Introduction

Malnutrition in all its forms, including obesity, is a major cause of death and disease globally, as documented in the latest Global Burden of Disease Study (GBD) 2016 [1]. Available food energy per capita has increased in most regions of the world, and this increase was previously found to be sufficient to explain concurrent increases in average population body weight in many countries [2]. Globally, between 1990 and 2010, the consumption of healthy food items improved, while at the same time consumption of unhealthy food items worsened, with heterogeneity across regions and countries [3].

In Belgium, dietary risks are the top third contributor to the burden of disease (2016), following tobacco and high blood pressure [4]. Belgians are generally not meeting food-based dietary guidelines [5, 6], in particular for fruits and vegetables and for restricting energy-dense, nutrient-poor foods. Fruit and vegetable intakes among the Belgian population aged 15–64 years slightly decreased between 2004 and 2014, but the difference was not significant [5, 6]. Consumption of energy-dense, nutrient-poor foods among individuals 15–64 year old was excessive in 2004 (730 kcal/day) and decreased to 674 kcal/day in 2014. This decrease was also not statistically significant [5, 6]. In 2014, only about 2.1% of children (3–9 years), 2.4% of adolescents (10–17 years) and 6.6% of adults complied with the recommendations limiting the consumption of energy-dense nutrient-poor foods [6].

Recently, it has been suggested that food processing, more specifically the type, intensity and purpose of food processing may be linked to human health [7], and a new method of food classification has been proposed [8]: the NOVA classification (with NOVA being a name, not an acronym), which is increasingly applied by researchers to investigate the link between nutrition and human health. The NOVA classification divides foods into four groups according to their degree of processing:

  1. 1.

    unprocessed or minimally processed foods,

  2. 2.

    processed culinary ingredients,

  3. 3.

    processed foods and

  4. 4.

    ultra-processed foods.

Ultra-processed foods (UPF) are products made mostly or entirely from substances extracted from foods or derived from food constituents with little if any intact food, which often contain flavours, colours and other additives that mimic or intensify the sensory qualities of foods or culinary preparations made from foods [9].

UPF are designed to be convenient, attractive and accessible and they are highly profitable and heavily marketed [7, 9]. Minimally processed foods are more satiating and less hyperglycemic than UPF [10]. In addition, it has been shown that the environment created in the gut by UPF could be an evolutionarily unique selection ground for microbes with behaviours that promote diverse forms of inflammation-related diseases [11].

UPF represent already more than 50% of total daily energy intake in some high-income countries, such as the United States [12, 13] and Canada [14]. Consumption of UPF has been associated with unhealthy dietary patterns [15,16,17,18,19,20,21,22] and with overweight and obesity in studies conducted in the US [23], Canada [24], France [25], Brazil [26, 27] and both across Latin American [28] and European [29] countries. A recent cohort study from Spain found that participants in the highest quartile of UPF consumption were at a higher risk of developing overweight or obesity than those in the lowest quartile of consumption [30]. Other recent cohort studies from Spain and France found a link between consumption of UPF and hypertension [31] and between consumption of UPF and cancer [32], respectively.

However, to date, the dietary share of UPF, meaning the proportion of daily energy consumed from UPF, has not been determined for the Belgian population. The aim of this study was therefore to assess the dietary share of UPF in a representative sample of children, adolescents and adults in Belgium and its association with dietary quality using the data from the Belgian national Food Consumption Surveys 2004 and 2014–2015.

Methods

National food consumption survey data

The Belgian food consumption surveys 2004 and 2014/2015 were conducted according to the guidelines published by The European Food Safety Authority in view of the EU Menu project [33]. The surveys were approved by the Ethical Committee of the University of Ghent and the Commission for the Protection of Privacy. Participants signed a written informed consent form. The study design and methodology of both surveys have been explained in detail elsewhere [5, 34].

In brief, a representative sample of the Belgian population (N = 3083, individuals ≥ 15 years in 2004 and N = 3146, individuals 3–64 years in 2014–2015) was randomly selected from the National Population Register according to a multistage stratified sampling procedure.

Dietary intake in adolescents and adults [10–64 years old (or 15 years and older in 2004)] was assessed using the 24-h dietary recall method, carried out on two non-consecutive days. GloboDiet © (formerly EPIC-SOFT), a computerised 24-h recall program designed for the standardized collection of dietary data within a pan-European survey, was used and adapted to the Belgian situation [35]. GloboDiet involves a structured and standardized approach to collect very detailed descriptions and quantities of consumed foods, recipes and dietary supplements. Food portion sizes were quantified using household measures (e.g., glasses, cups, spoons, etc.), food portions (obtained from manufacturer’s information) and a picture book including a selection of country-specific dishes in different portion sizes.

Dietary assessment in children (3–9 years old) was done using two self-administered non-consecutive 1-day food diaries (open-ended without pre-coded food lists) followed by a GloboDiet completion interview with the proxy respondent (parent or legal guardian).

The collected food consumption data was afterwards linked with detailed information on the nutrient composition of each specific food item, using the Belgian Food Composition Data NUBEL (including branded foods) and the Dutch Food Composition data (NEVO).

The plausibility of reported energy intake was estimated by comparing the reported energy intakes with the presumed energy requirements using the Goldberg cutoff method, which has been revised by Black [36]. Misreporters included individuals with a ratio below the lower cutoff (under-reporters) and individuals with a ratio above the upper cutoff (over-reporters) [36].

Classification of foods according to the NOVA classification

All foods and ingredients consumed were classified as ‘ultra-processed’, ‘processed’, ‘unprocessed or minimally processed’ or ‘processed culinary ingredient’ according to the NOVA classification [8, 37], which is to date the most commonly used system to classify foods by level of processing. NOVA is now recognised as a valid tool for nutrition and public health research, policy and action, in reports from the Food and Agriculture Organization [38] of the United Nations and the Pan American Health Organization [39].

UPF are products made mostly or entirely from substances extracted from foods or derived from food constituents with little if any intact food, which often contain flavours, colours and other additives that mimic or intensify the sensory qualities of foods or culinary preparations made from foods [9]. Ingredients of UPF include versions of oils and fats, flours and starches, sugar, and proteins, including those resulting from further processing, such as hydrogenated oils and fats, modified starches, hydrolysed proteins, and crushed or extruded ‘mixes’ of meat offals or remnants [8].

Some food descriptors, used within the GloboDiet recall program, were particularly useful to classify foods consumed according to the NOVA classification [8, 37] (Supplementary material 1), such as sugar content (sweetened with sugar, sweetened with artificial sweeteners or unsweetened), conservation method (canned, frozen, dried, salted, marinated, candied, fresh/untreated, etc.), medium (in oil, in water, in own juice, in syrup, etc.), production method (home prepared, industrially prepared, artisanal, catering, etc.), and ingredient (salted or unsalted).

Home-prepared dishes and recipes were disaggregated into ingredients and those ingredients were classified according to the NOVA classification. However, some composite foods that were home prepared could not be disaggregated into core ingredients (e.g., some milk-based desserts, cakes, pies and pastries, some soups and sauces) and represented about 0.9% of all foods consumed in 2014 and 2.6% of all food items consumed in 2004. In such cases, these home-prepared composite foods were classified as processed. Alcoholic beverages were not classified using the NOVA classification and were kept as a separate group.

Other variables

Height was accurately measured to 0.5 cm using a stadiometer (type SECA 213) and weight to 0.1 kg using an electronic scale (type SECA 815 and 804). Height and weight were measured during the second home visit (same day as the second 24-h dietary recall). Data on sex, region, educational level (higher education long type, higher education short type, secondary education or lower), self-perceived health (very good, good, fair, bad, very bad), frequency of breakfast consumption (never, less than 1 day a month, 1–3 days a month, 1 day a month, 2–4 days a week, 5–6 days a week, every day), smoking (every day, once in a while, no) and frequency of consumption of meal with family (two or more meals a day, one meal a day, only in the weekend, only on days of celebration, never) were retrieved from a CAPI (computer-assisted personal interview) conducted during the first home visit (same day as first 24-h dietary recall). In children (3–9 years), a parent or legal guardian was used as a proxy respondent. Smoking and frequency of consumption of meal with family were only collected for adults and adolescents.

Data analysis

Analyses were conducted in SAS 9.3. All analyses took the survey design into account. The comparison of the proportions of foods/ingredients consumed classified into the different NOVA groups between 2004 and 2014/2015 was done using χ2 test.

The repeated 24-h recall data were used to estimate the usual intake distribution of the proportion of daily energy consumed from UPF and unprocessed/minimally processed foods, overall and by population group, in both 2004 and 2014/15. A correction for within-subject variation [40, 41] was applied using the Statistical Program to Assess Dietary Exposure (SPADE) [42]. The usual intake distribution was modelled as a function of age. Uncertainty in the habitual intake distribution was quantified with ready for use bootstrap which provided confidence intervals with the required confidence level [42]. The association between the usual proportion of daily energy consumed from UPF and the usual proportion of daily energy consumed from unprocessed/minimally processed foods was assessed trough Spearman rank correlation.

The mean energy contribution of food subgroups to the dietary share of UPF and unprocessed/minimally processed foods for different age groups and by sex was analysed using data from the first 24-h recall of the 2014/2015 survey.

The hypothesis that people meeting dietary recommendations consume lower proportions of daily energy intake from UPF and higher proportions of daily energy intake from unprocessed/minimally processed foods was tested. To do so, using the first interview day, the proportion of daily energy intake consumed from UPF and unprocessed/minimally processed foods was compared between consumption days on which the World Health Organization (WHO) recommendations for fruit and vegetables (≥ 400 g/day), salt (≤ 5 g/day), saturated fat (≤ 10% of daily energy intake), and trans-fatty acids (≤ 1% of daily energy intake) were met or not met (using Wilcoxon rank sum tests). Salt added at the table or during cooking was not collected in the surveys. Consumption data on added or free sugars were not available at the time of the study and therefore the WHO recommendation for intake of free sugars was not considered in this study.

Generalized linear models were used to investigate the association of proportion of daily energy intake from UPF and unprocessed/minimally processed foods with sex, age, educational level, body mass index class, region of residence, self-perceived health, frequency of breakfast consumption, smoking, and frequency of consumption of meal with family. The mean consumption over 2 days was used as the dependent variable. The models were adjusted for mean total daily energy intake and included an interaction term sex × BMI class. The analysis was repeated excluding misreporters. A p value of < 0.05 was considered statistically significant for all analyses conducted.

Results

In 2014/2015, among the Belgian population 3–64 years, the proportion of foods and ingredients consumed classified as ultra-processed was 36.4% and the proportion of foods and ingredients consumed classified as unprocessed/minimally processed was 42.4%. When comparing 2004 and 2014–2015 for the population group 15–64 years, the proportions of foods/ingredients consumed classified into the different NOVA groups were statistically significantly different (p < 0.001) (Table 1):

Table 1 Proportion of foods and ingredients consumed in function of the extent of processing (according to the NOVA classification [8]), by survey year and by age group, Belgian national food consumption survey, 2004 and 2014–2015
Full size table

  • The proportion of foods consumed classified as processed was lower in 2014

  • The proportion of foods consumed classified as unprocessed or minimally processed was higher in 2014/2015

  • There was only a small difference in the proportion of foods consumed classified as ultra-processed between the two survey periods: 33.6% in 2004 and 34.8% in 2014/2015.

Overall, in 2014–2015 in Belgium the average usual proportion of daily energy intake from UPF and unprocessed or minimally processed foods was 29.9% and 21.3%, respectively (Table 2). Consumption of UPF was negatively associated with consumption of unprocessed/minimally processed foods (r = − 0.17, p < 0.001).

Table 2 Usual dietary share of (% daily energy intake from) unprocessed/minimally processed and ultra-processed food products in the Belgian population (3–64 years) by population group (Belgian national food consumption survey 2014–2015)
Full size table

The consumption of daily energy from UPF was not significantly different between men and women. However, women consumed a significantly higher proportion of their daily energy from unprocessed or minimally processed foods, compared to men (Table 2). After excluding misreporters (n = 818), the average usual proportion of daily energy intake from UPF increased to 32.6% (95% CI 31.0–33.4%) for the total population 3–64 years with no difference between men and women (data not shown).

Young children (3–9 years) consumed a significantly higher proportion of their daily energy from UPF (33.3% on average) compared to adolescents and adults (29.2% and 29.6% on average, respectively) in 2014–2015. There were no differences in the consumption of UPF between different socio-economic population groups (secondary education or lower, short-term education, long-term education). However, individuals with long-term education (or parental education in the case of children) consumed a significantly higher proportion of their daily energy from minimally or unprocessed foods compared to those with secondary education or lower. There were no differences in consumption of either UPF or unprocessed/minimally processed foods between people with obesity, overweight or normal weight (Table 2).

For the population group 15–64 years, there was no significant difference for both men and women in regards to the consumption of UPF between 2004 and 2014–2015. There was an increase in the consumption of unprocessed/minimally processed foods between 2004 and 2014–2015, but this was only significant for the total population (Table 3).

Table 3 Usual dietary share of (% daily energy intake from) unprocessed/minimally processed and ultra-processed food products in Belgium for the population group 15–64 years (Belgian national food consumption survey 2004 and 2014–2015)
Full size table

For the total population 3–64 years, processed meat products (14.3%), cakes, pies and pastries (8.9%) and dry cakes and sweet biscuits (7.7%) and carbonated soft drinks (6.7%) were the biggest contributors to the dietary share of UPF in 2014–2015 and pasta, rice and other grains (19.8%), fruits (18.5%), potatoes (9.4%) and chicken (7.8%) were the biggest contributors to the dietary share of unprocessed and minimally processed food products. There were some slight differences between age groups by gender in the contributions of food subgroups to UPF intakes (Table 4).

Table 4 Top contributing food subgroups (results for the top contributing food sub groups are the same for interview day 1 and 2) to the consumption of unprocessed/minimally processed and ultra-processed foods in Belgium for the total population and specific age groups by sex (Belgian national food consumption survey 2014–2015)
Full size table

The UPF dietary share was significantly lower (p = 0.0194) and the dietary share of unprocessed/minimally processed foods was significantly higher (p < 0.001) during consumption days when the survey participants met the WHO salt intake recommendation (≤ 5 g/day); this was the case for all age groups (Figs. 1, 2). The UPF dietary share was significantly lower (p = 0.0034; most age groups) and the dietary share of unprocessed/minimally processed foods was significantly higher (p < 0.001; all age groups) during consumption days when saturated fat was ≤ 10% of energy intake.

Fig. 1
figure 1

Dietary share of ultra-processed food products comparing days when WHO guidelines for fruit and vegetables (FV), sodium, saturated fat and trans-fat are met/not met for different age groups by sex; Belgian national food consumption survey 2014/2015. *%E from UPF between meeting and not meeting the guidelines is significantly different for sodium and saturated fat; p < 0.05

Full size image
Fig. 2
figure 2

Dietary share of unprocessed/minimally processed food products (MPF) comparing days when WHO guidelines for fruit and vegetables (FV), sodium, saturated fat and trans fat are met/not met for different age groups by sex; Belgian national food consumption survey 2014/2015. *%E from MPF between meeting and not meeting the guidelines is significantly different for fruit and vegetables, sodium and saturated fat; p < 0.001

Full size image

For most age groups, the UPF dietary share was lower on consumption days during which the individuals were meeting the WHO guideline for fruit and vegetable consumption (≥ 400 g/day), however this was not significant in the overall population (p = 0.0993). The dietary share of unprocessed/minimally processed foods was significantly higher during consumption days when the participants met the WHO salt and fruit and vegetable intake (≥ 400 g/day) recommendations; this was the case for all age groups (p < 0.001). For trans-fatty acids, the differences are not statistically significant but the number of individuals not meeting the WHO guidelines is very low in Belgium (≤ 15 individuals for each of the age groups) (Figs. 1, 2).

For the total population, age, region, BMI class, total energy intake and frequency of breakfast consumption were significantly associated with UPF dietary share. Compared to the youngest age group (3–5 years), all but one age group (51–64 years) consumed a significantly lower proportion of daily energy from UPF. Compared to Flemish region residents, residents from Brussels and the Walloon region consumed a significantly higher proportion of energy from UPF. Compared to normal weight individuals, individuals with obesity consumed a significantly lower proportion of their daily energy from UPF (Table 5). These differences between normal weight individuals and individuals with obesity became non-significant after exclusion of misreporters (data not shown). For the total population, age, region, self-perceived health and frequency of breakfast consumption were significantly associated with dietary share of unprocessed/minimally processed foods (MPF). Total energy intake was significantly inversely associated with MPF dietary intake. We observed a significant effect modification in the relationship between BMI class and MPF consumption by sex.

Table 5 Association of socio-demographic, socio-economic and health status factors with mean proportion of daily energy intake from unprocessed/minimally processed foods and ultra-processed foods over the two consumption days, all age groups (Belgian national food consumption survey 2014/2015)
Full size table

Compared to men, women with obesity consumed a significantly higher proportion of daily energy from MPF than normal weight women. Compared to participants from low educated households, those from higher educated households consumed a significantly higher proportion of daily energy from MPF (Table 5). While individuals with good and fair self-perceived health consumed significantly less MPF than individuals with very good perceived health, consumption of MPF between individuals with very good and very bad self-perceived health was not significantly different. When excluding misreporters, there was no more association between consumption of MPF and self-perceived health (data not shown).

In the model with adolescents and adults only, individuals who consumed more than two meals a day with family compared to those who did never, consumed a significantly higher proportion of energy out of MPF and never-smokers also consumed a significantly higher proportion of energy out of MPF compared to those smoking every day (data not shown).

Discussion

About one-third of daily energy intake is contributed by ultra-processed foods (UPF) in the Belgian population, and young children consume the largest proportion of their daily energy intake from UPF. Other studies, such as from the US [43], Canada [21] and Chile [18] also found that children consume the most UPF compared to other population groups.

Higher consumption of UPF is associated with lower diet quality in Belgium (lower fruit and vegetable consumption for some age groups, higher salt and saturated fat intake); the opposite is observed for higher consumption of unprocessed/minimally processed foods. The association of intake of UPF with the intake of trans-fatty acids is mixed, but this is probably due to the fact that the average population intake of trans fatty acids has decreased since 2004 in Belgium and is below the WHO recommended intake of < 1% of total energy intake in 2014 [44].

Estimates of UPF purchases calculated from national household budget surveys (conducted in Europe between 1991 and 2008) showed that the average household availability of UPF ranged from 10% of total purchased dietary energy in Portugal to 50% in the UK. After adjustment for confounders, each percentage point increase in the household availability of UPF resulted in an increase of 0·25 percentage points in obesity prevalence [29]. In Belgium UPF were found to contribute about 46% to total purchased dietary energy, which is higher than the average usual proportion of daily energy intake from UPF (30%) found in this study. However, food consumption surveys usually provide more details on the foods consumed compared to household budget surveys which are based on purchases. The data of this study can thus be considered as more accurate.

Some countries like Brazil [45] and Uruguay [46] include the concept of UPF in their food-based dietary guidelines. France has recently set a target to reduce population consumption of ultra-processed foods with 20% by 2022 [47]. In addition, some countries, like Chile [48], are taking action to limit the marketing of UPF products to children, and other countries, like Mexico [49, 50] and Hungary [51], introduced fiscal measures, such as a tax on junk food. The tax in Mexico has already demonstrated a positive effect with significant declines in the purchases of both solid and liquid UPF observed in a national urban sample [49, 52].

In Belgium, the latest food consumption survey showed that about 64% of the Belgian population was in favour of regulating the restriction of unhealthy food marketing to children, 78% in favour of measures on food reformulation and 57% in favour of fiscal measures to increase taxes on unhealthy foods and simultaneously decrease taxes on healthy foods [44]. Such measures are likely needed to reduce UPF consumption since these products are considered highly palatable, cheap, conducive to excessive consumption and aggressively marketed, making them highly profitable for food manufacturers [53].

In addition to policies, improving food preparatory skills is important, as a UK study found that better home food preparation skills and more frequent use of those skills was associated with lower consumption of UPF [54]. Interestingly in our study, sharing meals with family more often, was significantly associated with higher consumption of unprocessed/minimally processed foods.

Strengths of this study include the use of national food consumption survey data representative for the Belgian population and different population groups, the detailed descriptions on the foods consumed and the use of statistical modelling to calculate usual intakes (i.e., remove within-person day-to-day variability). In addition, this is the first study assessing UPF intake among the Belgian population.

The limitations include lack of data on the intake of added and free sugar and salt added at the table or during cooking and the fact that almost one-third of survey participants were identified as misreporters. In addition, it was not possible to disaggregate some of the home-prepared composite foods into ingredients.

Conclusion

About one-third of daily energy intake is contributed by UPF in Belgium and young children have the highest intakes of those food products. Internationally recommended policies to limit availability, affordability and marketing of UPF need to be implemented to reduce UPF consumption in Belgium.