Who consumes ultra-processed food? A systematic review of sociodemographic determinants of ultra-processed food consumption from nationally representative samples · Dicken SJ, Qamar S, Batterham RL

Results

The database search retrieved 1,131 results. 245 results were retained for full-text screening. 55 studies were included in the systematic review. The results are detailed in the PRISMA flowchart (Figure 1).

Several studies appeared to meet the inclusion criteria, but were excluded as: authors did not respond to emails requesting further detail on statistical associations (19,20), statistical values were not available after request (21), or were not nationally representative (22,23). One study was excluded, as despite being a nationally representative cohort the authors explicitly outlined that the analytical sample could not be considered nationally representative (24). One case-control study was not deemed to be nationally representative, although the authors suggested it was (25). One paper from Brazil was excluded based on the use of a crude FFQ, reporting the frequency of consumption of 10 items to determine UPF intake (26). One paper from Mexico was included as participants were randomly selected from a nationally representative cohort, despite the small sample size (27).

Study characteristics

In total, 55 studies were included in the systematic review, covering 32 countries and 36 nationally representative surveys. 54 studies were from 17 individual countries from 19 nationally representative surveys, with one additional study spanning 22 European countries (27–81). The main study characteristics are reported in Table 1. Sample size varied from 359 (32), to 57,423 (50). Five of the studies were included after considering citations of the included papers from the systematic search (28,35,39,61,63).

13 sociodemographic characteristics were assessed as predictors of UPF, including age, gender, race/ethnicity, income, education level, socioeconomic status/occupation/occupational social class, food security, marital status, household status (number and type of individuals (child, adult, elderly), rural/urban location, region of the country, immigrant status/country of birth and indigenous identity. One sociodemographic association was reported in 15 countries, and at least three sociodemographic associations were reported in 17 countries. Seven or more sociodemographic associations were reported in eight countries.

The included studies are detailed in Table 1, with statistical associations between sociodemographic variables and UPF intake reported in each study presented in Table 2.

Associations by country

Australia

In Australia, average UPF intake was 38.8% of total energy [SE: 0.2] (31). In unadjusted models, younger age, male gender, second/third/fourth household income quintiles (the combined income of all household members), a lower education level, a lower Socio-Economic Index for Areas (SEIFA) (greater area-level disadvantage), living in inner regional Australia (vs. living in a major city of Australia) and being Australian born or from an English-speaking country were associated with higher intakes of UPF (28–31). The associations remained unchanged in adjusted models (adjusted for all other sociodemographic variables and diet quality) (31), except for gender and rurality/urbanisation, which were no longer significant.

In the adjusted model, 19-30-year-olds consumed 8.3% [95%CI: 6.4, 10·3] more UPF as a proportion of total energy than 51-70-year-olds (31). The middle quintiles (second, third and fourth) of income had the highest intakes of UPF, with the highest income quintile associated with the lowest intake of UPF, 4.6% less than the second quintile (36.5% [SE:0.7] vs. 41.1% [SE: 0.8] (31). Individuals with the lowest education level (incomplete high school education or lower) consumed 2.3% [95%CI: 0.2, 4.5] more UPF as a proportion of total energy than individuals with the highest levels of education (tertiary qualification) (39.9% [SE: 0.7] vs 37.6% [SE: 0.6]).

Barbados

In Barbados during 2012-13, the average intake of UPF was 41% of total energy (838kcal from UPF/day [95% confidence interval (95%CI): 791, 885]). A younger age was associated with a roughly 20% greater UPF intake (25-44 years old: 889kcal/day [95%CI: 836, 942], 45-64 years old: 737kcal/day [95%CI: 693, 781], but gender and level of education were not significantly associated with UPF intake (32).

Belgium

In Belgium, the average UPF intake was 29.9% [95%CI: 29.0, 30.8] in 2014-15 (33,34). Unadjusted UPF intakes were higher with a younger age, but there was no difference in UPF intakes across education levels or between genders (males: 29.6% [95%CI: 28.0, 31.0] vs. female: 29.2% [95%CI: 28.0, 30.9]) (33,34). However in 2004, males had a higher UPF intake than females (32.3% [95%CI: 30.9, 34.3] vs. 28.9% [95%CI: 27.1, 30.2]) (33).

In adjusted models (adjusted for age, region, BMI and breakfast consumption frequency), age and region of Belgium were significantly associated with UPF intake. 3-5- and 51-64-year-olds consumed around 6-8% more UPF as a proportion of total energy than 6-50-year-olds, with 3-5-year-olds consuming 8.6% [SE: 2.1] more UPF as a proportion of total energy than 35-50-year-olds. Individuals living in Brussels region (+6.1% [SE: 1.2] or Walloon region (+8.1% [SE: 0.8] also had significantly higher UPF intakes than individuals living in Flanders region (33).

Brazil

In Brazil during 2008-9, average UPF intake varied across studies, from 20-30% of total energy (20.5% [95%CI: 20.2, 20.8] (38), 22.1% (35), 23.8% (36), 29.6% (37)). A younger age, female gender, white ethnicity (vs. African-descendent or other ethnicity), higher income, higher education level, urban residence and living in the South and South East regions of Brazil were associated with higher intakes of UPF, or were more likely to be in the highest vs. lowest quintile of UPF intake ((≥ 44% vs. ≤ 13% of TEI) (35–37,37).

In unadjusted associations, females (21.8% [95%CI: 21.3, 22.2]) consumed 2.6% more UPF as a proportion of total energy than males (19.2% [95%CI: 18.7, 19.7]). 20-39-year-olds consumed 6.3% more energy as UPF than adults 60 years or older (21.3% [95%CI: 20.8, 21.9] vs. 15.0% [95%CI: 14.2, 15.8]), and 10-19-year-olds (26.8% [95%CI: 26.1, 27.6]) consumed over 11% more. Individuals in the highest income tercile consumed over 10% more UPF as a proportion of total energy compared with individuals in the lowest income tercile (26.3% [95% CI: 25.7, 26.9] vs. 15.1% [95%CI: 14.6, 15.5]). Similar magnitude differences in UPF intake were seen across the highest (28.5%) vs. lowest (15.7%) terciles of income in another study (p <0.001) (35), and when grouped by multiples of minimum wage (MW), with individuals earning >3xMW consuming 30.3% of total energy from UPF, compared to those earning <0.5MW, who consumed 16.3% of total energy from UPF (36). Individuals living in an urban residence also consumed nearly 10% more total energy from UPF than individuals in rural settings (22.1% [95%CI: 21.7, 22.5] vs. 12.7% [95%CI: 12.3, 13.2]). Those in South and Southeast Brazil consumed 25.7% [95%CI: 25.0, 26.4] and 23.6% [95%CI: 23.0, 24.2] of energy as UPF, respectively, around 10% more than in the North (14.8% [95%CI: 14.3, 15.4]) and North East (14.9% [95%CI: 14.5, 15.3]) regions (38).

In 2017-18, a younger age was associated with a higher UPF intake. 15-19-year-old males (25.1% [95%CI: 23.3, 26.9]) consumed relatively over 100% more UPF than males aged 80+ (12.7% [95%CI: 10.5, 15.0]), and over a 50% relative increase in 15-19-year-old females (26.2% [95%CI: 24.5, 28.0]) compared with older females (80+) (17.9% [95%CI: 13.7, 22.1]) (39).

Canada

In Canada in 2004, a younger age, male gender, lower education level, rural residence and non-immigrant status were associated with greater unadjusted intakes of UPF (40–42). Sociodemographics remained significantly associated with UPF intake in the adjusted model (adjusted for age, gender, education, income, physical activity, smoking status, immigration status, and residential area), except for rural residence, which became non-significant (41). Family income per capita was not significantly associated with UPF intake in either unadjusted or adjusted models from 2004 (40,41). Non-immigrants consumed over 10% more energy from UPF than non-immigrants (47.8% [SE: 0.3] vs. 36.5% [SE: 0.3], p < 0.05) (41).

In 2015, the average UPF intake in 2015 was around 46-47% (45.7% [95%CI: 45.0, 46.4] (42), 46.8% (SE: 0.4) (44)). A younger age, male gender, higher income, lower education level, higher level of food insecurity, rural residence, non-immigrant status and Indigenous identity were associated with higher intakes of UPF, or were more likely to be in the highest vs. lowest tercile of UPF intake (72.8% vs. 24.4%) (42–44).

Adult males (45.4% [95%CI: 43.8, 47.0] consumed more energy from UPFs than adult females (41.6% [95% CI: 40.2, 43.0]), with no significant difference between males and females aged 55 or older (42). Income was associated with UPF intake in 2015 (p = 0.0143), with the highest income quintiles more likely to be in the highest vs. lowest tercile of UPF intake (44). One paper in 2015 reported age and gender associations with UPF intake across levels of food insecurity (43). Males (47.0% [SE: 3.7]) and females (45.8% [SE: 2.4]) aged 19-64 with severe food insecurity consumed around 8-10% more of total energy from UPFs than males (37.5% [SE: 0.66]) and females (37.6% [SE: 0.56]) aged 19-64 with food security (both comparisons p < 0.05, p-trend across levels of food insecurity: males = 0.009, females = 0.003) (43).

In trend analyses from 2004 to 2015, UPF intake significantly increased in older (55 or older) males (from 42.5% [95%CI: 41.5, 43.6] to 45.3% [95%CI: 43.9, 46.7]) and females (from 41.7% [95%CI: 40.6, 42.8] to 45.2% [95%CI: 44.0, 46.4]), but significantly decreased in children aged 2-12, adolescent males and females aged 13-18, and adult males and females 19-54 (42).

Chile

In Chile in 2010, the average UPF intake was 28.6% [95%CI: 27.7, 29.6] (45). A younger age, higher family income, urban residence and living in the Metropolitan region were associated with higher intakes of UPF in both unadjusted and adjusted (for all other sociodemographic variables) models (45). Gender and the level of education of the head of the household were not significantly associated with UPF intake in either unadjusted and adjusted models.

In the adjusted model, 2-19-year-olds consumed over double the quantity of UPF of adults 65 or older (2-19-year-olds: 38.6% [95%CI: 36.7, 40.6] vs. 65 or older: 18.3% [95%CI: 16.8, 19.8]). There was a linear trend, with 20-49-year-olds consuming 26.7% [95%CI: 25.2, 28.2] and 50-64-year-olds consuming 21.8% [95%CI: 19.7, 24.0] of total energy from UPFs (p-trend < 0.001) (45). There was also a linear trend in UPF intake across family incomes. Individuals from families with the highest incomes consumed over 4% more of total energy from UPFs than individuals from families with the lowest family incomes (≥6xMW: 30.1% [95%CI: 27.7, 29.6], 3-5xMW: 30.0% [95%CI: 27.8, 32.2], 2xMW: 28.7% [95%CI: 27.2, 30.3], 1xMW: 25.8% [95%CI: 24.0, 27.6]). Individuals living in urban residences consumed 29.3% [95%CI: 21.9, 25.5] of total energy from UPFs, compared with 23.7% [95%CI: 28.3, 30.4] in rural residences, nearly a 25% greater relative intake. Individuals living in the Metropolitan region of Chile had the highest UPF intakes of all regions (30.2% [95%CI: 28.6, 31.8]), approximately 2-3% greater than other regions, and significantly greater than the South, where the average adjusted UPF intake was 26.7% [95%CI: 24.8, 28.6] (45).

Colombia

In Colombia, average UPF intake was 15.9% of total energy in 2005, spanning from 0.2% to 41.1% across quintiles of UPF intake (46). A younger age, female gender, higher socioeconomic status (based on the System for the Selection of Beneficiaries of Social Programs (SISBEN) composite index), urban residence and living in Bogotá were significantly associated with higher intakes of UPF in unadjusted and adjusted (adjusted for all aforementioned sociodemographic variables) models (46).

In the adjusted model, 2-19-year-olds consumed nearly 8% more UPF as a proportion of total energy than adults aged 50 or over (2-19-year-olds: 19.3% [SE: 0.3], 10-19-year-olds: 19.3% [SE: 0.2], 50 or over: 11.4% [SE: 0.4]). Individuals in the highest socioeconomic level (22.8% [SE: 1.0]) consumed nearly twice the quantity of UPF (over 10% more as a proportion of total energy) of individuals in the lowest socioeconomic level (12.7% [SE: 0.3]). There was a small but significant difference between genders, whereby males consumed 15.5% [SE: 0.2] of total energy from UPF compared to 16.2% [SE: 0.2] in females (P = 0.007). Individuals living in urban residences (17.3% [SE: 0.2]) had 4.6% and 6.1% higher intakes of UPF than people from central (12.6% [SE: 0.5]) or rural (11.2% [SE: 0.6]) residences, respectively (both comparisons, p < 0.001). Regionally, the highest intakes of UPF were in residents of Bogotá (21.6% [SE: 0.5]) followed by the Eastern region of Colombia (18.1% [SE: 0.4]). The lowest intakes were in the Atlantic region of Colombia (12.7% [SE: 0.3]), nearly half the levels reported in Bogotá (46).

France

The average UPF intake in France was 31.1% [95%CI: 30.3, 31.9] in 2006-7 (47), and 30.6% [SD: 15.8] in 2014-15 (48). In 2006-7, a younger age, a complete high school or greater education level, occupation (management/intermediate profession, self-employed/farmer, manual worker/employee, homemaker or disabled person or other vs. retired persons) and urban residence were associated with higher intakes of UPF (47). 18-39-year-olds consumed nearly double the amount of UPF as a proportion of total energy than adults aged 60 or older (39.1% [95%CI: 37.8, 40.5] vs. 21.6% [95%CI: 20.4, 22.8]). Homemakers, disabled persons, and other occupations had the highest UPF intake (35.9% [95%CI: 34.1, 37.7]), followed by management or intermediate professions (32.2% [95%CI: 30.9, 33.4]), manual workers or employees (32.7% [95%CI: 31.3, 34.2]) and self-employed individuals or farmers (28.1% [95%CI: 25.1, 31.2]). Retired individuals had significantly lower UPF intake than all other occupations, around a third less, at 22.3% of total energy [95%CI: 21.1, 23.5]. Individuals with complete high school education (32.9% [95%CI: 31.8; 34.1]), completing a technical course (32.2% [95%CI: 30.3, 34.0] or University education (31.9% [95%CI: 30.4, 33.4]) had similarly high intakes, whereas individuals with an incomplete high school education had around 6% lower intake as a proportion total energy (26.5% [95%CI: 24.9, 28.1]), a roughly 20% lower relative intake (47). Urban residents consumed 3% more UPF as a proportion of total energy than rural residents (31.9% [95%CI: 30.9; 32.8] vs. 28.9 [95%CI: 27.4, 30.4]).

During 2014-15, age, education level, occupation, marital status, food insecurity and rurality/urbanisation were associated with higher intakes of UPF (48). Individuals in the highest vs. lowest tercile of UPF intake (34.1-78.9% vs. 0.1-20.6% of energy from UPF) were more likely to be younger, have middle or secondary school education, be an employee, manual worker, have an intermediate profession or be inactive, have moderate or severe food insecurity, be single or in an unmarried couple, or live in a city with 100,000 or more inhabitants (48). Individuals in the lowest vs. highest tercile of UPF intake were more likely to be older, have a primary school education, be retired, a farmer, craftsman, shopkeeper or business owner, have a rural residence, be married or widowed or have food security (48).

Across both 2006-7 and 2014-15, UPF intake did not significantly differ with gender (47,48), nor with the region of France in 2014-15.

Italy

In Italy in 2010-13, the average unadjusted UPF intake among Italian adults (aged 20-97) was 17.3% of total energy [95%CI: 17.1, 17.6], rising to 25.9% [95%CI: 24.8, 27.0] in children and adolescents aged 5-19. A younger age, female gender, occupation other than being retired, marital status, urban residence and region of Italy were associated with higher intakes of UPF in the adjusted model (adjusted for the aforementioned sociodemographic predictors, education level, smoking, physical activity and disease history) (49). Education level was not significantly associated with UPF intake.

In the adjusted model, differences in adult UPF intake across levels of sociodemographics varied by around 10% as a relative proportion of UPF intake (49). 20-40-year-olds consumed 3.1% [95%CI: 1.8, 4.4] more of total energy from UPF than adults aged 65 or older. Females consumed 1.28% [95%CI: 0.68, 1.89] more UPF than males. Individuals who were unmarried (+1.26% [95%CI: 0.37, 2.15]), separated or divorced (+1.88% [95%CI: 0.38, 3.38]) and widowed (+1.16% [95%CI: 0.07, 2.24]) consumed more energy from UPF than individuals who were married. Individuals in North Italy consumed 0.73% [95%CI: 0.14, 1.32] more UPF as a proportion of total energy than individuals in South Italy, but intakes in North or South Italy did not significantly differ to those living in Central Italy. Urban residents consumed 1.64% [95%CI: 0.87, 2.42] more UPF as a proportion of total energy than rural residents (49), and retired persons consumed significantly less UPF than all other occupations (manual, non-manual, housewife, student or unemployed), nearly 2% less than manual occupations (-1.87% [95%CI: -0.91, -2.83]).

Korea

In Korea across 2010 to 2018, the average UPF intake was 24.9% [SE: 0.1] (50). A younger age, male gender, lower income, mid/high education level and urban residence were associated with higher unadjusted intakes of UPF. All sociodemographic predictors remained significant in the adjusted model (adjusted for all aforementioned sociodemographics), except for household income (50).

In the adjusted model, there was a linear trend of decreasing UPF with increasing age (p-trend < 0.05). Adolescents (13-19-year-olds) consumed the highest amount of UPF, over double the amount of UPF of adults 65 or older as a proportion of total energy (33.8% [95%CI: 32.9, 34.6] vs. 16.3% [95%CI: 15.8, 16.7]). Individuals with a high school (26.4% [95%CI: 25.9, 26.9]), or college or higher education level (26.3% [95%CI: 25.8, 26.9]) consumed about 10% relative greater UPF intake than individuals with a middle school or lower education level (23.4% [95%CI: 23.0, 23.8]). Males and urban residents consumed 3% more UPF relatively than females and rural residents, respectively (both comparisons: 25.8% [95%CI: 25.5, 26.1] vs. 25.0% [95%CI: 24.4, 25.6]) (50).

In 2016-18, a younger age (greatest in adolescents), male gender, mid/high education level, living alone and urban residence were associated with higher unadjusted intakes of UPF, or were more likely to be in the highest (43.6%) vs. lowest (6.9%) tercile of UPF intake (50–52,81). Unadjusted UPF intakes did not significantly vary across household income levels in 2016-18 in two smaller KNHANES samples (n = 7,364, aged 19-64 (51), and n = 9,188, aged 30-79 (52)). However, UPF intake did significantly vary in a larger sample from 2016-18 (n = 19,216, aged 1 or older) (50), whereby individuals in the second and third quartiles (26.7% [95%CI: 26.1, 27.2]) or highest quartile (27.2% [95%CI: 26.5, 27.8]) of household income had significantly higher UPF intakes than individuals in the lowest household income quartile (22.0% [95%CI: 20.9, 23.1]). In the only study reporting adjusted (adjusted for aforementioned sociodemographics, smoking, alcohol and physical activity) UPF intakes from 2016-18, where the average UPF intake was 26.8% [SE: 0.3], age, gender and education level remained significantly associated with UPF intake, but household income, household status, marital status and rurality/urbanisation were not significantly associated with UPF intake (51). Adjusted intakes of UPF in males were 1.4% higher than females (27.6% [SE: 0.4] vs. 26.2% [SE: 0.4]), p = 0.0165). 19-29-year-olds consumed two-thirds more UPF than 50-64-year-olds (34.6% [SE: 0.8] vs. 20.6% [SE: 0.4]), p <0.0001), with a linear trend of decreasing UPF intake with older age (p-trend < 0.0001). Similar to the associations in 2010-18, individuals with a high school education (27.6% [SE: 0.4], p < 0.01), or college or higher education (26.8% (SE: 0.4), p < 0.05), consumed about 10% more UPF relative to individuals with a middle school or lower education level (25.0% [SE: 0.4], p < 0.05) (51).

Average UPF intake increased over time in Korea, from 23.1% [95%CI: 22.7, 23.5] in 2010-12, to 25.5% [95%CI: 25.1, 25.9] in 2013-15, to 26.1% [95%CI: 25.7, 26.5] in 2016-18 (p < 0.0001). UPF intake significantly increased across all sociodemographic strata (age, gender, rural/urban residence, education level and household income). In particular, 20-49-year-olds increased their UPF intake to the greatest extent, by 5% from 2010 to 2018, from 24.8% [95%CI: 24.3, 25.4] to 29.8% [95%CI: 29.2, 30.4] (50).

Mexico

In Mexico in 2012, the average UPF intake (aged 1 or older) was 29.8% [SE: 0.4] of total energy (54). A younger age, higher head of household education level, higher socioeconomic status, urban residence and living in the Northern region of Mexico were more likely to be in the highest (64.2% [range: 51.8-100%]) vs. lowest (4.5% [range: 0-11.8%]) quintile of UPF intake (all sociodemographic variable distributions across quintiles p < 0.001) (53). Sociodemographic variables remained significantly associated with UPF intake after adjustment for all other sociodemographic variables (54). However, gender was not significantly associated with UPF intake in unadjusted or adjusted models (53,54).

In one study of adults only from ENSANUT, adults younger than 60 consumed 21.4% of total energy from UPF [95%CI: 18.8, 24.0], compared with 14.2% [95%CI: 10.7, 17.6] in adults 60 or older, a 50% relative increase (27). In the adjusted model across all age groups, pre-school-aged children (+12.5% [95%CI: 10.9, 14.1]), school-aged children (+3.8% [95%CI: 2.2, 5.4], and adolescents (+3% [95%CI: 1.1, 4.9]) all consumed greater amounts of UPF than adults. Individuals in North Mexico and Central Mexico consumed 8.4% [95%CI: 6.6, 10.1] and 2.7% [95%CI: 1.2, 4.1] more UPF as a proportion of total energy than individuals in South Mexico. Individuals from households with the highest head of household education level (college education) consumed 7.8% [95%CI: 4.3, 11.4] more UPF as a proportion of total energy than individuals from a household with a head of household without any education. Individuals from a household with a head of household with an intermediate education level consumed 1.9% [95%CI: -0.5, 4.3] (elementary), 3.4% [95%CI: 0.8, 6.1] (middle school) and 4.3% [95%CI: 1.1, 7.4] (high school) more UPF as a proportion of total energy than individuals from a household with a head of the household without any education. Individuals in the middle and highest terciles of socioeconomic status (index based on household characteristics and basic goods and services) consumed 4.5% [middle tercile: 95%CI: 2.8, 6.2, highest tercile: 95%CI: 2.5, 6.5] more UPF as a proportion of total energy than individuals in the lowest tercile. Urban residents also consumed 5.6% [95%CI: 4.2, 7.0] more UPF as a proportion of total energy, compared to rural residents (54).

Netherlands

In the Netherlands in 2012-2016, the average UPF intake was 893g of UPF per 2000kcal [95%CI: 879, 907], or 61% of total energy intake (55). A younger age, middle education level (vs low or high) and urban residence were associated with higher UPF intake, whereas gender was not significantly associated with UPF intake (55).

Children and adolescents consumed around double the amount of UPF as older adults, and around 30-50% more than younger and middle-aged adults. For example, 4-8-year-olds consumed 1,252g of UPF per 2000kcal [95%CI: 1,217, 1,288], compared with 962g of UPF per 2000kcal [95%CI: 921, 1,003] in 19-30-year-olds, and 632g of UPF per 2000kcal [95%CI: 607, 656] in 71-79-year-olds. Individuals with a middle education level (intermediate vocational or higher secondary education) consumed around 8-10% more UPF relatively than individuals with lower (primary, lower vocational or advanced elementary) or higher (higher vocational or university education) education levels, consuming 939g of UPF per 2000kcal [95%CI: 916, 962], compared with 871g [95%CI: 838, 903] in individuals with a low education level, and 850g [95%CI: 830, 871] in individuals with a high education level. Individuals living in regions with a high degree of urbanisation (≥1,500 addresses/km2) consumed 916g of UPF per 2000kcal [95%CI: 891, 942], compared with 876g of UPF per 2000kcal [95%CI: 856, 896] in individuals living in regions with a low degree of urbanisation (<1,000 addresses/km2) and 898g of UPF per 2000kcal [95%CI: 868, 928] in individuals living in regions with a moderate degree of urbanisation (1000–1500 addresses/km2) (55).

Portugal

In Portugal in 2015-16, average UPF intake was 23.8% of total energy, or 257g [interquartile range: 141, 426] per day (57). Younger adults consumed more UPF than elderly adults (56). Other associations were stratified by male and female gender (57). In unadjusted models stratified by gender, crude UPF intake was significantly higher with a younger age (highest in adolescents), mid-high education level, in single, divorced or widowed individuals, individuals living in a larger household and living in Lisbon and Azores regions of Portugal (57). All variables remained significantly associated with UPF intake in adjusted (adjusted for age, education level and non-UPF intake) models, except for household status in both males and females, and marital status in females, which became non-significant. Rurality/urbanisation and food insecurity were not significantly associated with UPF intake in unadjusted or adjusted models(57).

In the adjusted models, adolescents (aged 10-17) had the highest UPF intake, with female adolescents consuming 192g [95%CI: 135, 249] more than female older adults, and male adolescents consuming 327g [95%CI: 277, 377] more than male older adults (aged 45-64). Female and male older adults (aged 45-64) consumed 63g more [95%CI: 34, 91] and 51g more [95%CI: 9, 93] than elderly females and males (aged 65-84). The difference in UPF intake across ages was greater in males than females (57). Females in Alentejo (+50g [95%CI: 9, 90]) and Algarve regions (+36g [95%CI: 1, 70]) consumed more UPF than females in North Portugal, and males from Lisbon consumed 76g [95%CI: 19, 133] more UPF than males in North Portugal. Males and females with the highest level of education (more than 12 years) consumed 68g [95%CI: 12, 124] and 51g [95%CI: 16, 86] more UPF per day than males and females with the lowest level of education (6 years or less), respectively, a roughly 20-25% relative increase. However, a lower education level was associated with a higher UPF intake in children. Single, divorced or widowed males consumed 48g [95%CI: 1, 96] more UPF per day than married males, or males in a couple.

Spain

In Spain, the average intake of UPF significantly increased from 1991 to 2008, accounting for 24.4% [SD: 14.0] of total energy in 1991, 25.6% [SD: 16.3] in 1996, 27.5% [SD: 19.2] in 2004, and 31.1% [SD: 19.0] in 2008 (58).

In 1991, a younger age was inversely related with higher intakes of UPF (ρ = -0.53, p < 0.0001) (59). In 2008-10, Individuals in the highest (42.8% [SE: 0.2]) vs. lowest (8.7% [SE: 0.1]) quartile of UPF intake were more likely to be younger and have a primary level education, compared with individuals in the lowest vs. highest quartile of UPF intake, who were more likely to have no formal education and be living alone (60). Gender proportions did not significantly differ across quartiles of UPF intake.

In the adjusted model (adjusted for year of cohort, age, gender, BMI and total energy intake) across 1991 to 2008, a younger age (-0.15% [SE: 0.01] per year of age) and female gender (1.1% [SE: 0.3] greater UPF intake than males) had significantly higher UPF intakes (58). UPF intake also varied across regions. In 2008, UPF intake was more than 30% of total energy in all regions, but 5% higher in North Spain (36.0% [SD: 18.3]) than South or Central South Spain (South: 31.3% [SD: 18.3], Central South: 30.2% [SD: 17.1]) (58).

Switzerland

In Switzerland during 2014-15, the average UPF intake was 28.7% of total energy [IQR: 19.9, 38.9] (61). In unadjusted associations, a younger age, living in the German-speaking region (vs. French- or Italian-speaking regions) and Swiss nationality (vs. non-Swiss) were significantly associated with a higher intake of UPF (61). After adjustment for other sociodemographics including income, male gender also became significantly associated with higher UPF intake. Household size and education level were not significant in adjusted or unadjusted models.

Across ages, unadjusted median UPF intakes were 8.5% higher as a proportion of total energy in 18-29-year-olds (34.8% [IQR: 24.5, 45.0], compared with 65-75-year-olds (26.3% [IQR: 17.1, 35.0], p = 0.001). Individuals from German-speaking region (29.6% [IQR: 20.9, 39.6]) also consumed 1.5-2.5% more UPF than other regions (French-speaking: 27.2% [IQR: 17.7, 37.1]; Italian-speaking: 28.0% [IQR: 16.9, 39.4], p = 0.002). Swiss nationals consumed relatively around 12% more UPF as a proportion of total energy than non-Swiss nationals (29.2% [95%CI: 20.3, 39.0] vs. 26.1% [IQR: 17.5, 37.1], p = 0.002) (61).

UK

In the UK, average UPF intake was 51.3% [SD: 13.1] in 2008-9 (≥19 years) (62), 53.1% across 2008-12 (≥18 years)(64), 56.8% [SE: 0.2] across 2008-14 (≥1.5 years)(65), 54.3% [SE: 0.4] across 2008-16 (19-96 years) (66), and 54.0% in 2014-16 (≥4 years) (67). From 2008 to 2016, UPF intakes have been relatively consistent, with no significant linear trends in UPF intake across sociodemographic strata (63).

In 2008-9, a younger age was significantly associated with higher intakes of UPF (−0.16% [95%CI: −0.24 to −0.09] per year of age), but gender, occupational social class (routine and manual or intermediate vs. managerial and professional) and household status (living with other adults or living with children) were not significantly with higher intakes of UPF in adjusted models (adjusted for the aforementioned sociodemographic variables and food preparation skill/behaviours) (62). Across 2008-12, a younger age (−0.18% [95%CI: −0.21, −0.14] per year of age) and male gender (1.38% [95%CI: 0.09, 2.67]), but not occupational social class, were significantly associated with higher intakes of UPF in adjusted models (adjusted for aforementioned variables and percentage of energy intake from alcohol) (64). Across 2008-14, younger ages were significantly associated with higher intakes of UPF, with the highest intakes in 11-18-year-olds (68.0% [SE: 0.4]) then 1.5-10-year olds (63.5 [SE: 0.34]) and lowest in adults (19-64: 54.9% [SE: 0.4]) and the elderly (≥65: 53.0% [SE: 0.52]) (all p < 0.001 with 1.5-10 years as reference) (65). In 2014-16, children (65.7% [95%CI: 64.2, 67.1]) and adolescents 67.1% [95%CI: 65.7, 68.5]) consumed greater quantities of UPF than adults (54.0% [95%CI: 53.0, 55.0]) (67). Across 2008-16, a younger age, male gender, white ethnicity (vs. non-white ethnicity), lower occupational social class and living in Northern Ireland were associated with higher intakes of UPF as a proportion of TEI (66). 19-29-year-olds (59.2% [SE: 1.3]) consumed around 8% more UPF as a proportion of total energy than adults aged 60 or older (51.5% [SE: 0.5]), and 5% more than 30-59-year-olds (54% [SE: 0.4]). Males consumed 3% more UPF as a proportion of energy intake than females (55.9% [SE: 0.6] vs. 52.8% [SE: 0.4]). White ethnicity was associated with a 10% higher intake of UPF as a proportion of total energy than other ethnicities (55.4% [SE: 0.4] vs. 45.4% [SE: 1.2]). Individuals in routine and manual occupations consumed 57.3% of energy from UPF, compared with 53.4% [SE: 0.8] in intermediate occupations, 53.8% [SE: 0.7] in lower managerial and professional occupations, and 50.3% [SE: 0.7] in higher managerial and professional occupations (p-trend < 0.05). Individuals living in Northern Ireland consumed 58.7% of total energy [SE: 0.8] from UPF, compared with 51.7% [SE: 0.6] in individuals living in the South of England (including London), who had the lowest intakes of UPF. Average UPF consumption in North England, Central England/Midlands, Scotland and Wales was around 55-57% of total energy (66).

US

In the US, average UPF intake increased from 53.5% [95%CI: 52.5, 54.6] in 2001-2 to 57% [95%CI: 55.0, 58.9] in 2017-18 (69).

In 1988, individuals in the highest (5.2 to <29.8 times/day) vs. lowest quartile (0 to <2.6 times/day) of frequency of UPF intake were more likely to be younger, male, non-Hispanic white, and were less likely to be Mexican or other ethnicity, have an education below high school level or have a high income/poverty ratio (≥350% the poverty level) (68).

Across 2005 to 2018, younger age, non-Hispanic white or black ethnicity, a lower income/poverty ratio (<350% of the poverty level) were significantly associated with a higher UPF intake, or were more likely to be in the highest quantile of UPF intake (70–73,75–79). Income was not significant in one study from 2013-14 (79). Hispanics and other ethnicities including non-Hispanic Asians and non-Hispanic Asian Americans had low UPF intakes, or were less likely to be in the highest quantile of UPF intake (70–73,75,76,78,79). Non-Hispanic Asians consumed nearly 20% less unadjusted UPF (39.3% [95%CI: 38.1, 40.5]) than non-Hispanic white (57.7% [95%CI: 56.9, 58.5]) or black (60.1% [95%CI: 58.8, 61.3]) ethnicities (77). Education was also significantly associated with UPF intake, typically with higher UPF intakes with a mid-low education level (69–71,77), or mid-level education (73,75,76). One paper stratifying by ethnicity reported higher UPF intakes were seen in non-Hispanic Asian Americans with higher education levels (lowest vs. highest: 32.1% [95%CI: 29.2, 35.1] vs. 39.7% [95%CI: 38.3, 41.1]), whereas higher UPF intakes were seen in non-Hispanic white or non-Hispanic other ethnicities at lower education levels (non-Hispanic white lowest vs. highest: 61.7% 95%CI: 59.9, 63.5] vs. 53.3% [95%CI: 52.2, 54.5]; non-Hispanic other lowest vs. highest: 62.0% [95%CI: 55.9, 68.0] vs. 50.6% [95%CI: 46.7, 54.6]), and higher UPF intakes in non-Hispanic blacks at mid-low education levels (77). Similarly a higher income/poverty ratio, was associated with a higher UPF intake in non-Hispanic Asian Americans (lowest vs highest: 35.0% [95%CI: 31.9, 38.1] vs. 40.8% [95%CI: 39.2, 42.4]), but associated with a lower UPF intake in non-Hispanic whites (lowest vs highest: 61.1% [95%CI: 59.7, 62.4] vs. 55.8% [95%CI: 54.8, 56.8]) (77).

Across 2011-16, US-born individuals consumed over 12% more UPF as a proportion of total energy (adjusted for age, gender, family income/poverty ratio, education level and race/ethnicity) than foreign-born individuals (US-born: 57.9% [95%CI: 57.3, 58.5] vs. foreign-born: 45.4% [95%CI: 44.0, 46.8]) (76). This difference was seen across all sociodemographics (age, gender, income/poverty ratio, education level, ethnicity). Differences between US- and foreign-born UPF intakes were smaller at the highest income/poverty ratio and education levels (around a 10% difference compared with around a 15% difference at low and middle income/education levels), where UPF intake in foreign-born individuals tended to be higher at a higher income/poverty ratio or education level compared to lower levels, whereas UPF intake in US-born individuals tended to be lower at a higher income/poverty ratio or education level compared to lower levels (p-interaction = 0.001). US-born non-Hispanic blacks also consumed 50% more relative UPF as a proportion of total energy than foreign-born non-Hispanic blacks (60.7% [95%CI: 59.7, 61.8] vs. 40.4% [95%CI: 37.0, 43.8], p < 0.001) (76).

With marital status, unmarried, single or widowed individuals were more likely to be in the highest quartile of UPF intake, compared with being married or living with a partner (70,75). Across ethnicities, the association was present within non-Hispanic Asian American, non-Hispanic black and Hispanic ethnicities, but not in non-Hispanic white or other ethnicities (77).

Gender was not significantly associated with UPF intake in most studies (71–74,76–79); but across 2005-14, females were more likely to be in the highest quantile of UPF intake (70,75), and across 2011-16, gender proportions across quartiles of UPF intake significantly differed, with no clear relationship (75).

Adjusted mean intakes (adjusted for age, gender, education, family income/poverty ratio and race/ethnicity) from 2007-2012 show that younger ages (2-9- and 10-19-year-olds) consumed two-thirds of total energy from UPF (2-9: 65.9% [95%CI: 65.0, 66.8], 10-19: 66.8% [95%CI: 65.9, 67.7]), 13-14% more than adults aged 60 or older (52.8% [95%CI: 51.9, 53.7]). 20-39-year-olds also consumed nearly 60% of total energy from UPFs (59.5% [95%CI: 58.7, 60.3]), and 40-59-year-olds over 55% (55.2% [95%CI: 54.1, 56.4]) (71). Non-Hispanic white (60.2% [95%CI: 59.4, 60.9]) and black ethnicities (60.6% [95%CI: 59.7, 61.5] had the highest adjusted UPF intakes, 5% more than Mexican-Americans (54.8% [95%CI: 53.2, 56.3]) and 8% more than other Hispanics (52.0% [95%CI: 50.3, 53.7]), with all other races having the lowest UPF intake (49.6% [95%CI: 47.3, 51.8]), more than 10% lower than non-Hispanic white or black ethnicities. Individuals with a college level education or higher (55.9% [95%CI: 54.6, 57.2]) consumed nearly 4% less UPF as a proportion of total energy than individuals with a high school education (59.7% [95%CI: 59.1, 60.3]) or individuals with less than a high school education (59.5% [95%CI: 58.4, 60.6]) (71). Across the lowest to highest levels of family income/poverty ratios, there was a 2% difference in UPF intake as a proportion of total energy (≤1.30: 59.6% [95%CI: 58.6, 60.7] vs. >3.5: 57.7% [95%CI: 56.9, 58.6]).

From 2001 to 2018, adjusted trends in UPF intake (adjusted for age, gender, race/ethnicity, education level and income/poverty ratio) showed that males (+4.3%) and females (+2.7%) increased their UPF intake over time, to 57.2% [95%CI: 55.2, 59.1] in males and 56.8 % [95%CI: 54.6, 59.1] females in 2017-18 (p-trend = 0.001 and 0.002, respectively) (69). UPF intake increased across all ages (aged 19 or older). Older adults (60 or older) had the lowest UPF intake in 2001-2 (51.7% [95%CI: 49.4, 54.0), but the highest in 2017-18 (57.4% [95%CI: 54.3, 60.4]). UPF intake increased in non-Hispanic black or white individuals (p-trend = 0.001), but not Hispanics (p-trend = 0.081). Hispanics consistently consumed around 5% less UPF than non-Hispanic black or white individuals. Adults of all income levels increased their UPF intake from 2001 to 2018 (p-trends all < 0.05), and UPF intake increased across all education levels (p-trends all < 0.05), with the lowest intake in college graduates across time, about 5% lower than adults with lower education levels (69).

Multinational

Across 22 European countries (Austria, Belgium, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Portugal, Romania, Slovenia, Spain, Sweden, The Netherlands and the UK), gender was not significantly associated with UPF intake when expressed as a proportion of total energy, except for in Portugal (p < 0.01), where females had higher intakes than males (24.5% vs. 19.8%). UPF intakes typically varied by 1-4% between genders within each country.

Associations by sociodemographic predictor

Sociodemographic associations with UPF intake by country are presented in Table 3.

Age

Age was assessed across 17 countries. There was a consistent association of a younger age (in adults, or in adults and children) having higher UPF intakes in all countries, with some studies showing the highest intakes in adolescents. Differences in absolute UPF intake across ages were large, typically between 5-20% as a proportion of total energy, reflecting 15-100% relative differences in UPF intake. Two studies in the US (69), and Belgium (33), also reported relatively high UPF intakes in the elderly or in older adults.

Gender

Gender was assessed across 32 countries. Most of the national differences in UPF intake between genders were not significant, or varied in significance across studies within the same country (eight countries). Where significant differences were seen in Australia, Korea, Canada, Switzerland and the UK, males consumed around 1-4% more as a proportion of total energy. In Brazil, Colombia, Italy, Portugal and Spain, females consumed around 1-3% more UPF as a proportion of total energy.

Race/ethnicity

Three countries assessed race/ethnicity (Brazil, the UK and the US). Significant and large differences in UPF intake were seen across race/ethnicities, with 10-20% absolute differences in UPF intake as a proportion of total energy in UK and US, corresponding to 20-50% relative differences in UPF intake. In Brazil, the lowest quintile (≤ 13% of total energy) of UPF intake constituted 34% white and 64% African-descendent, compared with 57% white and 41% African-descendent in the highest quintile (≥ 44% of total energy) (82).

Income

Income was assessed in six countries (Australia, Brazil, Canada, Chile, Korea and the US). Five countries reported significant associations between income and UPF intake.

A higher income was associated with a higher UPF intake in Chile and Brazil, with 10-15% absolute differences in UPF consumption as a proportion of total energy across the highest and lowest income levels in Brazil, and over 4% absolute differences in Chile, reflecting a 15-100% relative increase in UPF intake with higher income.

In Australia, the second, third and fourth income quintiles had 2-3% higher adjusted absolute intakes of UPF as a proportion of total energy, compared with the lowest quintile (31). In the US, there was a 2% difference in adjusted UPF absolute intake as a proportion of total energy across income:poverty levels, increasing with lower income:poverty levels (71). The association between income and UPF intake in the US differed based on ethnicity. In Korea, there was a 4-5% crude difference in UPF intake between low and mid-to-high incomes across 2010-18, but adjusted mean intakes were non-significantly different (50). In Canada, income levels significantly differed across terciles of UPF intake in 2015, but no difference was seen in 2004.

Education level

Education level was assessed in 15 countries, with no assessment in individual studies from Colombia or the UK. There was a significant association between education level and UPF intake in 10 countries, with no significant association in Barbados, Belgium, Chile, Italy or Switzerland.

A lower education level was associated with a higher UPF intake in Australia (adjusted: 2.3% absolute difference, 6% relative difference) and Canada (adjusted: 1.8% absolute difference, 4% relative difference), and in the US after 2001. Some US studies showed higher UPF intakes with mid-low vs. high education levels (adjusted: 4% absolute difference, 7% relative difference) (71), or middle education levels (73,76). The association between education level and UPF intake in the US also differed based on ethnicity.

In the Netherlands, a middle education level (intermediate vocational education, higher secondary education) was associated with about an 8-10% greater relative UPF intake than lower or higher education levels (in grams of UPF per 2000kcal). In France, a mid-high education level had the highest UPF intakes in 2006-7, with the lowest intakes in the lowest education level (6% absolute difference, 20% relative difference). By 2014-15, middle education levels were more likely, and low education levels less likely, to be in the highest UPF intake quartile in France. Similarly in Spain, low education levels (no formal education) were less likely to be in the highest quartile of UPF intake, and mid-level (primary) education were more likely be in the highest quartile of UPF intake, with similar proportions of individuals with a high education level (secondary or higher education) across quartiles of UPF intake (60).

A higher education level was associated with a higher UPF intake in Brazil (across quartiles), Korea (adjusted: 2-3% absolute difference, 10% relative difference), Mexico (adjusted: 7.8% absolute difference, 25% relative difference) and Portugal (adjusted: 20-25% relative difference in grams), ranging from 2-8% higher as an adjusted proportion of total energy from UPF.

Socioeconomic status

Socioeconomic status, assessed via indices of occupation, social class or deprivation was assessed in six countries, with significant, but varying associations with UPF intake in all six countries.

A higher socioeconomic status was associated with higher UPF intake in Colombia (adjusted: 10% absolute, 80% relative) and Mexico (adjusted: 4.5% absolute, 15% relative).

A lower compared with higher socioeconomic status was associated with a higher UPF intake in Australia (adjusted: 2.5% absolute difference) and in the UK across 2008-2016 (based on occupational social class; 7% absolute difference, 14% relative difference) (66), but not in 2008-9 (62), or across 2008-12 (64).

Based on occupation in Italy and France, retired individuals had the lowest UPF intakes (Italy adjusted: 1.9% lower absolute intake than manual occupations, 10% lower relative intake; France in 2006: 6-13% lower absolute intake, 20-40% lower relative intake), with broadly similar intakes or higher proportions in the highest terciles of UPF intake in manual occupations, non-manual occupations, unemployed individuals or students (proportions of professional executives were similar across terciles of UPF intake in France in 2014-15).

Food insecurity

Food insecurity was assessed in three countries (Canada, France and Portugal). Higher levels of food insecurity were associated with a higher UPF intake in Canada (10% absolute difference, 20-25% relative increase in UPF across age-gender groups) and more likely to be in the highest tercile of UPF intake in France. Food insecurity was not significantly associated with UPF intake in Portugal.

Marital status

Marital status was assessed in five countries, with significant differences reported in four countries (France, Italy, the US and in Portuguese males), and tended to find unmarried, single, separated or divorced individuals had higher UPF intakes, or were more likely to be in the highest quantiles of UPF intake, compared with married individuals or individuals living together. Marital status was not significant in Korea, after adjustment for other sociodemographic factors, or in Portuguese females.

Household status

The number of individuals in the household was assessed in five countries, with a significant association with UPF intake in one country. In Spain, people living alone were less likely to be in the highest quartile of UPF intake, but household status was not significant in Switzerland, the UK, Korea or Portugal after adjustment for other sociodemographic factors or health behaviours.

Rural/urbanisation

The level of urbanisation was assessed in 11 countries, with higher UPF intakes typically being reported in more urban than rural residences in eight countries: Brazil (10% absolute difference, 80% relative difference), Colombia (6% absolute difference, 50% relative difference), Chile (6% absolute difference, 25% relative difference), France (3% absolute difference, 10% relative difference), Italy (1.6% absolute difference, 10% relative difference), Korea in 2010-2018 (0.8% absolute difference, 3% relative difference), Mexico (5.6% absolute difference, about 20% relative difference) and the Netherlands (5% relative difference). Urbanisation was not significant in Korea during 2016-2018, after adjustment for other sociodemographic factors and health behaviours.

Individuals from rural residences were more likely to be in the highest tercile of UPF intake in Canada in 2015, but there was no significant association in 2004 after adjustment for other sociodemographic factors and health behaviours. Living in inner regional Australia was crudely associated with a higher UPF intake, and living in a major city associated with a lower intake, but did not remain significant after adjustment for other sociodemographic factors. Urbanisation was also not significantly associated with UPF intake in Portugal.

Region of country

Region of country was assessed in 11 countries, with ten countries demonstrating regional differences in UPF intake, typically varying by 5-10% as a proportion of total energy, or a 25-75% relative difference in UPF intake: Belgium (adjusted: 6-8% absolute difference, 20-25% relative difference), Brazil (10% absolute difference, 75% relative difference), Chile (adjusted: 4% absolute difference, 13% relative difference), Colombia (adjusted: 9% absolute difference, 70% relative difference), Italy (adjusted: 0.1% absolute difference, 9% relative difference), Mexico (adjusted: 8% absolute difference, 33% relative difference), Portugal (20% relative difference in grams), Spain (in 2008: 5% absolute difference, 20% relative difference), Switzerland (2.4% absolute difference, 8% relative difference), UK (7% absolute difference, 14% relative difference). There was no significant difference in UPF intake across regions of France.

Immigrant status / Country of birth

Immigrant status or country of birth was assessed in four countries. Home-born vs. foreign-born individuals in the US (adjusted: 12% absolute difference, 28% relative difference), home-born and English-speaking country born individuals vs. other individuals in Australia (6-9% absolute difference, 20-30% relative difference), Swiss nationals vs. non-Swiss nationals in Switzerland (3% absolute difference, 12% relative difference) and non-immigrants vs. immigrants in Canada (11% absolute difference, 30% relative difference), had around 3-13% higher absolute intakes of UPF as a proportion of total energy, or a 10-30% higher relative UPF intake.

Indigenous identity

Indigenous identity was assessed in Canada only. Individuals with Indigenous identity were more likely to be in the highest tercile of UPF intake (44).

Associations by country-level income and adjusted analyses

When considering upper-middle income countries (Brazil, Colombia and Mexico), a higher socioeconomic score was associated with higher UPF intake in Colombia and Mexico, a higher income with higher UPF intake in Brazil, higher education level with higher UPF intake in Brazil and Colombia, and a higher UPF intake in all three countries in a more urban residence.

When considering multivariate adjusted associations only, 19 studies across 13 countries reported sociodemographic associations adjusted for other sociodemographic characteristics and health behaviours. The majority of the significant crude associations between sociodemographic variables and UPF intake remained significant with adjustment for other variables. Further details are provided in the supplementary materials.

Assessment of methodological quality

Risk of bias scores for each study are presented in table 1. Most studies scored a 5 or 7 out of 10 (average: 5.7/10), depending on whether adjustment was made for one or more sociodemographic variables. Risk of bias scores were higher (indicating a lower risk of bias) for studies performing adjustment for other sociodemographic factors.