Obesity is a worldwide problem with increasing prevalence. In 2014, more than half a billion adults were obese, and 39% were overweight. In the Americas, ∼30% of women and 25% of men were obese, and in the Eastern Mediterranean ∼25% of women and 15% of men were obese (1).
Changes in the food system continuously promote obesity. There is now a greater availability of ready-to-eat or -heat foods known as ultraprocessed foods, which are products that have little, if any, whole foods and are manufactured with substances extracted from foods or synthesized in laboratories (dyes, flavorings, and other additives) (2). They have high amounts of fat, sugar, and salt and a high energy density and low fiber content; they are extremely palatable foods that are aggressively advertised (3) and contain a large diversity of chemical additives. Examples of ultraprocessed foods include breakfast cereals, reconstituted meat products, soft drinks, and ready-to-eat foodstuffs (2).
Several prospective studies have been carried out to assess the relation between dietary components and obesity. An analysis of 3 American cohorts, the Nurses’ Health Study, Nurses’ Health Study II, and Health Professionals Follow-Up Study, showed that the consumption of foodstuffs such as sweets, desserts, processed meats, fries, and sugar-sweetened beverages were strongly associated with weight gain in US adults (4). However, the relation between the consumption of foods aggregated according to their degree of processing (i.e., ultraprocessed foods) and excess weight has only recently been analyzed.
A cross-sectional times series study in 15 Latin American countries revealed an association between sales of ultraprocessed foods and changes in body weight in 12 countries from 2000 to 2009 (5). Cross-sectional studies have demonstrated an association between the consumption of these foods and the risk of metabolic syndrome in adolescents (6) and obesity in adolescents and adults (7, 8). A study of children aged 3–8 y in a low-income community in Brazil showed that ultraprocessed food consumption was an important predictor of the increase in total cholesterol and LDL cholesterol (9). On the other hand, a study with data from the 2008–2012 United Kingdom National Diet and Nutrition Survey showed no association between ultraprocessed food consumption and body weight. However, diets with smaller quantities of ultraprocessed foods have better nutritional quality (10).
Studies have shown an association between specific types of ultraprocessed foods, such as soft drinks, and being overweight or obese. However, no prospective cohort study to our knowledge has been performed to evaluate the association between all ultraprocessed foods as a group and the incidence of overweight and obesity in adults. Therefore, in this study (NCT02669602), we evaluated the association of ultraprocessed food consumption with the incidence of overweight and obesity in a Mediterranean cohort with a prolonged follow-up.
The SUN (University of Navarra Follow-Up) Project is a dynamic and multipurpose prospective cohort study with permanently open recruitment conducted in Spain among university graduates since December 1999. The participants are followed up biennially with the use of questionnaires distributed by post or electronic mail. Details of its design have been published elsewhere (11, 12).
Up to March 2012, the SUN data set included 21,291 participants who had answered the baseline questionnaire. In this study, we excluded those classified as overweight or obese [BMI (in kg/m2) ≥25] at baseline (n = 6340), individuals who reported total energy intake values outside of predefined limits [low: <3347 kJ/d or <800 kcal/d in men and <2092 kJ/d or <500 kcal/d in women; high: >16,736 kJ/d or >4000 kcal/d in men and >14,644 kJ/d or >3500 kcal/d in women (n = 1713)] (13), women who were pregnant at baseline or became pregnant during the follow-up period (n = 2739), and individuals who reported a previously diagnosed chronic disease at baseline (e.g., diabetes, cancer, cardiovascular disease) (n = 618). In addition, we excluded participants with a weight change >10 kg in the 5 y preceding entry into the study to reduce potential sources of confounding by other causes of weight changes (n = 260). Among the remaining participants, 1106 subjects were lost to follow-up, and 64 participants had missing values in ≥1 variable of interest. After these exclusions, a total of 8451 participants were included in the final analyses (Figure 1). The retention rate of the study was ∼89%.
Open in new tabDownload slide
Flowchart of participants.
Open in new tabDownload slide
Flowchart of participants.
This study was conducted according to Declaration of Helsinki guidelines, and all procedures involving human subjects were approved by the University of Navarra institutional review board. Voluntary completion of the baseline self-administrated questionnaire was considered to imply informed consent.
Exposure assessment: ultraprocessed food consumption
Dietary exposures were assessed at baseline through a self-administered 136-item semiquantitative food-frequency questionnaire (FFQ) that was previously validated in Spain (14, 15). Frequencies of consumption were measured in 9 categories (ranging from never or almost never to >6 servings/d), and the FFQ included a typical portion size for each item. Daily food consumption was estimated by multiplying the portion size by the consumption frequency for each food item.
The foods were classified according to NOVA (2) based on the extent and purpose of applied food processing. There are 4 groups in this classification scheme. The first group includes foods that are fresh or processed in ways that did not add substances such as salt, sugar, oils, or fats and infrequently contain additives. Processes used are aimed to extend life, allow storage for long use, and facilitate or diversify preparation (freezing, drying, and pasteurization). Examples are fruits and vegetables, grains (cereals), flours, nuts and seeds, fresh and pasteurized milk, natural yogurt with no added sugar or artificial sweeteners, meat and fish, tea, coffee, drinking water, spices, and herbs. The second group contains processed culinary ingredients. These are substances obtained from the first group or from nature and may contain additives to preserve the original properties. Examples are salt, sugar, honey, vegetable oils, butter, lard, and vinegar. The third group is processed food made with the addition of substances such as salt, sugar, or oil and the use of processes such as smoking, curing, or fermentation. Examples are canned or bottled vegetables and legumes, fruits in syrup, canned fish, cheeses, freshly made bread, and salted or sugared nuts and seeds. The fourth group is ultraprocessed food and drink products that are made predominantly or entirely from industrial substances and contain little or no whole foods. These products are ready to eat, drink, or heat. Examples include carbonated drinks, sausages, biscuits (cookies), candy (confectionery), fruit yogurts, instant packaged soups and noodles, sweet or savory packaged snacks, and sugared milk and fruit drinks (2). It is this fourth group that is the main subject of this study. Supplemental Table 1 describes the classification of FFQ foods according to NOVA. The frequency of ultraprocessed food consumption was estimated with the use of the sum of the food items from the fourth group in the FFQ (total of 33 items). The sample was divided into quartiles according to total consumption (servings/d).
Self-reported weight and height were validated with a previous study (16). The outcome used the incidence of overweight and obesity (BMI ≥25) during follow-up and was defined as the first time participants reached a BMI of 25 during follow-up.
Assessment of other variables
The baseline questionnaire also included questions relating to the following variables: sex, age, marital status, educational status, smoking status, physical activity, television watching, siesta sleep, diet and dietary habits, and snacking between main meals. Physical activity was evaluated with the use of a validated 17-item questionnaire (17). Total energy, macronutrient, fiber, and alcohol intake and fruit, vegetable, fast-food, fried-food, processed meat, nonprocessed meat, and sugar-sweetened beverage consumption was assessed with the use of the FFQ (15). Nutrient intake scores were computed with the use of a computer program developed specifically by our dietitians for this purpose. Adherence to the Mediterranean dietary pattern was evaluated with the use of a well-known score (18).
Differences in baseline characteristics of participants according to ultraprocessed food consumption quartiles were evaluated with ANOVA and adjusted for sex and age. To evaluate the relation between ultraprocessed food consumption at baseline and the subsequent risk of the development of overweight and obesity during follow-up, we used Cox proportional hazards models, and to estimate HRs and 95% CIs we used the lowest quartile as the reference category.
The follow-up period was defined as the interval between the date of recruitment and date of the return of the follow-up questionnaire in which the participant was classified as overweight or obese for the first time (for incident cases). The date of death or of the last questionnaire was used for noncases.
Tests for linear trends were conducted by assigning medians of ultraprocessed food consumption to each category and treating this variable as a continuous variable in the respective Cox regression model. We fitted a first model without any adjustment (crude), a second model adjusted for age and sex, and a third multivariable-adjusted model adjusted for age, sex, marital status, educational status, baseline BMI, physical activity, television watching, siesta sleep, smoking status, snacking between meals, and following a special diet. Total energy intake was not included as a covariate because it may plausibly mediate the association of ultraprocessed foods and overweight and obesity. We evaluated the interaction between exposure, sex, and BMI with the use of a likelihood ratio test that compared the fully adjusted Cox regression model and the same model with interaction product terms (3 df). Nelson-Aalen cumulative hazards estimates were plotted for overweight and obesity incidence according to ultraprocessed food consumption quartiles at baseline. We used inverse probability weighting to adjust the Nelson-Aalen curves for baseline potential confounders.
To test the proportional hazard assumption, we calculated a Cox regression with the exposure as a continuous time-varying covariate to check that the HR did not vary over time, obtaining a nonsignificant result, suggesting that the proportionality assumption was met. We also checked the proportionality of hazards model with the use of a Grambsch-Therneau test of the scaled Schoenfeld residuals from a Cox model on the 3 dummy variables of the upper ultraprocessed quartiles (19). The P value of the global test was 0.72.
To determine the contribution of each food item to the between-person variance in ultraprocessed food consumption (13), we constructed a series of nested least-squares linear regression models after stepwise-selection regression analyses. The additional contribution of a given food item was reflected in the change in the cumulative R2.
Sensitivity analyses were conducted by repeating the multivariable-adjusted Cox regression models with the following changes: 1) additional adjustment excluding fruit and vegetable consumption, 2) exclusion of those participants under the 5th percentile and over the 95th percentile of total energy intake, 3) additional adjustment for total energy intake, 4) additional adjustment for family history of obesity, 5) additional adjustment for weight gain >3 kg in the 5 y before entering the cohort, 6) exclusion of participants who were early incident cases of overweight (those who became overweight after only 2 y of follow-up), and 7) inclusion of subjects with prevalent chronic diseases.
All analyses were performed with Stata version 12.1 (StataCorp LP). P < 0.05 was considered significant.
A total of 2967 (35.1%) men and 5484 (64.9%) women were included in this analysis, and the mean ± SD age of the participants was 37.6 ± 11.0 y. The main baseline characteristics of participants according to quartiles of total ultraprocessed food consumption are presented in Table 1. Participants in the fourth quartile of ultraprocessed food consumption had the highest BMI, were more likely to be current smokers, watched more television, and had the highest total energy and fat intake and the lowest protein and total fiber than those in the first quartile. Moreover, on average, they consumed more fast food, fried foods, processed and other meats, and sugar-sweetened beverages. In contrast, they had the lowest intakes of vegetables and were less likely to follow special diets. In addition, they had the highest prevalence of snacking between main meals. In addition, the adherence to the Mediterranean diet score varied according to quartiles of ultraprocessed food (i.e., the higher the consumption of ultraprocessed foods, the less adherence to the Mediterranean diet).