A Systematic Review of Worldwide Consumption of Ultra-Processed Foods: Findings and Criticisms

A debate is ongoing on the significance and appropriateness of the NOVA classification as a tool for categorizing foods based on their degree of processing. As such, the role of ultra-processed food (UPF) on human health is still not completely understood. With this review, we aimed to investigate the actual level of consumption of UPF across countries and target populations to determine the impact in real contexts. Suitable articles published up to March 2021 were sourced through the PubMed and SCOPUS databases. Overall, 99 studies providing data on the level of UPF consumption expressed as the percentage of total energy intake were identified, for a total of 1,378,454 participants. Most of them were published in Brazil (n = 38) and the United States (n = 15), and the 24 h recall was the most-used tool (n = 63). Analysis of the results revealed that the United States and the United Kingdom were the countries with the highest percent energy intake from UPF (generally >50%), whereas Italy had the lowest levels (about 10%); the latter was inversely associated with adherence to the Mediterranean diet. High variability was also observed based on sex, age, and body mass index, with men, young people, and overweight/obese subjects generally having higher levels of consumption compared to older subjects. Overall, our findings underline the large differences in UPF intake. Since most of the observations derived from studies conducted with food questionnaires are not specifically validated for UPF, further efforts are essential to confirm the results previously obtained and to investigate further the association between UPF consumption and health status, also considering the actual contribution within different dietary patterns, which has been less investigated to date.


Introduction
Food processing includes all the processes needed to transform raw or harvested foodstuffs into new products, ensuring their safety, palatability, and shelf-life [1]. However, in recent decades, food processing has drastically changed to address consumer preferences. The demand for food items with longer shelf-life and improved palatability has led to other natural or artificial ingredients being added to processed foods, which may, to some extent, impact the nutritional quality of these foods, which are often characterized by high sugar, fat, and salt contents [2]. For this reason, interest is growing in elucidating whether the high consumption of these foods negatively impacts diet quality and, in turn, human health.
One of the main limits to the evaluation of the role and impact of food processing on health status is the lack of a proper definition and classification of "processed food". In 2009, the European Prospective Investigation into Cancer and Nutrition (EPIC) [3] proposed a
The number of studies performed in different countries is reported in Figure 2. As shown, Brazil had the highest number of studies (n = 38), followed by the United States (n = 15), France (n = 8), the United Kingdom (n = 6), Canada (n = 6), and Spain (n = 4). Fewer studies were performed in Malaysia, (n = 3), and in other countries such as Mexico, Italy, Australia, Korea, and Portugal (n = 2); only one study each was conducted in Chile, Japan, Indonesia, Lebanon, Israel, the Netherlands, Colombia, Belgium, and New Zealand. -Not pertinent (n = 67) -Not providing data expressed as quantity of UPF (g/day) and/or % energy from UPF (n = 23) -Not providing data for each NOVA class (n = 1) -Not in English (n = 4) -No full-text (n = 2)
The number of studies performed in different countries is reported in Figure 2. As shown, Brazil had the highest number of studies (n = 38), followed by the United States (n = 15), France (n = 8), the United Kingdom (n = 6), Canada (n = 6), and Spain (n = 4). Fewer studies were performed in Malaysia, (n = 3), and in other countries such as Mexico, Italy, Australia, Korea, and Portugal (n = 2); only one study each was conducted in Chile, Japan, Indonesia, Lebanon, Israel, the Netherlands, Colombia, Belgium, and New Zealand. Regarding the method of dietary assessment, the 24 h recall was the most-used tool (n = 62), followed by food-frequency questionnaires (FFQs) (n = 28) and food diaries (n = 10).

Levels of UPF Intake
The levels of consumption of UPF shown in the retrieved studies are reported in Table 1, while results stratified for specific target groups are provided in Tables 2-4.
The data revealed a high variability in terms of the percentage (%) of energy provided by the consumption of food belonging to group 4 based on the NOVA classification. Similar findings were observed both intra-and inter-country. The highest levels of consumption were observed in the United States and the United Kingdom with the percent of energy intake generally higher than 50% with respect to TEI [18,19,36,38,[46][47][48][49][50][51][52][53][54]. Conversely, the lowest levels were observed in Italy, in which the two studies identified reported about 10% energy obtained from UPF [55,56].
Regarding the variability within the same country, Brazil showed levels of UPF consumption ranging from 7.7% in 64 subjects aged 25-57 years [57] to 51.2% of total energy intake in over 4200 subjects [58]. A lower variability was identified in studies performed in the United States, with levels of intake ranging from 50% to 70%, with the only exception being a study observing an energy intake of 35.5% in almost 92,000 subjects aged 55-74 years [59]. Regarding the method of dietary assessment, the 24 h recall was the most-used tool (n = 62), followed by food-frequency questionnaires (FFQs) (n = 28) and food diaries (n = 10).

Levels of UPF Intake
The levels of consumption of UPF shown in the retrieved studies are reported in Table 1, while results stratified for specific target groups are provided in Tables 2-4. The data revealed a high variability in terms of the percentage (%) of energy provided by the consumption of food belonging to group 4 based on the NOVA classification. Similar findings were observed both intra-and inter-country. The highest levels of consumption were observed in the United States and the United Kingdom with the percent of energy intake generally higher than 50% with respect to TEI [18,19,36,38,[46][47][48][49][50][51][52][53][54]. Conversely, the lowest levels were observed in Italy, in which the two studies identified reported about 10% energy obtained from UPF [55,56].
Regarding the variability within the same country, Brazil showed levels of UPF consumption ranging from 7.7% in 64 subjects aged 25-57 years [57] to 51.2% of total energy intake in over 4200 subjects [58]. A lower variability was identified in studies performed in the United States, with levels of intake ranging from 50% to 70%, with the only exception being a study observing an energy intake of 35.5% in almost 92,000 subjects aged 55-74 years [59]. Table 1. Characteristics of the selected studies (n = 106) and the level of consumption of ultra-processed foods (UPF) expressed as the percent of energy provided by UPF intake with respect to total energy intake (TEI). Several studies stratified data based on different characteristics, such as age, sex, and BMI. The level of UPF consumption stratified for these parameters is reported in Tables 2-4. Slight differences between the sexes were observed in UPF intake, with men having often an overall higher intake compared to women [48,54,87,95,96,111]. However, in most of the studies, the levels of UPF intake appeared comparable [14,30,49,51,81,83,94,112] ( Table 2). Regarding age, studies generally reported large variations amongst the age groups. However, UPF intake generally decreased with increasing age, with the highest levels of UPF intake observed in children and adolescents, and the lowest in older subjects [51,54,87,104,108,110,112,118] (Table 3). Only five studies also stratified results based on body mass index (BMI) of the participants, generally finding a slight, but higher, UPF intake in subjects with the highest BMI (e.g., greater than 30 kg/m 2 ) [58,87,89,95] ( Table 4) [58]. Table 2. Level of consumption of ultra-processed foods (UPF) expressed as percent energy provided by UPF intake with respect to total energy intake (TEI) considering sex.

Author (Year) Sex UPF Consumption and Statistic
Adams and White (2015)    Data are reported as mean ± standard deviation (SD) or standard error of the mean (SEM) *; CI, confidence interval; F, female; M, male; ND, not determined or reported; UPF, ultra-processed food and drink products; TEI, total energy intake. Table 3. Level of consumption of ultra-processed foods (UPF) expressed as percent energy provided by UPF intake with respect to total energy intake (TEI) by considering age.    Data are reported as mean ± standard deviation (SD) or standard error of the mean (SEM) *; CI, confidence interval; mo, months; ND, not determined or reported; UPF, ultra-processed food and drink products; TEI, total energy intake. Table 4. Level of consumption of ultra-processed foods (UPF) expressed as % energy provided by UPF intake with respect to total energy intake (TEI) by considering the BMI factor. Data are reported as mean ± standard deviation (SD) or standard error of the mean (SEM) *; CI, confidence interval; BMI, body mass index; ND, not determined or reported; UPF, ultra-processed food and drink products; TEI, total energy intake.

Author (Year) Body Mass Index (BMI) UPF Consumption for BMI and Statistics
Two studies evaluated the intake of UPF during pregnancy, reporting conflicting results. Silva et al. [37] reported no difference between second and third trimester. Conversely, Gomes et al. [34] reported a significant difference in the second and third trimesters between the intervention and control group. This latter consumed more UPF compared to the intervention group that received training for the application of healthy food practices during prenatal care appointments (see Supplementary Table S1). Finally, Gehring and colleagues [85,86] compared UPF intake in subjects adhering to different dietary patterns, observing a higher UPF intake in vegans and vegetarians than pesco-vegetarians and meat eaters (Supplementary Table S1).

Discussion
Since Monteiro et al. proposed the NOVA classification to categorize foods based on the degree of processing [4], several studies have been conducted to estimate the level of consumption of UPF and its association with several health markers [24,35,52,67,120] as well as with disease risk and mortality [55,59,87,121,122], adjusting the models for energy intake and other potential confounding factors. It was hypothesized that a high level of UPF consumption may represent a health issue, being associated with weight gain and worsening of cardiovascular risk factors such as high waist circumference and low HDL cholesterol [11].
In the present study, we collected 100 unique studies published in 106 manuscripts estimating the UPF levels in different populations from 21 countries around the world. Overall, we found a large variability in the percent of TEI obtained from UPF in the different countries, with the United States and United Kingdom being the countries with the highest percent of TEI from UPF, and Mediterranean countries such as Italy showing the lowest level (~10% of TEI). These results are in line with previous evidence suggesting that adherence to the Mediterranean diet is inversely associated with UPF consumption [20]. This is further confirmed by findings showing that the highest tertiles or quartiles of UPF intake are associated with the lowest adherence to the Mediterranean diet [44,55]. The low levels of UPF consumption registered in Italy and other Mediterranean countries are those associated with the lowest risks for non-communicable diseases. For instance, da Silva et al. recently observed an association between UPF consumption and the increased presence of high waist circumference, overweight, and peripheral arterial disease when comparing the third and first tertiles of the UPF contribution to energy intake in a Brazilian cohort [123]. Intriguingly, the first tertile corresponded to <10.6% of TEI from UPF, which is similar to the levels registered in the study conducted in the Italian population. These levels are far lower than those detected in the first quartiles in the study by Rauber et al., who observed that participants in the highest quartile (>70.3% and >71.7% of TEI from UPF in women and men, respectively) had a significantly higher risk of developing obesity, and of experiencing a ≥5% increase in BMI waist circumference than those belonging to the lowest quartile (<24.1% and <26.3% of TEI from UPF in women and men, respectively) [47]. The type of foods contributing to UPF intake largely varied among countries, but, in accordance with previous findings [124], the most consumed UPF included: baked goods, dairy products, processed fruit and vegetables, and, among drinks, carbonated drinks.
In addition to country, the level of UPF intake was found to be inversely associated with the increase in age. In this regard, children generally showed the highest intake of UPF, which led the European Childhood Obesity Group to "a call to action" aimed to inform people about the potential harmful effects of UPF [118]. For example, it was found that in the United Kingdom, 65% of calories eaten by primary and secondary school children derived from white bread, biscuits, carbonated drinks, crisps, and chips. These findings are in line with the observations reported by others [125,126]. Similarly, the pediatric populations of the United States and Canada reported an intake of UPF above 55% by including breads, cookies, savory snacks, reconstituted meat products, milk-based drinks, breakfast cereals, juices and sodas, and frozen and ready-to-eat meals in the diet [18,96]. Among U.S. school-aged children and adolescents, UPF provided 66.2% and 66.4% of TEI, respectively, with pizzas, sodas, and juices being the most-consumed products [118]. In this scenario, it was proposed that the levels of UPF intake in the young may reflect, at least in part, their lifestyle. In this regard, a recent study documented that subjects consume more UPF when dining out than when eating at home [127]. Other reasons are related to socio-economic inequalities, including lower education status of the mother or unemployed parents, which may lead to a preference for cheaper and less nutritious foods [105]. A different trend was observed for older subjects who showed a lower intake of UPF compared to younger subjects; the main UPF products included cookies and pastries, but also processed breads, breakfast cereals, and yogurts [16,128]. Compared to age, a minor variability was found for sex and/or BMI, which might differ for the net amount of UPF consumption but not for the percent of TEI from UPF. Intriguingly, the adherence to specific dietary patterns represented an additional determinant of the levels of UPF consumption. In this context, based on the NOVA classification, vegans and vegetarians reported higher UPF consumption compared to pesco-vegetarians and meat eaters, mainly driven by a higher consumption of plant-based meat and dairy substitutes. These results highlight the high variability in the characteristics of these types of diet, which may differ widely for the consumption of several food groups [129]. However, these results were found only in two different publications belonging to the same cross-sectional trial performed in the NutriNet-Santé cohort [85,86]; thus, a confirmation of such an analysis deserves further investigation, also to comprehend better if all vegetarian diets have the same health benefits regardless of the levels of UPF consumed.
A thorough comparison of the findings from the different studies considered was challenging due to the differences in food classification and to the disparate definitions that were proposed. Descriptions of UPF within the NOVA system vary with distinguishing features including single vs. 2-3 vs. ≥5 more ingredients, or natural/fresh vs. imitation or industrial, and whole foods vs. fractioned substances [13]. This means that different studies may have classified the same food as UPF or not based on the distinguishing feature used for classifying foods.
Notably, one of the main sources of variation among study protocols is the tools used for estimating UPF intake. Overall, from the present review, it is difficult to provide conclusive findings about the influence of the tools used to determine UPF intake since its estimation was typically performed using a single method. In this regard, most of the studies (n = 49) used the 24 h recall. To reflect the typical diet, this tool needs to be administered several times; however, in some studies, data were derived from a single 24 h recall, which could have affected the significance of the findings. However, this tool has the strength of being able to assess the consumption of all food items since subjects can include/report specific information (e.g., brand), which may help with identifying the NOVA group. However, a lower number of studies assessed UPF intake using FFQs, which are not always specifically created and validated to estimate the consumption of products undergoing different food processing, and the food list cannot cover all the food items consumed, thus leading to underreporting [130]. Among studies using FFQs, some of them used the questionnaire developed within the European Prospective Investigation into Cancer and Nutrition (EPIC) [131], which is not able to distinguish among products belonging to different NOVA groups. This is, for instance, the case for artisanal or industrial breads or cakes that belong to two different NOVA groups (Groups 3 and 4, respectively). Therefore, the use of tools not specifically validated for estimating UPF consumption may potentially lead to the misclassification of foods in the UPF categories, which, in turn, may lead to the misinterpretation of the associations found with markers of health. Only two studies [17,51] used a questionnaire specifically validated for estimating the levels of UPF consumption in children and adults. The approach to validating and using ad hoc developed FFQs, when possible, created for specific populations to consider the different dietary habits, should be recommended for more accurately estimating the consumption of UPF and the actual impact on health-related outcomes.
This study has some strengths worth highlighting, the first of which is the rigorous search and selection strategy that identified available studies examining the energy intake from UPF. However, we cannot exclude that the use of further databases may have allowed the identification of additional studies. Secondly, food processing level was always determined according to the criteria of NOVA classification, to facilitate the comparison amongst findings, although it was reported that some NOVA definitions are open to researchers' interpretation [132], leading to a not-always-uniform categorization [133]. Finally, we included only studies reporting results as the percent of TEI from UPF and not as grams per day. In our opinion, the ratio of energy intake from UPF compared to total energy intake is more useful for reflecting the impact of these products on the whole diet. However, this method may not be useful for detecting the consumption of energy-free products (such as energy-free drinks with artificial sweeteners); thus, this choice can also be considered as a limitation. Notably, the NOVA system used has been largely criticized for different reasons, mostly because this system focuses on the role of food processing, regardless of the nutritional characteristics of foods, on human health [132]. Thus, further efforts should be directed toward elucidating the impact of the different foods belonging to the same NOVA category but with different nutrient profiles on human health. To conclude, it seems worthwhile to implement a critical and constructive discussion able to clarify the potential and applicability of this type of approach.

Conclusions
In conclusion, this review showed high levels of UPF consumption, especially in some countries and in specific target groups (i.e., children and adolescents). However, most of the data on UPF consumption have been derived from FFQs and 24 h dietary recall, which are not specifically validated for estimating UPF; thus, such data should be considered with caution. In this scenario, tools specifically validated to estimate the levels of UPF consumption can be useful to avoid misinterpretation of the findings, especially when used to investigate the association with health status. Despite several studies reporting a positive association between UPF and obesity and cardiometabolic health, on the whole, the evidence is not yet totally convincing. In addition, whether this association is dependent on the nutritional characteristics of UPF and/or related to the applied processing is unclear. In this context, despite the NOVA system classifying foods based on the food processing technology without providing any information about the nutritional content of the food, the UPF group has been suggested to be an indicator of poor food quality due to the generally high amounts of free or added sugars, fats, low levels of fiber, and high energy density. For instance, it was recently observed that UPF consumption is associated with a deterioration in diet quality, with UPF intake being negatively correlated with fiber and protein and positively correlated with sugar, fat, and saturated fat intake [115]. This highlights that the concept of UPF may be somehow misleading, with the effect on human health mediated more by the nutritional quality of products rather than the processing. The association between UPF and nutritional adequacy is not surprising, since, for instance, the presence of added sugar or fat is a major element in defining UPF. Finally, the classification systems based on processing are not often aligned with dietary guidelines (e.g., some products considered UPF are recommended within a balanced diet). Thus, confirmation of the results already obtained should be accompanied by an evaluation of the association between UPF consumption and health status, estimating the contribution within different dietary patterns.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/nu13082778/s1, Supplementary Table S1. Level of consumption of ultra-processed foods (UPF) expressed as percent energy provided by UPF intake in the total energy intake (TEI) in pregnant women and following different dietary patterns. Funding: This research received no external funding.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.