32.6) for adolescents and 29% (95% CI: 25.2, 32.2) for adults.

FIGURE 3: Percent of participants who correctly identified that products were high in nutrient(s) of concern by study arm. GDA, guideline daily allowance; HSR, health star rating; MTL, multiple traffic light; WL, warning label.

Exploring how each FOPL format compared with other FOPLs on participants’ ability to identify HFSS, the pattern of results revealed the strongest effect for WL, with the HSR performing worst. Among both adolescents and adults, the WL increased correct identification of HFSS compared with the GDA, the MTL, and the HSR (P < 0.05 for all comparisons). Among adolescents and adults, the MTL also increased correct identification of HFSS compared with the GDA and the HSR (P < 0.05). Among adolescents and adults, the GDA increased the correct identification of HFSS compared with the control (P < 0.05).

Intentions to purchase

All FOPL formats significantly reduced intentions to purchase compared with the control (P < 0.05 for all comparisons) (Figure 4). HFSS food purchase intentions were highest under the control (adolescents 89%, adults 80%) and HSR labels (adolescents 87%, adults 81%), indicating that these formats exert minimal influence in discouraging product choice. In contrast, interpretive labeling systems, particularly WL (adolescents 75%, adults 69%) and MTL (adolescents 75%, adults 73%), reduced intentions to purchase. The GDA format yielded intermediate effects (adolescents 81%, adults 76%), likely reflecting its reliance on numerical information rather than interpretive cues.

FIGURE 4: Percentage of participants who intend to purchase the products after intervention.

Secondary outcomes

Table 2 summarizes secondary psychological outcomes among adolescents and adults by FOPL format. All FOPL formats (HSR, WL, GDA, and MTL) produced significantly higher mean scores across all outcomes compared with the control in both age groups (all P < 0.001). Among adolescents, WL and MTL showed the highest scores for perceptions of unhealthiness, concern about health problems, PME, and cognitive elaboration. Similar trends were observed among adults, with WL yielding the strongest effects, followed by MTL. Thus, FOPL exposure increased negative health perceptions and message processing relative to no label, with interpretive labels showing stronger effects.

 | Control mean (95% CI) | HSR mean (95% CI) | P value1 | WL mean (95% CI) | P value1 | GDA mean (95% CI) | P value1 | MTL mean (95% CI) | P value1
Adolescents
Perceptions of unhealthiness | 1.7 (1.7, 1.8) | 2.2 (2.1, 2.3) | <0.001 | 3.0 (2.9, 3.0) | <0.001 | 2.7 (2.6, 2.7) | <0.001 | 2.9 (2.8, 3.0) | <0.001
Attention | 3.1 (3.1, 3.2) | 3.7 (3.6, 3.7) | <0.001 | 3.8 (3.8, 3.9) | <0.001 | 3.8 (3.7, 3.8) | <0.001 | 3.8 (3.8, 3.9) | <0.001
Concern about health problems | 1.3 (1.2, 1.4) | 2.3 (2.2, 2.4) | <0.001 | 3.2 (3.1, 3.3) | <0.001 | 2.7 (2.6, 2.8) | <0.001 | 3.1 (3, 3.2) | <0.001
Seem Unpleasant | 1.2 (1.1, 1.3) | 2 (1.9, 2.1) | <0.001 | 2.9 (2.8, 3.0) | <0.001 | 2.4 (2.3, 2.5) | <0.001 | 2.7 (2.6, 2.8) | <0.001
Discouraged consumption | 1.2 (1.1, 1.3) | 2.1 (2, 2.2) | <0.001 | 3.0 (2.9, 3.1) | <0.001 | 2.5 (2.4, 2.6) | <0.001 | 2.8 (2.7, 2.9) | <0.001
PME2 | 1.2 (1.2, 1.3) | 2.1 (2.1, 2.2) | <0.001 | 3 (2.9, 3.1) | <0.001 | 2.5 (2.5, 2.6) | <0.001 | 2.8 (2.8, 2.9) | <0.001
Cognitive elaboration | 1.4 (1.3, 1.5) | 2.4 (2.3, 2.5) | <0.001 | 3.2 (3.2, 3.3) | <0.001 | 2.8 (2.8, 2.9) | <0.001 | 3.2 (3.1, 3.2) | <0.001
Adults
Perceptions of unhealthiness | 1.9 (1.8, 1.9) | 2.3 (2.2, 2.4) | <0.001 | 3.0 (2.9, 3.1) | <0.001 | 2.7 (2.6, 2.7) | <0.001 | 2.9 (2.8, 3.0) | <0.001
Attention | 3.2 (3.1, 3.2) | 3.7 (3.6, 3.7) | <0.001 | 3.8 (3.8, 3.9) | <0.001 | 3.7 (3.7, 3.8) | <0.001 | 3.8 (3.8, 3.9) | <0.001
Concern about health problems | 1.3 (1.2, 1.4) | 2.4 (2.3, 2.4) | <0.001 | 3.3 (3.2, 3.3) | <0.001 | 2.8 (2.7, 2.9) | <0.001 | 3.1 (3.0, 3.2) | <0.001
Seem Unpleasant | 1.2 (1.1, 1.3) | 2.1 (2.0, 2.2) | <0.001 | 2.9 (2.8, 3.0) | <0.001 | 2.4 (2.4, 2.5) | <0.001 | 2.8 (2.7, 2.9) | <0.001
Discouraged consumption | 1.3 (1.2, 1.4) | 2.1 (2.0, 2.2) | <0.001 | 3.0 (2.9, 3.1) | <0.001 | 2.6 (2.5, 2.7) | <0.001 | 2.9 (2.8, 3.0) | <0.001
PME2 | 1.3 (1.2, 1.4) | 2.2 (2.1, 2.3) | <0.001 | 3.1 (3, 3.2) | <0.001 | 2.6 (2.5, 2.7) | <0.001 | 2.9 (2.9, 3.0) | <0.001
Cognitive elaboration | 1.3 (1.3, 1.4) | 2.5 (2.4, 2.6) | <0.001 | 3.3 (3.2, 3.4) | <0.001 | 2.8 (2.8, 2.9) | <0.001 | 3.2 (3.1, 3.2) | <0.001

Tertiary outcomes

Across both adolescents and adults, all FOPL formats performed substantially better than the control across all tertiary outcomes (all P < 0.001) (Table 3). Interpretive labels (WL and MTL) consistently showed the strongest responses for almost all tertiary outcomes, followed by GDA and HSR. Compared with the control, exposure to any FOPL format markedly increased perceived label understanding, learning something new, perceived truthfulness, and desire to have the label on products. WL was most frequently identified as the label that discouraged consumption and elicited the highest levels of health concern and perceived unpleasantness, with MTL showing similar but slightly lower effects. These patterns were highly consistent across adolescents and adults, indicating robust effects of interpretive labeling on attention, comprehension, and perceived risk.

 | Control percent(95% CI) | HSR percent(95% CI) | P value1 | WL percent(95% CI) | P value1 | GDA percent(95% CI) | P value1 | MTL percent(95% CI) | P value1
Adolescents
Label understanding | 33.1 (29.4, 36.9) | 83.4 (80.2, 86.1) | <0.001 | 97.9 (96.5, 98.8) | <0.001 | 92.2 (89.8, 94.0) | <0.001 | 97.6 (96.0, 98.5) | <0.001
Learning something new | 35.7 (32.0, 39.6) | 83.5 (80.4, 86.3) | <0.001 | 98.1 (96.6, 98.9) | <0.001 | 93.1 (90.9, 94.9) | <0.001 | 98.1 (96.6, 98.9) | <0.001
Label seems true | 52.7 (48.7, 56.6) | 85.7 (82.6, 88.2) | <0.001 | 97.9 (96.5, 98.8) | <0.001 | 94.4 (92.3, 96.0) | <0.001 | 99.2 (98.1, 99.7) | <0.001
Want the label on product | 48.3 (44.3, 52.2) | 82 (78.8, 84.9) | <0.001 | 96.0 (94.1, 97.3) | <0.001 | 92.8 (90.5, 94.6) | <0.001 | 98.5 (97.2, 99.2) | <0.001
Which label most discourages consumption | 9.2 (7.1, 11.7) | 50.6 (46.6, 54.5) | <0.001 | 80.8 (77.5, 83.7) | <0.001 | 62.3 (58.4, 66.0) | <0.001 | 74.8 (71.2, 78.0) | <0.001
Grab attention | 81.2 (77.9, 84.1) | 98.8 (97.6, 99.4) | <0.001 | 98.7 (97.5, 99.4) | <0.001 | 99.2 (98.1, 99.7) | <0.001 | 99.2 (98.1, 99.7) | <0.001
Make concerned about the health consequences. | 12.3 (9.9, 15.1) | 57.3 (53.4, 61.2) | <0.001 | 85.6 (82.6, 88.1) | <0.001 | 69.3 (65.6, 72.8) | <0.001 | 82.7 (79.5, 85.5) | <0.001
Seem unpleasant | 9.8 (7.7, 12.4) | 45.1 (41.2, 49.1) | <0.001 | 76.0 (72.5, 79.2) | <0.001 | 58.5 (54.6, 62.3) | <0.001 | 68.3 (64.5, 71.8) | <0.001
Adults
Label understanding | 33.8 (30.1, 37.6) | 80.7 (77.5, 83.6) | <0.001 | 95.2 (93.2, 96.6) | <0.001 | 87.8 (85.0, 90.2) | <0.001 | 96.8 (95.0, 97.9) | <0.001
Learning something new | 49.1 (45.2, 53.0) | 85.7 (82.7, 88.2) | <0.001 | 96.4 (94.7, 97.6) | <0.001 | 89.3 (86.6, 91.5) | <0.001 | 97.3 (95.6, 98.3) | <0.001
Label seems true | 42.8 (39.0, 46.7) | 80.9 (77.6, 83.8) | <0.001 | 95.5 (93.5, 96.9) | <0.001 | 88.8 (86.0, 91.1) | <0.001 | 97.9 (96.4, 98.8) | <0.001
Want the label on product | 10.7 (8.5, 13.4) | 50.1 (46.2, 54.0) | <0.001 | 81.1 (77.8, 84.0) | <0.001 | 65.9 (62.0, 69.6) | <0.001 | 78.0 (74.6, 81.1) | <0.001
Which label most discourages consumption | 15.0 (12.4, 18.1) | 64.0 (60.2, 67.7) | <0.001 | 90.6 (88.1, 92.7) | <0.001 | 77.1 (73.6, 80.3) | <0.001 | 87.7 (84.9, 90.1) | <0.001
Grab attention | 85.1 (82.1, 87.7) | 99.4 (98.3, 99.8) | <0.001 | 99.5 (98.5, 99.8) | <0.001 | 99.8 (98.8, 100) | <0.001 | 99.5 (98.5, 99.8) | <0.001
Make concerned about the health consequences | 12.3 (9.9, 15.1) | 58.3 (54.3, 62.1) | <0.001 | 87.4 (84.5, 89.8) | <0.001 | 71.7 (67.9, 75.1) | <0.001 | 85.0 (81.9, 87.6) | <0.001
Seem unpleasant | 10.0 (7.9, 12.6) | 47.5 (43.6, 51.4) | <0.001 | 77.1 (73.6, 80.2) | <0.001 | 60.8 (56.8, 64.6) | <0.001 | 73.2 (69.6, 76.5) | <0.001

Sociodemographic differences in correct identification under FOPL

Supplemental Tables 3 and 4 show that all 4 FOPLs (HSR, WL, GDA, and MTL) significantly improved correct identification of products high in nutrients of concern compared with the control across nearly all sociodemographic subgroups (P < 0.001). Among all labels, WL consistently demonstrated the strongest performance in both adolescents and adults, in product identification, followed by MTL. Among adolescents and adults, WL remained almost equally effective across both sexes. Regarding the area of residence, rural adolescents demonstrated slightly higher WL effectiveness than urban peers (74.7% compared with 69.5%), whereas the pattern reversed among adults. In urban adult residents, the WL was substantially more effective than the control (72.5% in WL compared with 18.6% in control, P < 0.001), whereas in rural residents, the WL was slightly less effective (65.2% in WL compared with 12.4% in control, P < 0.001).

In our additional stratified models using 0 to 4 y of schooling as a proxy for illiteracy, we found that participants with 0 to 4 y of schooling consistently showed similar patterns of findings as those with ≥5 y of schooling (Supplemental Table 5).

RR between the different FOPLs

Supplemental Figures 3 and 4 illustrate the relative likelihood (unadjusted relative risk, RR) of correctly identifying nutrients of concern (e.g., high sugar, sodium, saturated fat, or calories) across different FOPL systems. An RR >1.0 indicates a higher probability of correct identification than the reference label, which therefore represents the less effective format. Cluster-adjusted weighted analysis (Supplemental Figures 5 and 6) showed similar findings (<5% variation in RR for most cases) and the direction of association as well as the statistical significance levels remained unchanged. This illustrated the negligible effect of cluster on the outcome estimates. The confounders (age, sex, area of residence, education, and financial condition) adjusted models produced comparable estimates for almost all the cases (Supplemental Tables 6–8).

WL compared with GDA

Across all products and nutrients, WLs were associated with substantially higher relative risks of correct identification (RR > 1.0). For example, adolescents were more likely to correctly identify high-sugar content in cake (RR: 1.24; 95% CI: 1.17, 1.31), and adults were more likely to identify high-saturated fat in salty biscuit (Olympia Lexus) (RR: 1.27; 95% CI: 1.18, 1.37). These findings indicate that WLs are markedly more effective than GDA labels in enabling both adolescents and adults to recognize nutrients of concern.

MTL compared with GDA

The MTL system consistently showed higher RRs of correct identification across products. For instance, adolescents were more likely to identify high-sugar content in cake (RR: 1.17; 95% CI: 1.10, 1.24), and adults were more likely to identify high-saturated fat in salty biscuit (Olympia Lexus) (RR: 1.19; 95% CI: 1.10, 1.28). These results suggest that the color-coded, interpretive nature of MTL labels facilitates better understanding than numerical GDA labels. However, the magnitude of effect was generally smaller than that observed for WLs, indicating that although MTLs are more effective than GDAs, WLs remain the most effective approach for communicating nutrients of concern.

WL compared with MTL

The WL system has a small advantage over MTL among adolescent consumers. The relative risk is >1.0, indicating that adolescents exposed to WL had a higher probability of correctly identifying certain products with excessive nutrients of concern than those exposed to MTL labels. Only in the case of high-sugar (cake and sweet biscuit) RRs did not touch 1.00, which means WL shows more probability of correctly identifying excessive nutrients of concern than those exposed to MTL labels in these products. For example, identification of high sugar in cake RR [95% CI: 1.06 (1.02, 1.11)] and high sugar in sweet biscuits RR [95% CI: 1.04 (1.01, 1.08)]. In practical terms, this means the simplistic, salient, and negatively framed nature of WL (e.g., a black stop signs) is more effective for this age group than the color-coded MTL system.

Similar to the adolescent results, Supplemental Figure 3B shows that WL and MTL labels have a relatively same effectiveness for adult consumers. The data show that adults using WL had only a bit higher relative likelihood of correctly identifying high-saturated fat [RR 95% CI: 1.07 (1.01, 1.14)] in salty biscuit (Olympia Lexus) than those using MTL.