Ultra-Processed Food Consumption and Incidence of Obesity and Cardiometabolic Risk Factors in Adults: A Systematic Review of Prospective Studies

Ultra-processed foods (UPF) are energy-dense, nutritionally unbalanced products, low in fiber but high in saturated fat, salt, and sugar. Recently, UPF consumption has increased likewise the incidence of obesity and cardiometabolic diseases. To highlight a possible relationship, we conducted a systematic review of prospective studies from PubMed and Web of Science investigating the association between UPF consumption and the incidence of obesity and cardiometabolic risk factors. Seventeen studies were selected. Eight evaluated the incidence of general and abdominal obesity, one the incidence of impaired fasting blood glucose, four the incidence of diabetes, two the incidence of dyslipidemia, and only one the incidence of metabolic syndrome. Studies’ quality was assessed according to the Critical Appraisal Checklist for cohort studies proposed by the Joanna Briggs Institute. Substantial agreement emerged among the studies in defining UPF consumption as being associated with the incident risk of general and abdominal obesity. More limited was the evidence on cardiometabolic risk. Nevertheless, most studies reported that UPF consumption as being associated with an increased risk of hypertension, diabetes, and dyslipidemia. In conclusion, evidence supports the existence of a relationship between UPF consumption and the incidence of obesity and cardiometabolic risk. However, further longitudinal studies considering diet quality and changes over time are needed.


Introduction
Obesity is a growing worldwide health problem. It is characterized by excessive adiposity that can compromise health status. According to the World Health Organization, obesity affects more than 1 billion people worldwide, 650 million of whom are adults [1]. Obesity is closely linked with metabolic syndrome [2], defined by the National Institutes of Health as a cluster of interconnected metabolic abnormalities, including central adiposity, dyslipidemia, high blood pressure, and impaired fasting glucose [3]. Both obesity and metabolic syndrome are associated with increased risk for mortality and many non-communicable diseases (NCDs) [2].
Obesity and metabolic syndrome are complex, multifactorial diseases whose causes are not yet fully elucidated. However, it is well known that dietary habits play a crucial role

Search Strategy
Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines (PRISMA) were followed to carry out the study [7]. Studies included in the present review were identified by searching in two electronic databases, including PubMed and Web of Science, using the following search string: (ultraprocessed food* OR ultra-processed food* OR ultra processed food* OR NOVA food*) AND (obesity OR overweight OR waist circumference OR blood pressure OR hypertension OR dyslipidemia OR triglycerides OR cholesterol OR impaired fasting glucose OR diabetes OR metabolic syndrome OR cardiovascular disease OR cardiovascular risk). Electronic search was carried out in September 2022. This systematic review was registered in PROSPERO with registration number CRD42023423112.

Study Selection, Inclusion and Exclusion Criteria
Initially, we proceeded to exclude duplicates. Then, two independent investigators (S.P.M. and S.R.) selected articles based on title and abstract. The selected articles were then evaluated for eligibility. To be included in the present review, articles needed to be original, include healthy participants aged 18 years or older, written in English, have a prospective cohort study design, use NOVA classification to define UPF, and have as outcomes general or central obesity and cardiometabolic risk factors. No country/region/ethnicity nor date restrictions were applied. Cross-sectional and case-control studies were excluded. Studies limiting the evaluation only to a specific food category included in the definition of UPF, such as reconstituted meat products or sugar-sweetened beverages, or that assess household availability or purchase of UPF were excluded. We further excluded metaanalyses, review articles, congress abstracts, letters, and comments. Disagreements in study selection were resolved through consensus or by seeking the opinion of a third investigator (A.L.) if consensus could not be reached.

Data Extraction
From each article, we extracted the following data: main author, country, year of publication, number of participants, outcomes, dietary assessment method, confounding factors, and main results. Two independent investigators (S.P.M. and F.M.) reviewed selected articles and performed data extraction. A third investigator (A.L.) supervised data extraction and solved inconsistencies and disagreements.

Quality Assessment
Two independent investigators (S.P.M. and M.P.) conducted the quality assessment. The Critical Appraisal Checklist for cohort studies proposed by Joanna Briggs Institute was used to assess the methodological quality of the selected studies [16]. The checklist included 11 items related to the following critical domains: population characteristics, follow-up, outcomes, exposure, confounders, and statistical analysis. For each item, it was possible to respond with "no", "yes", "unclear", or "not applicable". Based on the responses, an overall critical assessment of the quality of the study was obtained. In cases where the two investigators disagreed in answering individual items, the opinion of a third investigator (A.L.) was sought. Studies that received a positive score in at least half of the items were considered to be of acceptable quality for inclusion in this Review [9].

Results
A total of 2662 articles were initially found on Pubmed and Web of Science ( Figure 1). We then removed 717 duplicates and discarded an additional 1852 articles based on title and/or abstract, as they were deemed irrelevant to the review. The remaining 93 records were evaluated for eligibility. Of these, 2 articles were not written in English, 42 were review, meta-analysis, editorial, commentary, or congress abstracts, and 32 were original studies but with a study design different from the cohort study (mainly cross-sectional), and therefore were removed. At the end of the evaluation process, 17 studies were included in this systematic review. The quality assessment of the selected studies is shown in Figure 2.
Non-obese subjects with low UPF consumption had a lower risk of developing obesity during follow-up (HR Q4 vs.Q1 = 1.20, 95% CI: 1.08, 1.33, P trend < 0.001) than those with a higher intake. Moreover, a 10% increment of UPFs intake was associated with a higher risk of developing obesity (HR = 1.11, 95% CI: 1.07, 1.15; P < 0.001).  Non-obese participants in the uppermost quartile of UPF consumption were at a higher risk of developing obesity (HR = 1.79, 95% CI: 1.06, 3.03) than participants in the lowest quartile. Similarly, participants with normal waist circumference at baseline but in the first quartile of UPF consumption were at a higher risk of developing abdominal obesity (HR = 1.30, 95% CI: 1.14, 1.48) than participants in the lowest quartile.

Consumption of Ultra-Processed Food, Excess Body Weight, and Abdominal Obesity
Eight cohort studies investigated the relationship between UPF consumption and the risk of weight excess and abdominal obesity, all finding a positive relationship [17,20,21,25,26,[32][33][34]. Four studies focused on the risk of overweight and obesity [17,21,26,33] and two studies on the risk of abdominal obesity [20,25,31], while two other studies investigated both the risk of overweight and obesity and of abdominal obesity [31,32]. Mendonca et al. [26] analyzed data from the SUN cohort, reporting that normal-weight participants consuming higher amounts of UPF, expressed as consumption of servings per day, had a 26% higher risk of developing overweight or obesity during follow-up (HR = 1.26; 95% CI: 1.10, 1.45, P trend = 0.001), than participants with lower UPF consumption. Similarly, data from the ELSA cohort [17] showed that normal-weight participants in the uppermost quartile of UPF consumption had a 20% higher risk of overweight and obesity during follow-up than participants in the lowest quartile (RR = 1.03, 95% CI: 1.0, 1.40). However, no association between UPF consumption and incident risk of obesity was found among participants who were overweight at baseline (fourth vs. first quartile RR = 1.02, 95% CI: 0.85, 1.21). Results from the French NutriNet-Santè cohort [21], including 110260 adults, reported an 11% increase in the risk of developing overweight or obesity among normal-weight participants (HR = 1.11, 95% CI: 1.08, 1.14; p < 0.001) and a 9% increase in the risk of developing obesity among overweight participants (HR = 1.09, 95% CI: 1.05, 1.13; p < 0.13), associated with a 10% increase in the % of energy from UPF. Data from the EPIC cohort [33], including a multi-national population of 348748 adults, also reported that normal-weight participants in the fifth quintile of UPF consumption had a 15% higher risk of developing overweight or obesity (RR = 1.15, 95% CI: 1.11, 1.19; P trend < 0.001) than participants in the first quintile of UPF consumption. Similarly, overweight participants in the highest quantile of UPF consumption had a 16% higher risk of developing obesity (RR = 1.16, 95% CI: 1.09, 1.23; P trend < 0.001) than overweight participants with low consumption of UPF. Data from the China Health and Nutrition Survey (CHNS) [32], including 12451 adults of both sexes, showed a higher risk of overweight and obesity (OR = 1.45, 95% CI: 1.21, 1.74) and abdominal obesity (OR = 1.50, 95% CI: 1.29, 1.74) in participants consuming ≥50 g/day of UPF than non-consumers. Additionally, Rauber et al. [34] found that participants in the fourth quartile of UPF consumption presented a 79% and 30% greater risk of developing obesity (HR = 1.79, 95% CI: 1.06, 3.03) and abdominal obesity (HR = 1.30, 95% CI: 1.14, 1.48), respectively, than participants in the first quartile of UPF consumption. Sandoval et al. [25] reported that, in the Seniors-ENRICA-1 cohort, the incidence of abdominal obesity in elders was significantly higher in participants in the uppermost tertile of UPF consumption than participants in the lowest one (OR = 1.61; 95% CI: 1.01, 2.56; P trend = 0.048). Finally, DaSilva Magalhães et al. [20] assessed UPF consumption in 896 men and women aged 23-25 years and related it to the incidence of metabolic syndrome and its components at ages 37-39. They found that UPF consumption was associated with a higher risk of abdominal obesity in women (RR = 1.01, 95% CI: 1.00, 1.02) but not in men.

Consumption of Ultra-Processed Food and Hypertension
Four studies focused on the relationship between UPF consumption and the incidence of hypertension [18,20,23,27]. Three studies [18,20,27] evaluated this relation both in men and women, whereas only one [23] did so for women. Additionally, two studies evaluated the outcome as self-reported diagnoses of hypertension [23,27] while in the other two [18,20], the outcome was defined by measuring the systolic and diastolic blood pressure during the follow-up. Mendonça et al. [27] observed a 21% higher risk of hypertension among participants in the uppermost tertile of UPF consumption compared with participants in the first tertile (HR = 1.21, 95% CI: 1.06, 1.37, P trend = 0.004). Similarly, Scaranni et al. [18] found participants of the ELSA-Brasil cohort with high UPF consumption to have a 17% increased risk of developing hypertension (OR = 1.17; 95% CI: 1.00, 1.37) than participants with low UPF consumption. In contrast, in the Mexican Teachers' Cohort (MTC), including 64934 women, Monge et al. [23] did not find UPF consumption significantly associated with the incident risk of hypertension when comparing extreme categories of UPF consumption (IRR = 0.96; 95% CI: 0.79, 1.16; P trend = 0.95). Finally, DaSilva et al. [20] reported that the %UPF at 23-25 years is marginally associated with the risk of hypertension at 37-39 years old (%kcal adjusted RR = 1.01; 95% CI: 1.00, 1.02).

Consumption of Ultra-Processed Food and Lipid Profile
Among the studies selected, three studies investigated the association between UPF consumption and the incidence of dyslipidemia [19,20,29]. Two of them focused on adults [19,20] and the other one on elders (≥60 years old) [29]. Of the 1821 participants from the Seniors-ENRICA cohort, Donat-Vargas et al. [29] reported that participants in the third tertile of energy intake of UPF had a higher risk for hypertriglyceridemia (OR = 2.66; 95% CI: 1.20, 5.90; P trend = 0.011) and low HDL cholesterol (OR = 2.23; 95% CI: 1.22, 4.05; P trend = 0.012) than participants in the first tertile. No association between UPF consumption and high LDL cholesterol emerged. Scaranni et al. [18,19], in the ELSA-Brasil cohort, observed that participants with medium and high UPF consumption had a higher risk of developing isolated hypertriglyceridemia (OR = 1.14, 95% CI: 1.03 and 1.26; OR = 1.30, 95% CI: 1.17 and 1.45), isolated hypercholesterolemia (OR = 1.12, 95% CI: 1.00 and 1.27; OR = 1.28, 95% CI: 1.12 and 1.47), low HDL cholesterol (OR = 1.12, 95% CI: 1.00 and 1.24; OR = 1.18, 95% CI: 1.05 and 1.32), and mixed hyperlipidemia (OR = 1.21, 95% CI: 1.05 and 1.39; OR = 1.38, 95% CI: 1.18 and 1.62) than participants consuming lower amounts of UPF. However, the association with low HDL cholesterol was lost when BMI was included in the model. On the contrary, DaSilva et al. [20] reported that UPF consumption at 23-25 years old was not associated with the risk of hypertriglyceridemia at the age of 37-39. On the other hand, UPF was associated with a higher risk of low HDL only in women (RR = 1.02, 95% CI: 1.01, 1.04).

Consumption of Ultra-Processed Food and Metabolic Syndrome
Only one study evaluated the relationship between UPF consumption and the incident risk of MetS [20]. The authors [20] reported that UPF consumption at 23-25 years was not associated with the risk of MetS at 37-39 years (RR = 1.00; 95% CI: 0.99, 1.01).

Discussion
In this systematic review, we summarized all available prospective studies focused on the association between UPF consumption and the incidence of obesity and cardiometabolic risk factors in adults. All studies included reported UPF consumption associated with the risk of developing overweight and obesity [17,21,26,[31][32][33]. Moreover, although more limited in number, the studies included in this review agreed on the association between UPF consumption and abdominal obesity [17,20,25]. Much more limited and heterogeneous were the prospective studies investigating the association between UPF consumption and cardiometabolic risk factors. However, most evidence supports the existence of a relationship with an increased risk of dyslipidemia, hypertension, and diabetes.
Traditionally, UPFs are energy-dense products with low nutritional quality. They contribute to increasing dietary intakes of saturated and trans fatty acids, sugars, refined carbohydrates, and sodium, and to reducing dietary intakes of fiber, micronutrients, and other protective bioactive compounds naturally present in foods [14]. In addition, it has been reported that these products are less satiating and characterized by a greater glycemic response than minimally processed foods [36]. Because of the higher energy density, low satiating effect, and large portion packing [37], consumption of these products may promote excess energy intake [38]. The minimal preparation skills required for UPF consumption can then alter eating patterns, leading to the rapid and unconscious consumption of food while engaged in routine alternative activities [39,40], altering neural and digestive functions that signal hunger and satiety, leading to overconsumption [41,42]. In addition, given their high fat and sugar contents, they can alter the reward neurocircuit mechanism, leading to increased food cravings and further exacerbating overconsumption [43,44]. To this it should be added that, under conditions of energy excess, the increased glucose response induced by UPF consumption may alter the insulin response, favoring the storage of excess nutrients in adipose tissue rather than their oxidation [45]. Excessive energy intake and obesity resulting from UPF consumption are certainly reasons for the development of cardiometabolic risk factors. However, this cannot entirely explain the associations observed between UPF and cardiometabolic risk factors, as many studies controlled their models for BMI and total energy intake. Many UPFs, such as condiments, broths, soup powders, and processed meats, have high levels of salt, contributing to higher sodium intake, a known risk factor for developing hypertension [46]. Added sugar could also alter fructose metabolism in the liver, promoting insulin resistance in the liver and throughout the body. Added fructose has been found to contribute to low-grade inflammation and oxidative stress, potentially causing β-cell damage and reducing insulin secretion [47]. Moreover, excess dietary fructose has been reported to impair the catabolism of very lowdensity lipoprotein cholesterol (VLDL-C) and increase VLDL-C synthesis, leading to an increase in triglycerides [48]. UPFs are also a source of trans and saturated fatty acids, which may contribute to an increased risk of dyslipidemia. Several RCTs found trans fatty acids having adverse effects on lipid profile [49], such as decreasing HDL cholesterol [50]. In addition, the intake of saturated fatty acids may have a negative impact on lipid metabolism, especially by virtue of the fact that UPFs are simultaneously low in PUFA [51]. Finally, UPFs contain plenty of chemical additives, synthetic antioxidants, and preservatives; many of these have been shown to increase the risk of obesity, deteriorate the lipid [52] and glucose profiles [53], and induce low-grade inflammation and metabolic syndrome [54]. In addition, the packaging of UPF can release known endocrine-disrupting chemicals (e.g., bisphenol A) into the food, increasing the risk of obesity and cardiometabolic risk [55][56][57]. Finally, it is presumed that those who consume high amounts of UPF have lower consumption of whole grains, fruits, and vegetables, limiting the intake [58] of micronutrients and bioactive compounds that may reduce cardiometabolic risk.
Despite the associations found, some considerations need to be made to evaluate the associations between UPF consumption and the incidence of obesity and cardiometabolic risk factors and to compare the results between studies. Only six studies controlled for dietary patterns or quality [19,21,22,24,25,33]. Considering the overall dietary pattern avoids potential confounding by other aspects of the diet, allows for evaluation of the interaction between synergistic components, and increases the ability to assess stronger effects due to the cumulative effects of many dietary characteristics [59]. An approach focused only on UPF consumption does not take into account the substitution effects of foods and associated foods [60]. Consumption of UPFs in a varied and balanced diet may not have the same effect as when they are consumed in a high-calorie diet, in which consumption of UPFs leads to the reduction of foods of higher nutritional value [9]. In addition, a very limited number of studies have repeatedly measured exposure. Dietary habits may change over time according to the food offered and living or environmental conditions, and consequently, they may influence the risk of obesity and cardiometabolic risk factors. An additional source of bias may be the method used for diet assessment. Most studies used 24 h recall and FFQ, while only two studies used dietary history. Although they are all accepted methods for evaluating dietary consumption, they are susceptible to recall bias and to difficulties in quantifying portions, compared with prospective methods based on recording and weighing foods consumed. Moreover, although a single 24 h recall is generally considered valid for assessing a population's food intake, to have a better estimation of habitual UPF consumption, especially given the wide range of products that are part of it, it may be necessary to consider multiple food days. In addition, although the FFQ is a commonly used method to assess the diet-health association, it suffers from some measurement errors. The dietary assessment is often limited to a specific list of foods that varies according to the questionnaire used and the quantification of intake is not as accurate as with the 24 h recall or food diary [61]. Moreover, it should be remembered that all of these methods are not specifically designed to assess UPF consumption as it is defined by the NOVA classification. This can determine an overestimation or underestimation of UPF consumption. Finally, the use of different units of measurement to assess exposure to UPF (e.g., %UPF energy , servings, g/day, (%UPF intake ) may have contributed to increased heterogeneity among studies. Future studies should therefore standardize the units of measurement to facilitate the comparison of results. Since obesity, as well as cardiometabolic risk and other NCDs, is strongly related to caloric intake, it is important to discern the effect of UPFs from that of total energy intake. Using a nutrient density model (%UPFenergy), without further adjustment for total calories, is not sufficient to remove the effect of total energy intake [62]. In addition, this approach does not allow for the consideration of UPFs that do not provide energy (e.g., artificially sweetened beverages). Similarly, the use of daily consumption frequency alone, without portion quantification, does not allow for true quantification of the foods consumed (e.g., many small portions might be equivalent to one large portion). These limitations can be overcome by using the total weight of foods consumed. In addition, using energy-adjusted food weight with the residual method would control for confounding factors by total energy intake and remove extraneous variations due to total energy intake [62].
Among the limitations of the present review is that many of the studies assessed exposure only at baseline. It must be considered that dietary habits assessed at baseline may have changed during follow-up, affecting risk estimates. To obtain a better representation of dietary habits and identify the direction of their relationship with cardiometabolic risk, more longitudinal studies with repeated assessments of food intake are needed. Moreover, several studies had a retention rate during follow-up that was potentially suboptimal (<80%). In addition, although the adjustment for confounders was considered satisfactory overall, several studies did not consider diet quality, which may have influenced the result. Poor geographic representativeness is a further limitation. The majority of the studies were conducted in Brazil, Spain, France, and England, limiting the generalizability of the results for other countries. Although the NOVA classification is internationally recognized, it may not be appropriate in all countries due to different cultural and dietary habits, as well as different industrial food production technology. For example, it was found that 23% of UPFs sold in Italy were of high nutritional quality considering three front-of-pack labeling schemes [63]. Therefore, further studies need to be conducted on other populations in order to develop correct nutrition policies and recommendations.
Nevertheless, this systematic review also has some strengths. We included only prospective cohort studies that, by measuring events in a temporal sequence, allow us to distinguish causes from effects [64]. This also made it possible to limit the variability due to the use of different study designs. In addition, we only included studies that used the NOVA classification, limiting the variability among studies due to different methods of defining UPF.

Conclusions
In conclusion, studies currently available in the literature agree that the consumption of ultra-processed foods is associated with the incidence of obesity. Less clear is its relationship with the incidence of outcomes related to cardiometabolic risk. Despite the positive associations found between the consumption of ultra-processed foods and cardiometabolic risk, the studies reported in the literature are still very limited, especially for some outcomes, and some results are conflicting, probably due to the adoption of different methods for assessing dietary habits, adjustment for possible confounders not always optimal, and other methodological limitations. Further longitudinal studies are therefore needed to better compare these associations, possibly considering overall dietary quality and dietary changes over time.