"Warning: ultra-processed": an online experiment examining the impact of ultra-processed warning labels on consumers' product perceptions and behavioral intentions · D'Angelo Campos A, Ng SW, Duran AC et al.

Discussion

In this online study with Brazilian adults, we experimentally examined whether adding a label warning consumers that a product is ultra-processed to Brazil’s existing nutrient warning labels influenced participants’ ability to identify the product as ultra-processed, their product perceptions, and behavioral intentions. The addition of the ultra-processed warning label led to a higher share of participants being able to identify UPFs compared to nutrient warning labels alone. On the other hand, the addition of the ultra-processed warning label did not influence participants’ intentions to purchase the products, how healthful they perceived the products to be, or how effective they perceived the labels to be. To contextualize our experimental findings, we also qualitatively explored underlying understandings of the meaning of the term “ultra-processed” in the sample. Most participants exhibited a moderate understanding of the UPF concept.

Our results suggest that front-of-package ultra-processed warning labels could help consumers more easily identify which products are UPFs, which is a key initial step toward better-informed consumer choices. The moderate understanding of the term “ultra-processed” observed in our sample underscores the importance of providing consumers with this sort of practical guidance for navigating a food supply abundant in UPFs. Along similar lines, previous studies in Brazil and other Latin American countries have shown that, although consumers can sometimes distinguish between unprocessed and ultra-processed foods, most have difficulty with foods that fall in intermediate categories (e.g., processed culinary ingredients and processed foods) [54–56], which are an important part of and facilitate healthy dietary patterns. However, it is worth noting that although the presence of the ultra-processed warning label improved participants’ ability to identify UPFs, 42% still failed at this identification, suggesting that complementary strategies such as public health campaigns may also be necessary to improve consumer literacy around UPFs. Additionally, our sample had a high level of educational attainment compared to the general Brazilian population, so it would be important determine whether ultra-processed warning labels would also help consumers with lower educational attainment identify UPFs.

On the other hand, our finding that ultra-processed warning labels did not influence participants’ product perceptions or behavioral intentions differs from previous studies on ultra-processed food labels, which observed effects on these outcomes when comparing ultra-processed labels to control [34, 35]. This difference suggests that, although ultra-processed labels may have an influence on consumers when compared to the absence of any FOPLs, their influence may not go beyond the influence already exerted by nutrient labels. A possible explanation may be that the difference between nutrient content and degree of processing is not evident to consumers, in which case the added ultra-processed warning label may have seemed redundant. While there is currently no scientific consensus about the importance of this distinction, emerging epidemiological evidence suggests that level of processing may matter for health independently from nutrient content [18–24] – therefore, future health communication research should examine whether results would differ when making this distinction clearer. Additionally, different UPF labeling approaches would be also worth exploring in future research. One such approach recently proposed in the United States Senate are health warning labels – i.e., labels warning consumers about the possible health effects of UPF consumption rather than simply identifying a product as ultra-processed [57]. Another approach would be signaling specific non-nutrient related UPF characteristics, such as the presence of food additives, which may be more easily understood or more effectively impact consumers’ health risk perceptions than the broader and more complex term “ultra-processed.” An example of this approach can be found in Mexico and Colombia’s FOPL systems, which currently signal the presence of non-sugar sweeteners [58].

Lastly, it is crucial to consider that, in countries adopting nutritional warning label schemes like Brazil and other nations in the Americas, incorporating ultra-processed warning labels could result in many products that do not display nutrient warnings (because they do not exceed pre-determined amounts of nutrients of concern) being obligated to carry ultra-processed warnings. For instance, a study analyzing the 80 products that constitute a food basket in Chile found that 33% those classified as UPFs based on the Nova system did not carry any nutrient warnings [59]. Similarly, a recent study examining a nationally representative sample of packaged foods purchased in the US found that a Nova-based approach would target 50% of products for intervention, while an exclusively nutrient-based approach would only target 43% of products and thus miss an important share of UPFs [60]. Recent studies also indicate that, after the implementation of nutrient-based FOPLs, manufacturers tend to reformulate several products to avoid mandatory labels – and while such reformulations generally improve the nutrient content of products, they often result in manufacturers incorporating additives, such as non-sugar sweeteners, that can categorize products as UPFs [61–63]. To maximize power with our pre-determined sample size, this study did not examine the potential benefits offered by ultra-processed warning labels in these scenarios, which merit further investigation.

This study’s strengths include an experimental design allowing for causal inference, the use of professionally developed stimuli featuring multiple types of real products, and a diverse national sample. However, this study also has limitations. As an online experiment with a convenience sample, we cannot establish how generalizable our findings would be in real-world settings – and although previous evidence suggests that online convenience samples tend to produce experimental results similar in direction to nationally representative samples [64, 65], our sample’s small share of participants with low educational attainment may have specifically precluded us from findings differences in the effectiveness of ultra-processed warning labels by education. The experiment exposed participants to only four products, so we cannot establish how generalizable our findings would be to a broader set of products. Additionally, given a lack of applicable measures validated for Brazilian samples specifically, we relied on measures validated in samples from other countries and commonly used in food labeling studies around the world [45–48, 66]. The study also did not inform participants about the meaning and intended use of ultra-processed warning labels prior to label exposure, which could influence effectiveness in a real-world implementation scenario. Lastly, to increase realism, we designed rectangular ultra-processed labels that would be compatible with and proportional in size to Brazil’s current FOPL visual identity, but alternative design elements such as octagonal or triangular shape [38, 67–72] and a larger size could possibly heighten the effectiveness of ultra-processed warning labels.