Chronic Non-Communicable Diseases (NCD), responsible for 71% of deaths worldwide, have modifiable risk factors such as physical inactivity, smoking, unhealthy diet, alcohol abuse, in addition to socioeconomic factors [1]. In Brazil, in order to fight NCD, the goal is to change the diet of individuals, aiming at a healthy diet [2].

In this perspective, the Dietary Guidelines for the Brazilian Population (DGBP) were developed, which at the forefront of food and nutritional recommendations, adopted the degree of processing as a criterion for choosing foods, holding as a golden rule: "Always prefer fresh foods or minimally processed foods and culinary preparations over ultra-processed foods (UPF)” [3]. The NOVA food classification was recognized as important in scientific research and its investigation was encouraged in food consumption assessments [4]. As a result, the number of studies that assess food consumption according to degree of processing has increased, especially focused on associations between UPF consumption and health outcomes [5-7]. Some others include the assessment of food consumption also for other levels of processing [8-10].

In addition, it is known that just as there is an interaction between health and society, in which factors in different layers have an influence on the health of the individual [11,12], so too food consumption is influenced by different determinants, such as clinical, psychological, social, economic, demographic, cultural, and contextual factors [13,14].

In this context, knowing the food consumption of a population and its associated factors provides a means to apply effective interventions, contributing to reducing modifiable risk factors to improve the health of individuals and reduce subsequent expenses with NCD.

Therefore, the objective of this study was to evaluate the food consumption of Brazilian graduates, participants of the Cohort of Universities of Minas Gerais (CUME Project), according to the degree of processing, and its association with socioeconomic, behavioral, and individual factors.

Considering food consumption, total energy (non-parametric variable) is best described by the median, which presented a value of 2,224 kcal and an Interquartile Range (IQ) of 1,108 kcal.

In a percentage of 100% of the total energy contribution of the 3 groups, each had the following contribution: Group 1, 65.5% (95% CI: 65.2-65.9%); Group 2, 10.0% (95% CI: 9.8-10.2%); Group 3, 24.5% (95% CI: 24.1-24.8%). In order to describe the food consumption of the sample in more detail, the energy contribution was grouped into subgroups (Table 1).

Table 1 -

Food groups and consumables	Energy contribution (% DEI)
	Mean	95% CI
Group 1: In natura food §	65.5	(65.2-65.9)
Rice	4.5	(4.4-4.6)
Noodles	0.9	(0.9-1.0)
Legumes	3.9	(3.8-4.0)
Roots and Tubers	1.7	(1.6-1.7)
Vegetables	2.1	(2.0-2.1)
Fruits	12.8	(12.5-13.1)
Natural fruit juice	2.3	(2.3-2.4)
Coffee, teas, and chimarrão	0.7	(0.7-0.8)
Milk	3.3	(3.2-3.4)
Beef and pork	9.3	(9.1-9.6)
Chicken meats	4.2	(4.1-4.3)
Fish meats	1.7	(1.6-1.8)
Eggs	1.4	(1.4-1.5)
Culinary preparations a	2.0	(1.9-2.1)
Fried preparations b	1.8	(1.7-1.9)
Added sugar	1.7	(1.6-1.8)
Oils and fats c	4.6	(4.5-4.7)
Other in natura foods † d	6.3	(6.1-6.5)
Group 2: Processed foods	10.0	(9.8-10.2)
French bread	3.3	(3.2-3.4)
Cheeses	3.4	(3.3-3.5)
Processes meats e	0.7	(0.7-0.8)
Fruit jams and jellies	0.5	(0.5-0.6)
Processed alcoholic beverages f	2.0	(1.9-2.1)
Group 3: Ultra-processed foods	24.5	(24.1-24.8)
White bread	1.1	(1.1-1.2)
Others types of bread g	5.7	(5.6-5.9)
Margarine, mayonnaise, and cream cheese	2.3	(2.2-2.4)
Sausages h	2.2	(2.1-2.3)
Fast foods i	2.3	(2.2-2.4)
Ready or semi-ready dishes j	1.9	(1.9-2.0)
Sweetened dairy drinks	1.2	(1.2-1.3)
Soft drinks and industrialized juices	1.5	(1.5-1.6)
Ultra-processed alcoholic beverages k	0.3	(0.3-0.4)
Sweets l	3.4	(3.3-3.5)
Other ultra-processed foods m	2.3	(2.2-2.4)

Note: ^§ In natura, minimally processed foods, culinary preparations and culinary ingredients; † Other in natura, minimally processed foods and culinary ingredients; ^a Meatballs, hominy, gnocchi, polenta/cornmeal, noodle or rice soups; ^b Polenta, cassava, and French fries; ^c Olive oil, canola, soybean, sunflower, and corn oils, butter, pork fat; ^d Cassava flour, corn flour, acai pulp, raisins, peanuts, nuts, chestnuts, shrimp, offal, textured soy protein, sushi, oats, granola, peppers, and honey; ^e Smoked meats, dried meat, sardines and canned tuna, cod; ^f Beers and wines; ^g Toast, whole grain bread, light bread, sweet bread, cheese bread; ^h Mortadella, turkey breast, sausage links, sausage, bacon; ⁱ Hot dogs, hamburgers, fried and baked snacks; ^j Lasagna, pizza; ^k Cachaça, distilled beverages; ^l Chocolates, bonbons, candies, ice creams, puddings, dulce de leche, rice pudding, other non-fruit sweets; ^m Breakfast cereals, cereal bars, chocolate milk, mustard, soluble soy extract, popcorn, snacks, and chips. CI: Confidence Interval; DEI: Daily Energy Intake.

The population of the present study had a minimum age (non-parametric variable) of 20 years and a maximum of 86 years, with a median of 34 years and an IQ of 12 years. Most were female (68.1%), not legally married or without a stable union (51.9%), and white skin color (65.1%). The minimum education level was undergraduate, which is characteristic of this population, and 72.5% of the individuals had graduate-level degrees. Regarding income, 52.4% of the participants had an individual income equal to or greater than 5 x MW, and 78.6% had a family income equal to or greater than 5 x MW. Considering their behavioral characteristics, 55.8% of the individuals were active, 41.6% presented alcohol abuse, and 8.6% reported currently smoking. Regarding self-perception of health, most of the sample (88.5%) self-reported their health status as very good or good. Regarding the presence of NCD, 11.7% were obese, 12.1% had SAH, 3.1% had DM, and 12.4% reported a medical diagnosis of depression (Table 2).

Table 2 -

Variable	n	%
Block 1 - Socioeconomic Factors
Marital Status
Married legally or in a stable union	1,982	48.1
Not Married/No Stable Union	2,142	51.9
Education
Graduate	1,136	27.5
Graduate degrees	2,988	72.5
Professional situation
Works	3,127	75.8
Does not work	997	24.2
Individual Income
<5 x MW	1,965	47.6
≥5 x MW	2,159	52.4
Family Income
<5 x MW	884	21.4
≥5 x MW	3,240	78.6
Block 2 - Behavioral Factors
Physical Activity
Inactive or Insufficiently	1,823	44.2
Active	2,301	55.8
Binge Drinking
No	2,409	58.4
Yes	1,715	41.6
Tobacco use
No	3,769	91.4
Yes	355	8.6
Block 3 - Individual Factors
Sex
Male	1,316	31.9
Female	2,808	68.1
Skin Color
White	2,683	65.1
Non- White	1,441	34.9
Self-reported Health Status
Very Good/Good	3,645	88.5
Fair/Poor/Very Bad	473	11.5
Presence of Obesity
No	3,640	88.3
Yes	482	11.7
Presence of Systemic Arterial Hypertension
No	3,626	87.9
Yes	498	12.1
Presence of diabetes Mellitus
No	3,996	96.9
Yes	128	3.1
Presence of Depression
No	3,611	87.6
Yes	513	12.4

Note: MW: Minimum Wage R$ 880,00 in 2016; R$ 954,00 in 2018. IQ: Interquartile Range.

Table 3 presents the final hierarchical model, with the consumption outcome for Group 1. In Block 1, marital status was the only socioeconomic factor that remained in the final model, in which the association was negative in relation to individuals who were unmarried or without a stable union (B:-0.76; p=0.049). From Block 2, the physical activity variable (B:2.56; p<0.001) was positively associated, while alcohol abuse (B:-2.50; p<0.001) and tobacco use (smoker) (B:-1.64; p=0,014) were negatively associated. Regarding Block 3, female gender (B:1.51; p=0.001), “non-white” skin color (B:1.71; p<0.001), and age (B:0.19; p<0.001) were positively associated. Regarding the presence of NCD, a positive association was observed with DM (B:2.12; p=0.049) and a negative one with obesity (B:-2.65; p<0.001) and depression (B:-1.91; p=0.001).

Table 3 -

Variable	Group 1 - In natura foods §
	B	(CI 95%)	p-value
Block 1 - Socioeconomic Factors
Marital Status
(Reference category: Married Legally or in a Stable Union)
Not Married/No Stable Union	-0.76	(-1.52; -0.03)	0.049*
Block 2 - Behavioral Factors
Physical Activity
(Reference category: Inactive or Insufficiently active)
Active	2.56	(1.82; 3.30)	<0.001*
Binge Drinking
(Reference category: No)
Yes	-2.50	(-3.27; -1.75)	<0.001*
Tobacco use
(Reference category: No)
Yes	-1.64	(-2.95; -0.33)	0.014*
Block 3 - Individual Factors
Sex
(Reference category: male)
Female	1.51	(0.72; 2.31)	0.001*
Age
(Continuous variable)	0.19	(0.14; 0.23)	<0.001*
Skin Color
(Reference category: white)
Non-White	1.71	(0.96; 2.46)	<0.001*
Self-reported Health Status
(Reference category: Very Good/Good)
Fair/Poor/Very Bad	-1.003	(-2.19; -0.18)	0.097
Presence of Obesity
(Reference category: No)
Yes	-2.65	(-3.82; -1.47)	<0.001*
Presence of diabetes Mellitus
(Reference category: No)
Yes	2.12	(0.01; 4.23)	0.049*
Presence of Depression
(Reference category: No)
Yes	-1.91	(-3.02; -0.80)	0.001*

Note: *Variables with p<0.05. § In natura, minimally processed foods, culinary preparations and culinary ingredients. Multiple Hierarchical Linear Regression. ANOVA (p<0.001) / R Square Change= 0.061/ Durbin-Watson=1.97. All graphs were adequate for normality and homoscedasticity. CI: Confidence Interval.

Table 4 presents the final hierarchical model, with the outcome of food consumption for Group 2. In Block 1, no variable remained in the final model. In Block 2, the physical activity variable (B:-0.56; p=0.002) was negatively associated with the consumption of food in Group 2, while alcohol abuse (B:2.41; p<0.001) and tobacco use (smoker) (B:1.39; p<0.001) were positively associated. In relation to Block 3, female gender (B:-1.18; p<0.001), “non-white” skin color (B:-0,53; p=0.004) and age (B:0.06; p<0.001) were associated with the increase in energy from the consumption of food in Group 2.

Table 4 -

Variable	Group 2 - Processed foods
	B	(CI 95%)	p-value
Block 1 - Socioeconomic Factors
No variables from this block remained in the final model	-	-	-
Block 2 - Behavioral Factors
Physical Activity
(Reference category: Inactive or Insufficiently active)
Active	-0.56	(-0.92; -0.21)	0.002*
Binge Drinking
(Reference category: No)
Yes	2.41	(2.04; 2.78)	<0.001*
Tobacco use
(Reference category: No)
Yes	1.39	(0.75; 2.02)	<0.001*
Block 3 - Individual Factors
Sex
(Reference category: male)
Female	-1.18	(-1.56; -0.81)	<0.001*
Age
(Continuous variable)	0.06	(0.04; 0.08)	<0.001*
Skin Color
(Reference category: white)
Non-White	-0.53	(-0.89; -0.16)	0.004*

Note: *Variables with p<0.05. Multiple Hierarchical Linear Regression. ANOVA (p<0.001) / R Square Change= 0.074/ Durbin-Watson=2.01. All graphs were adequate for normality and homoscedasticity. CI: Confidence Interval.

Table 5 presents the final hierarchical model, with the consumption outcome for Group 3. In Block 1, marital status was the only socioeconomic factor that remained in the final model, in which the association was positive for individuals who were unmarried or without a stable union (B:1.004; p=0.003). In Block 2, being active (a) was negatively associated (B:-1.98; p<0.001). In Block 3, “non-white” skin color (B:-1.20; p=0.001) and age (B:-0.24; p<0.001) were negatively associated with the increase in energy from the consumption of food in Group 3. Regarding the presence of NCD, there was a positive association with obesity (B:2.58; p<0.001) and depression (B:1.88; p<0.001) and a negative association with DM (B:-2.03; p=0.035).

Table 5 -

Variable	Group 3 - Ultra-processed foods
	B	(CI 95%)	p-value
Block 1 - Socioeconomic Factors
Marital Status
(Reference category: Married Legally or in a Stable Union)
Not Married/No Stable Union	1.004	(0.33; 1.68)	0.003*
Block 2 - Behavioral Factors
Physical Activity
(Reference category: Inactive or Insufficiently active)
Active	-1.98	(-2.63; -1.33)	<0.001*
Block 3 - Individual Factors
Age
(Continuous variable)	-0.24	(-0.27; -0.21)	<0.001*
Skin Color
(Reference category: white)
Non-White	-1.20	(-1.87; -0.53)	0.001*
Self-reported Health Status
(Reference category: Very Good/Good)
Fair/Poor/Very Bad	1.002	(-0.54; 2.06)	0.063
Presence of Obesity
(Reference category: No)
Yes	2.58	(1.54; 3.63)	<0.001*
Presence of diabetes Mellitus
(Reference category: No)
Yes	-2.03	(-3.92; -0.15)	0.035*
Presence of Depression
(Reference category: No)
Yes	1.88	(0.89; 2.86)	<0.001*

Note: *Variables with p<0.05. Multiple Hierarchical Linear Regression. ANOVA (p<0.001) / R Square Change= 0.069/ Durbin-Watson=1.96. All graphs were adequate for normality and homoscedasticity. CI: Confidence Interval.

The assessment of food consumption in this study was not limited to factors associated with UPF consumption, but also included the analysis of factors associated with the consumption of processed foods and in natura, minimally processed foods, culinary ingredients and culinary preparations, based on the DGBP [3]. It also considered that this consumption has interaction with different factors, thus applying the hierarchical analysis of associated, socioeconomic, behavioral, and individual factors, and found important associations as shown in the results.

Food consumption, divided into degrees of food processing, presented frequencies similar to other studies with the Brazilian population [8,29,30]. UPF consumption receives attention, as it is directly associated with the presence of NCD [5,6,31] and its consumption is growing at a fast pace globally, especially in high-income countries, where these foods are the main source of daily energy [32]. In the US, the average daily energy contribution found was more than half (58.5%) from UPF [33]. In samples from Canada [34] and the United Kingdom [35], the average daily UPF consumption found was 46.8% and 53%, respectively.

Soft drinks and industrialized juices appear as a marker of an unhealthy diet and may be associated with greater abdominal adiposity and obesogenic eating behaviors [36,37]. The 1.5% daily energy frequency of these drinks in the present study is similar to other Brazilian populations [8,30], and different from the percentage of 4.6% found in a US study [33]. Fast foods also constitute the UPF group and appear in the present study with a mean daily energy contribution of 2.3%. Such foods are frequent in food eaten away from home and may be related to obesity [38]. In a study in Latin America, only Argentina and Venezuela (due to financial crisis) did not observe an increase in the consumption of beverages and UPF sold in retail fast food [39].

Starting with the distal block of socioeconomic factors, both the relative percentage of energy from foods in Group 1 and foods in Group 3 were associated with marital status, while Group 2 was not associated with any factor in this block. This study showed that those married or in a stable relationship tended to consume more foods from Group 1, and less UPF. Canuto et al. [40] reviewed Brazilian surveys, in which only one article found an association with marital status, finding greater consumption of fruits and vegetables, which are in the group of in natura foods. In the USA, a study showed a significant association between UPF consumption and marital status, in which in the last quintile, the total for unmarried individuals tended toward higher UPF consumption [41]. Suggesting that married individuals tend to have a healthier diet than those without a spouse [40].

Considering behavioral factors, physical activity was directly associated with higher food consumption in Group 1, inversely with Group 2, and inversely with Group 3, similar to a study with Brazilians that showed that better food consumption was associated with leisure-time physical activity [42]. Alcohol abuse and tobacco use were associated only with Groups 1 and 2, being inversely with Group 1 food consumption and directly with Group 2, similar to a study conducted with young adults, in which the diet quality index was worse for those who smoked at least once a week, and who were in the habit of consuming alcoholic beverages [43]. In this case it is important to emphasize that 2% of the daily energy intake of Group 2 foods comes from processed alcoholic beverages, which would obviously have a direct association between Group 2 foods and higher alcohol consumption. Another relationship that can be suggested is that tobacco use is associated with alcohol abuse. Such results may be linked to risk behaviors, among which are the alcohol abuse, cigarette use, low consumption of fruits and vegetables, physical inactivity, and non-use of sunscreen, in which individuals present an average of three health risk behaviors, and these behaviors are interrelated [44].

Regarding individual characteristics, women tended to have a higher percentage of energy from foods in Group 1, and a lower percentage of energy from foods in Group 2, confirming what a review of Brazilian studies shows, that gender is a determinant of food consumption, demonstrating different intake of food groups and micronutrients between men and women [40]. Further demonstrating that the women in the present study have a protective diet, since UPF consumption is associated with excess weight and abdominal obesity, more pronounced in women [41].

The present study found that individuals with “non-white” skin color tended to have better food choices, being directly associated with the consumption of foods from Group 1 and inversely with the consumption of Groups 2 and 3, contrasting with another review, which showed white individuals showing greater consumption of fruits and vegetables [40]. Skin color tends to be associated with the presence of moderate to severe food insecurity [45], which in this study could be related to higher UPF consumption, however other characteristics, such as low education and more precarious socioeconomic conditions are not present to the population of the CUME project. Another important association was age, which was directly associated with the consumption of foods in Groups 1 and 2, and inversely with the consumption of UPF, corroborating other studies [8,46]. Such results may be related mainly to the emergence of UPF, which appear as substitutes for foods from the in natura food group, in 1987-8 [20], the period in which the older individuals in the present study probably already had their eating habits established [8].

Considering NCD, this study did not find a significant association between hypertension and food consumption in the three groups, although in a longitudinal analysis of the CUME Project, the association between UPF consumption and SAH can be observed [47].

However, there was an inverse association of obesity and depression with the consumption of foods in Group 1 and a direct association with the consumption of foods in Group 3. Similar results were found in relation to the higher consumption of UPF linked to higher BMI values ​​and the presence of obesity [48,49]. A study in the USA found that higher UPF consumption was associated with higher BMI [41]. In Australia, research has suggested that UPF negatively impact the intake of all nutrients related to NCD, especially in relation to excess free sugars, total, saturated, and trans fats, and fiber deficiency in these foods [50]. Longitudinal studies in Europe showed an association between UPF consumption and risk for overweight, obesity, and arterial hypertension [51,52].

Regarding depression, the findings were similar in France and Spain, in which the higher consumption of UPF was associated with the presence of depression, suggesting that more studies investigate this association, along with other factors related to diet and mental illness [53,54]. Unhealthy lifestyle habits can be related to both the presence of depression and poor diet quality.

The presence of diabetes in this population was directly associated with the increase in the percentage of energy from Group 1 and inversely with that from Group 3. This finding may be related to the impact of diabetes on the individual's health and the fear of developing comorbidities, causing people with diabetes to choose more appropriate and healthier options [55].

This study presented important results that corroborate the current literature, and some positive points should be noted: 1) the assessment of food consumption was not limited to factors associated with UPF consumption 2) the use of hierarchical analysis made it possible to contemplate the exposure factors at different levels 3) and although the data was collected using a virtual environment, the self-reported data was validated.

However, it should be noted that our study was limited to the baseline of the Cohort of Minas Gerais Universities (CUME Project), presenting a cross-sectional design, which does not allow establishing a causal relationship between exposure and outcome.

Energy contribution of Groups 1, 2, and 3 was found to be 65.5%, 10.0%, and 24.5%, respectively. Regarding the factors associated with food consumption, it is noteworthy that Group 1 was positively associated with the practice of physical activity, female gender, age, “non-white” skin color, and the presence of DM; and negatively with unmarried/non-stable marital status, alcohol abuse, tobacco use, presence of obesity, and presence of depression. Considering Group 2, it was positively associated with alcohol abuse, tobacco use, and age; and negatively with physical activity, female gender, and “non-white” skin color. As for Group 3, it was positively associated with a marital status of not married/without a stable union, presence of obesity, and presence of depression; and negatively with physical activity, age, “non-white" skin color, and presence of DM.