Articles

Plant-based dietary patterns and ultra-processed food

consumption: a cross-sectional analysis of the UK Biobank
Kiara Chang,a,∗ Jennie C. Parnham,a Fernanda Rauber,b,c Renata B. Levy,b,c Inge Huybrechts,d Marc J. Gunter,d,e Christopher Millett,a,f and
Eszter P. Vamosa
a
Public Health Policy Evaluation Unit, School of Public Health, Imperial College London, London, United Kingdom
b
Department of Preventive Medicine, School of Medicine, University of São Paulo, São Paulo, Brazil
c
Center for Epidemiological Research in Nutrition and Health, University of São Paulo, São Paulo, Brazil
d
Nutrition and Metabolism Branch, International Agency for Research on Cancer, Lyon, France
e
Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom
f
NOVA National School of Public Health, Public Health Research Centre, Comprehensive Health Research Center, CHRC, NOVA
University Lisbon, Portugal

Summary eClinicalMedicine
2024;78: 102931
Background Dietary shift towards more plant-based options is increasingly popular, but the quantity of ultra-processed
foods (UPFs) they contain is largely unknown. This study assessed the level of UPF and minimally processed food Published Online xxx
https://doi.org/10.
consumption among regular and low red meat eaters, flexitarians, pescatarians, vegetarians and vegans in a large
1016/j.eclinm.2024.
dataset of United Kingdom (UK) adults. 102931

Methods This is a cross-sectional analysis of the UK Biobank participants recruited between December 19, 2006, and
October 1, 2010. Responses to food frequency questions were used to identify diet types for vegans (never eating any
animal-based foods); vegetarians (never eating meat/fish); pescatarians (never eating meat); flexitarians (consumed
fish/meat under twice a week); low red meat eaters (consumed fish/poultry more than once a week but red/
processed meat under twice a week); and regular red meat eaters (consumed red/processed meat more than once
a week). Consumption of all food and drinks collected in 24-h recalls between April 29, 2009, and June 28, 2012,
were categorised using the Nova classification. The primary outcomes are the consumption of UPFs and
minimally processed foods, expressed as a percentage of daily food intake (grams/day). Multivariable linear
regression assessed the mean percentage point difference in UPF and minimally processed food consumption
between diet types.

Findings This study included 199,502 UK Biobank participants (mean age 58.2 [standard deviation 7.9] years; 55.1%
women). The mean UPF consumption was 24.2%, 21.9%, 22.0%, 20.4%, 23.8%, and 22.7% among 75,091 regular red
meat eaters, 70,144 low red meat eaters, 45,057 flexitarians, 4932 pescatarians, 4119 vegetarians and 159 vegans,
respectively. The adjusted results suggested that compared with regular red meat eaters, UPF consumption was 1.3
percentage points higher among vegetarians (95% confidence interval [CI]: 0.9, 1.7) and lower among low red meat
eaters (−1.3, 95% CI: −1.4, −1.1), flexitarians (−0.8, 95% CI: −1.0, −0.7), and pescatarians (−1.6, 95% CI: −1.9, −1.2).
The UPF consumption in vegans were not significantly different from regular red meat eaters (1.2 percentage points,
95% CI: −0.7, 3.2). Minimally processed food consumption was higher in all other types of diet than regular red meat
eaters, with an adjusted percentage point difference ranged from 0.4 (95% CI: 0.005, 0.9) for vegetarians to 3.2 (95%
CI: 1.0, 5.5) for vegans compared with regular red meat eaters.

Interpretation This UK-based study found higher UPF consumption in vegetarian diets and lower in diets with a
modest amount of meat or fish. It is important that policies which encourage the urgently needed transition to
more sustainable dietary patterns also promote rebalancing diets towards minimally processed foods.

Funding National Institute for Health and Care Research (NIHR) School for Public Health Research, World Cancer
Research Fund.

Copyright © 2024 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

*Corresponding author. 609 School of Public Health Building, 90 Wood Lane, London W12 0BZ, United Kingdom.
E-mail address: chu-mei.chang@imperial.ac.uk (K. Chang).

www.thelancet.com Vol 78 December, 2024 1

 Articles

Keywords: Ultra-processed foods; Plant-based diet; Flexitarian; Vegetarian; Vegan

Research in context

Evidence before this study Added value of this study

We searched PubMed from the inception of the database to This is the first UK study and the largest study conducted to
May 6, 2024, for studies published that have assessed ultra- the best of our knowledge. We found using data from the UK
processed food (UPF) consumption in two or more diet types Biobank study that the consumption of UPFs was high and
with different levels of animal-sourced food avoidance (e.g., represented more than 20% of daily food intake and more
vegetarians, meat eaters). No language restrictions were than 46% of daily energy intake in all types of diet. Compared
applied. Only two studies were identified and these were with diets of regular red meat eaters, vegetarians consumed a
smaller in sample size compared with our study. No previous significantly higher amount of UPFs while diets of low red
study has been conducted on a United Kingdom (UK)-based meat eaters, flexitarians, and pescatarians showed the
sample despite the UK being one of the leading consumers of opposite.
UPFs globally. Given the environmental impacts of animal-
Implications of all the available evidence
sourced foods, it is important to understand how the
The consumption of UPFs is universally high across all types
consumption of UPFs and minimally processed foods varies
of dietary patterns based on animal-sourced food avoidance.
across dietary patterns of animal-sourced food consumption.
The adaptation of plant-based meat and dairy alternatives is
evident particularly in the diets of pescatarians, vegetarians,
and vegans.

Introduction or no whole foods, and are higher in calorie, salt, fat,

The global temperature has risen to 1.3 ◦ C above pre- and sugar content but lower in fibre.8 Most industrially
industrial levels and fast approaching the 1.5 ◦ C target produced meat and dairy substitutes are UPFs, as are
set in the Paris Agreement.1 Dietary shift towards more soft drinks, mass-produced bakery goods and most
plant based options is increasingly common partly due breakfast cereals. They are made hyper-palatable, rela-
to considerations for planetary health, as meat and dairy tively cheap and convenient, heavily marketed and
are widely understood as the biggest contributors to purposefully designed to displace minimally processed
greenhouse gas emissions from individual diets.2,3 In foods in diets.8
the United Kingdom (UK), representative surveys The consumption of UPFs is rising globally and has
showed that 4% and 1% of the respondents identified as reached beyond 50% of daily calorie intake in the UK
vegetarian and vegan respectively, and another 10% and United States (US).9,10 The causal link between UPF
indicated being mainly vegetarian but occasional meat consumption and weight gain was demonstrated in a
consumer.4 Market research data for Western Europe randomised controlled trial, and large-scale cohort
suggested that between 2007 and 2022, the per capita studies have generally suggested an association between
sales volume of milk and milk drinks reduced by 3.9 kg higher UPF consumption and increased risk of car-
while plant-based milk increased by 1.2 kg.5 Notably, the diometabolic diseases, cancer, and mortality.11–15 A
sales volume of meat substitutes grew by 10.1% from recent prospective cohort study from the UK has shown
2021 to 2022 while processed meat sales declined by that diets high in plant-sourced UPFs were linked to
2.5%.5 These data are supportive of a concurrent shift increased cardiovascular risk.16 However, data are scarce
towards adaptation of a more plant-based diet, and raise on the level of UPFs and minimally processed foods
concerns about their quality and healthfulness in light consumed by people adhering to various dietary pat-
of the increased traction of plant-sourced but highly terns according to differing levels of animal-sourced
industrially processed products.6 food avoidance. This is an important gap in knowl-
The United Nations has recommended a sustain- edge as the UPF industry is well-aware of the increasing
able healthy diet consisting of a diverse range of demand of plant-sourced foods, and the production and
minimally processed foods while avoiding ultra- marketing of plant-based alternatives (including meat,
processed foods (UPFs).7 Based on the Nova classifi- seafood, eggs, and dairy) have greatly increased in the
cation, UPFs are industrial products made through a past decade.17 Therefore, this study aims to quantify and
sequence of extensive industrial processes that include compare the degrees of industrial food processing in
fractioning of original foods into substances and often dietary patterns of regular red meat eaters, low red meat
chemically modifying them, then recombining the eaters, flexitarians, pescatarians, vegetarians and vegans
substances into products during which various addi- in a large sample of middle-aged adults who participated
tives are frequently used.8 UPFs typically contain little in the UK Biobank.

2 www.thelancet.com Vol 78 December, 2024

 Articles

Methods same 24-h recall was sent out to all participants with a
Data source and study population valid email address (about 320,000 participants) on four
The UK Biobank is a large prospective cohort study that separate occasions during 2011–2012. The dietary recall
recruited half a million participants aged between 40 was designed to capture the consumption of over 200
and 69 years old identified from the National Health food and beverage items in the previous 24 h, and has
Service patient registry.18 Study participants attended been validated showing similar energy and nutrient in-
one of the 22 assessment centres across England, Wales, takes as an interviewer-administered 24-h recall.19 We
and Scotland for the collection of baseline data between considered a total of 210,975 participants with one or
December 19, 2006 and October 1, 2010. They more 24-h recall data collected between April 29, 2009
completed questionnaires covering socio-demographic, and June 28, 2012 for inclusion in this study (Fig. 1).
lifestyle (including smoking status and a 29-item short
food frequency questionnaire [FFQ]) and psychosocial Outcome measures
characteristics at recruitment.18 Physical measurements Participants’ 24-h dietary recall data were used to
were obtained using standardised procedures, and construct the dietary outcome measures. We applied the
medical history and medication use were verified by Nova classification assigning each food and beverage
trained research staff. item collected by the 24-h recall to one of the four Nova
In a subsequent addendum of the UK Biobank pro- food groups based on the extent and purpose of indus-
tocol, a web-based 24-h dietary recall (Oxford WebQ) trial food processing they underwent.8 Nova 1 includes
was introduced in 2009 and completed by the last 70,000 unprocessed and minimally processed foods that are
study participants attending baseline assessment. The edible parts of food as found in nature or have

Fig. 1: Flowchart for the derivation of study population from the UK Biobank.

www.thelancet.com Vol 78 December, 2024 3

 Articles

undergone minimal processing without adding any oily/non-oily fish but never consumed red meat, pro-
substances (e.g., salt, sugar) to the original food. Ex- cessed meat, or poultry. Flexitarians were defined as
amples of Nova 1 foods are fruits, vegetables, legumes, participants reported eating fish, red meat, processed
roots and tubers, milk and meat. Nova 2 includes pro- meat, or poultry no more than once a week. The
cessed culinary ingredients such as sugar, vegetable oils remaining participants were classified as meat eaters
and butter. Nova 3 represents processed foods that are and were further divided into two categories. Low red
manufactured by combining Nova 1 food with Nova 2 meat eaters were defined as those reported eating red/
ingredients, e.g., canned vegetables in brine, freshly processed meat no more than once a week (but
made breads and cheeses. Nova 4 represents UPFs that consumed fish or poultry more than once a week).
are industrial formulations made often with many Regular red meat eaters were frequent consumers re-
substances that are exclusively available for industrial ported eating red/processed meat more than once a
use (e.g., modified starches, protein isolates, high- week.
fructose corn syrup, emulsifiers, thickening agents) We cross-compared participants’ diet type with their
and undergo extensive food processing procedures that dietary intake of similar food groups (meat, fish, eggs,
cannot be conducted in a domestic kitchen. Examples of and dairy) as recorded in the 24-h recalls, and those with
Nova 4 foods are soft drinks, mass-produced industrial- inconsistencies in data reporting between the FFQ and
processed breads, sweet or savoury packaged snacks, 24-h recalls were excluded from the study. We addi-
breakfast ‘cereals’, reconstituted meat products and tionally presented the mean daily intake of the following
ready-to-eat/heat food. The methodological details have key nutrients as part of the characteristics for the overall
been previously described.12,13 sample and the six diet types: total energy intake (kcal/
We averaged dietary intakes across multiple 24-h day), free sugar intake (as a percentage of total energy
recalls. We assessed for primary outcomes the partici- intake), saturated fatty acids intake (as a percentage of
pants’ consumption of foods from each Nova group total energy intake), sodium density (mg per 1000 kcal
measured as a proportion of total daily food/energy intake), and fibre density (g per 1000 kcal intake).
consumed (%g/day and %kcal/day, respectively). The
secondary outcomes were the participants’ absolute Covariates
daily intake of food/energy from each Nova group (g/ Study covariates included age, sex (male, female),
day and kcal/day, respectively). We combined the dietary ethnicity (white, non-white), smoking status (never
intakes of Nova 1 and 2 foods in one measurement as smoked, ex-smoker, current smoker), physical activity
they are mostly consumed in combination, and refer to level using the International Physical Activity Ques-
‘minimally processed foods’ from here onwards as they tionnaire (low, moderate, high, missing),21 body mass
represent the largest quantity in diet. index (BMI) categorised as underweight (<18.5 kg/m2),
normal (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/
Diet types of the UK Biobank participants m2), or obese (≥30.0 kg/m2), highest educational
Study participants were categorised as regular or low red attainment (university degree, other), average annual
meat eater, flexitarian, pescatarian, vegetarian, or vegan household income (<£18,000, £18,000-£30,999,
based on their responses to selected FFQ questions £31,000-£51,999, >£52,000, missing), Index of Multiple
relevant for the classification of these diet types. Re- Deprivation (IMD) quintile, and total energy intake
sponses to seven questions were used that queried (kcal/day, included when the outcome being assessed
participants’ frequency of consuming oily fish, non-oily was based on food weight). IMD is a composite rank
fish, processed meat, poultry, beef, lamb, and pork over score for each small area in the UK derived based on
the past year.20 Participants were given the following seven domains of deprivation,22 and was mapped to
options to select from: never, less than once a week, participants’ postcode in the UK Biobank. We cat-
once a week, 2–4 times a week, 5–6 times a week, once egorised participants’ IMD scores into quintiles. Addi-
or more daily, do not know, or prefer not to answer. A tional covariates considered in the sensitivity analysis
further question assessed participants’ avoidance of included self-reported diagnosis of diabetes (or regular
dairy products and eggs or foods containing eggs. Par- insulin use), high blood pressure (or regularly taking
ticipants who did not respond to all eight FFQ questions blood pressure medication), depression, or history of
and those responded ‘do not know’ or ‘prefer not to cardiovascular disease (angina, heart attack, and stroke)
answer’ to one of these questions were considered as verified in a nurse interview. Missing data were un-
ambiguous for the purpose of diet type classification and der 3.0% of the study population for ethnicity, smoking
were excluded from the study. status, BMI, IMD and education. These were assumed
Participants were classified as vegans if they reported missing at random and were subsequently excluded
never eating any type of meat, fish, eggs or dairy prod- taking a complete case analysis approach. However,
ucts. Vegetarians included lacto- and ovo-vegetarians participants with missing physical activity or average
who reported never eating any type of meat or fish. household income were considerable (14.8% and 9.1%
Pescatarians were identified as those reported eating of the study population, respectively) and there may be

4 www.thelancet.com Vol 78 December, 2024

 Articles

common factors related to their missingness, such as Analyses were conducted using Stata version 15.
participants’ willingness to provide data. Therefore, a Statistical tests were two-sided with a p-value < 0.05
missing data category was employed to capture this and considered significant.
preserve sample size.
Role of funding source
Ethics The funders had no role in study design, data collection,
The UK Biobank received ethical approval from the data analysis, interpretation, or writing of the report.
North West Multi-centre Research Ethics Committee
(21/NW/0157), and all study participants provided
written informed consent at recruitment.23 Results
This study included 199,502 UK Biobank participants
Statistics after excluding 176 individuals who reported being
We compared characteristics of the study population pregnant or were unsure, 706 individuals with an
and dietary intake of key nutrients by diet type using implausible total energy intake beyond 500–5000 kcal/
analysis of variance for continuous and chi-squared day,12,13 2110 individuals with incomplete FFQ data,
tests for categorical variables. We examined the dis- 1110 individuals with an inconsistent reporting of major
tribution for each continuous variable graphically as food groups consumed between the FFQ and 24-h re-
well as assessing their skewness and kurtosis before an calls data, and 7371 individuals with missing covariates
analysis of variance test was performed. We also tested data (Fig. 1). The mean age was 58.2 (SD = 7.9) years,
using the Kruskal Wallis test which do not impose a 55.1% were women, and 96.0% were of white ethnicity
normality assumption on the distribution and the re- (Table 1). Most participants (72.8%) were frequent meat
sults were the same. The mean proportion of daily eaters, one in five were flexitarians (22.6%), and much
food/energy intake contributing to the Nova food fewer were pescatarian (2.5%), vegetarian (2.1%), or
groups (primary outcomes) and their subsidiary food vegan (0.08%). Pescatarians, vegetarians, and vegans
groups were computed separately and presented were more likely to be women and younger, had a non-
graphically for regular red meat eaters, low red meat white ethnic background, high physical activity level,
eaters, flexitarians, pescatarians, vegetarians, and BMI in the normal range, completed university educa-
vegans identified. The non-parametric Wilcoxon rank tion, and reside in a more deprived neighbourhood. On
sum tests were used to compare the distribution of average, regular red meat eaters had the highest total
primary outcomes and its subsidiary food groups be- energy and lowest fibre density while the opposite was
tween regular red meat eaters as the reference group found for vegans. Moreover, saturated fatty acids intake
and participants of other diet type. The non-parametric was highest among regular red meat eaters and flex-
tests were adopted because the normality assumption itarians, and lowest among vegans. Vegetarians had the
was less likely to be met for the subsidiary food groups highest sodium density while vegans had the lowest.
with smaller consumption relative to the overall diet. The mean free sugar intake was similarly high among
Multivariable linear regression models were per- regular red meat eaters and vegetarians (13.9% of total
formed to assess whether the mean proportion of daily energy intake), and was lowest among low red meat
food/energy intake contributed by each Nova food group eaters (13.1% of total energy intake).
differed between diet types while adjusted for age, sex,
ethnicity, smoking status, physical activity, BMI cate- Dietary intakes based on the percentage of total
gory, highest education attainment, average household food (g/day) consumed
income, IMD quintile, and total daily energy intake (for Overall, there were no clear patterns observed across
the food weight models). Similar regression models diet types for the contribution of minimally processed
were performed for the mean absolute daily food/energy foods (defined earlier as Nova 1 and Nova 2 combined)
intake contributed by each Nova food groups (secondary and Nova 4 UPFs in daily food consumed (Fig. 2a,
outcomes). Fig. 3). However, the mean proportion of Nova 3 pro-
Sensitivity analyses were performed on primary cessed food consumption was the highest among reg-
outcomes: (i) excluding the BMI variable from the ular red meat eaters (12.3%) and lower among those
models because an examination of the variance inflation with plant-based dietary patterns especially among
factor (VIF) suggested that BMI may be an influential vegans (5.6%) (Fig. 2a).
factor susceptible to multicollinearity (with a VIF>30.0 The mean proportion of minimally processed food
while all other variables had VIF<2.6); (ii) additionally consumption was the lowest among regular red meat
adjusted for the presence of self-reported diagnosis of eaters (63.5%) and highest among vegans (71.7%). The
diabetes (or regular insulin use), high blood pressure (or mean UPF consumption was the lowest among pesca-
regularly taking blood pressure medication), depression, tarians (20.4%) and highest among regular red meat
or history of cardiovascular disease; (iii) excluding par- eaters (24.2%). Compared with regular red meat eaters,
ticipants with fewer than two 24-h recalls. vegans consumed a moderately larger amount of Nova 1

www.thelancet.com Vol 78 December, 2024 5

 Articles

Overall Regular red Low red Flexitarian Pescatarian Vegetarian Vegan p-valuea
meat eater meat eater
N (%) 199,502 (100.0) 75,091 (37.6%) 70,144 (35.2%) 45,057 (22.6%) 4932 (2.5%) 4119 (2.1%) 159 (0.08%)
Age (year)
Mean (SD) 58.2 (7.9) 58.1 (8.1) 58.4 (7.9) 58.8 (7.7) 56.1 (7.8) 54.8 (7.7) 53.8 (8.0) <0.001
Sex (n, %)
Male 89,532 (44.8%) 43,765 (58.2%) 25,329 (33.7%) 17,645 (36.1%) 1362 (27.6%) 1367 (33.1%) 64 (40.2%) <0.001
Female 109,970 (55.1%) 31,326 (41.7%) 44,815 (59.6%) 27,412 (63.8%) 3570 (72.3%) 2752 (66.8%) 95 (59.7%)
Ethnicity (n, %)
White 191,614 (96.0%) 72,514 (96.5%) 67,217 (89.5%) 43,421 (95.8%) 4697 (95.2%) 3617 (87.8%) 148 (93.0%) <0.001
Non-white 7888 (3.9%) 2577 (3.4%) 2927 (3.8%) 1636 (4.1%) 235 (4.7%) 502 (12.1%) 11 (6.9%)
Smoking status (n, %)
Never smoked 112,991 (56.6%) 41,332 (55.0%) 40,494 (53.9%) 25,698 (57.7%) 2828 (57.3%) 2557 (62.0%) 82 (51.5%) <0.001
Ex-smoker 71,037 (35.6%) 26,873 (35.7%) 25,052 (33.3%) 15,913 (35.7%) 1814 (36.7%) 1323 (32.1%) 62 (38.9%)
Current smoker 15,474 (7.7%) 6886 (9.1%) 4598 (6.1%) 3446 (6.5%) 290 (5.8%) 239 (5.8%) 15 (9.4%)
Physical activity (n, %)
High 66,749 (33.4%) 23,872 (31.7%) 25,085 (33.4%) 14,156 (35.7%) 2010 (40.7%) 1559 (37.8%) 67 (42.1%) <0.001
Moderate 71,966 (36.0%) 27,351 (36.4%) 24,676 (32.8%) 16,692 (35.1%) 1724 (34.9%) 1462 (35.4%) 61 (38.3%)
Low 31,210 (15.6%) 12,754 (16.9%) 10,034 (13.3%) 7292 (14.3%) 545 (11.0%) 570 (13.8%) 15 (9.4%)
Missing 29,577 (14.8%) 11,114 (14.8%) 10,349 (13.7%) 6917 (14.7%) 653 (13.2%) 528 (12.8%) 16 (10.0%)
Baseline BMI status (n, %)
Underweight (<18.5 kg/m2) 1064 (0.5%) 282 (0.3%) 365 (0.4%) 268 (0.5%) 76 (1.5%) 70 (1.6%) 3 (1.8%) <0.001
Normal (18.5–24.9 kg/m2) 73,437 (36.8%) 23,201 (30.8%) 26,218 (34.9%) 18,981 (37.3%) 2766 (56.0%) 2175 (52.8%) 96 (60.3%)
Overweight (25–29.9 kg/m2) 82,966 (41.5%) 32,779 (43.6%) 29,120 (38.7%) 18,081 (41.5%) 1572 (31.8%) 1364 (33.1%) 50 (31.4%)
Obese (≥30 kg/m2) 42,035 (21.0%) 18,829 (25.0%) 14,441 (19.2%) 7727 (20.5%) 518 (10.5%) 510 (12.3%) 10 (6.2%)
Highest educational
attainment (n, %)
College/University degree 86,824 (43.5%) 31,090 (41.4%) 29,348 (39.0%) 20,732 (41.8%) 3075 (62.3%) 2483 (60.2%) 96 (60.3%) <0.001
Other 112,678 (56.4%) 44,001 (58.5%) 40,796 (54.3%) 24,325 (58.1%) 1857 (37.6%) 1636 (39.7%) 63 (39.6%)
Average household income (n, %)
>£52,000 57,361 (28.7%) 21,701 (28.8%) 20,514 (27.3%) 12,359 (29.2%) 1551 (31.4%) 1206 (29.2%) 30 (18.8%) <0.001
£31,000–£51,999 51,920 (26.0%) 19,957 (26.5%) 17,913 (23.8%) 11,568 (25.5%) 1324 (26.8%) 1106 (26.8%) 52 (32.7%)
£18,000–£30,999 44,219 (22.1%) 16,564 (22.0%) 15,422 (20.5%) 10,321 (21.9%) 1006 (20.3%) 870 (21.1%) 36 (22.6%)
<£18,000 27,789 (13.9%) 10,396 (13.8%) 9593 (12.7%) 6511 (13.6%) 685 (13.8%) 575 (13.9%) 29 (18.2%)
Missing 18,216 (9.1%) 6473 (8.6%) 6702 (8.9%) 4298 (9.5%) 366 (7.4%) 362 (8.7%) 12 (7.5%)
Index of multiple deprivation (n, %)
Quintile 1 (Least deprived) 44,711 (22.4%) 16,808 (22.3%) 15,782 (21.0%) 10,400 (22.4%) 988 (20.0%) 713 (17.3%) 20 (12.5%) <0.001
Quintile 2 44,436 (22.2%) 16,460 (21.9%) 15,988 (21.2%) 10,177 (22.7%) 1011 (20.4%) 778 (18.8%) 22 (13.8%)
Quintile 3 41,830 (20.9%) 15,685 (20.8%) 14,857 (19.7%) 9319 (21.1%) 1044 (21.1%) 890 (21.6%) 35 (22.0%)
Quintile 4 38,467 (19.2%) 14,331 (19.0%) 13,402 (17.8%) 8763 (19.1%) 997 (20.2%) 942 (22.8%) 32 (20.1%)
Quintile 5 (Most deprived) 30,058 (15.0%) 11,807 (15.7%) 10,115 (13.4%) 6398 (14.4%) 892 (18.0%) 796 (19.3%) 50 (31.4%)
Total energy intake (kcal/day)
Mean (SD) 2044.8 (611.7) 2166.3 (632.3) 1971.7 (584.0) 1973.9 (585.6) 1963.3 (597.5) 1957.7 (633.3) 1845.0 (621.6) <0.001
Free sugar intake
(% of total energy intake)
Mean (SD) 13.6 (7.3) 13.9 (7.2) 13.1 (7.2) 13.7 (7.3) 13.2 (7.5) 13.9 (7.5) 13.3 (8.5) <0.001
Saturated fatty acids intake
(% of total energy intake)
Mean (SD) 10.8 (3.4) 11.1 (3.4) 10.3 (3.3) 11.1 (3.4) 10.6 (3.5) 10.8 (3.7) 6.7 (2.4) <0.001
Sodium density (mg/1000 kcal)
Mean (SD) 932.8 (258.6) 957.3 (258.7) 910.2 (257.4) 922.2 (254.7) 939.5 (266.1) 979.1 (262.4) 882.0 (296.1) <0.001
Fibre density (g/1000 kcal)
Mean (SD) 12.6 (5.0) 11.8 (4.8) 13.0 (5.0) 12.6 (5.0) 15.0 (5.2) 16.0 (5.7) 20.9 (6.6) <0.001

SD, standard deviation; N, sample size. ap-values were from the comparison of characteristics between diet types were assessed by analysis of variance for continuous variables and chi-squared tests for
other categorical variables.

Table 1: Characteristics of the study population from the UK Biobank by diet types.

6 www.thelancet.com Vol 78 December, 2024

 Articles

Fig. 2: Mean proportion of the total diet from each Nova food group as measured by: a) daily food intake; and b) daily energy intake.
Nova 1 includes unprocessed and minimally processed foods, Nova 2 includes processed culinary ingredients, Nova 3 represents processed foods,
Nova 4 represents ultra-processed foods. †p < 0.05; ‡p < 0.01 from ranksum test comparing distribution of consumption against regular red
meat eaters.

fruit, vegetables, legumes and nuts (Appendix Table S1). of minimally processed foods in their diet (Fig. 4). In
Several UPFs, including industrially processed breads, particular, vegans consumed 3.2 percentage points
carbonated drinks, and ready meals, were commonly (95% confidence interval [CI]: 1.0, 5.5) higher amount
consumed by all diet types. However, the consumption of minimally processed foods than regular red meat
of Nova 4 plant-based milk and meat alternatives were eaters. The consumption of UPFs showed a mixed
substantially higher among pescatarians, vegetarians pattern. Compared with regular red meat eaters, the
and vegans (e.g., 4.9% vs 0.4% from plant-based milk, mean UPF consumption were higher among vegetar-
and 1.5% vs 0.03% from meat alternatives among ians by 1.3 percentage points (95% CI: 0.9, 1.7) but
vegans vs regular red meat eaters, respectively). lower among low red meat eaters, flexitarians, and
Results of fully adjusted regression models showed pescatarians. There was no evidence of a difference in
that compared with regular red meat eaters, all other UPF consumption between vegans and regular red
diet types had a significantly larger mean consumption meat eaters.

www.thelancet.com Vol 78 December, 2024 7

 Articles

Fig. 3: Mean proportion of daily food intake from subsidiary food groups of: a) Nova 1; b) Nova 2 and Nova 3; and c) Nova 4 food group.
Nova 1 includes unprocessed and minimally processed foods, Nova 2 includes processed culinary ingredients, Nova 3 represents processed foods,
Nova 4 represents ultra-processed foods.

Dietary intakes based on the percentage of total were significantly lower among low red meat eaters,
energy (kcal/day) consumed flexitarians, and pescatarians compared with regular red
Based on the total daily energy consumed, the mean meat eaters. However, UPF consumption was signifi-
dietary contribution of minimally processed foods cantly higher among vegetarians by 2.3 percentage
ranged from 36.1% in regular red meat eaters to 42.2% points (95% CI: 1.8, 2.8) while no evidence of a differ-
in vegans (Fig. 2b, Appendix Fig. S1 and Table S2). The ence was observed for vegans compared with regular red
mean dietary energy from UPFs were the lowest in meat eaters.
pescatarians (46.4%) and highest in vegetarians (51.7%).
Results from the fully adjusted regression models Consumption of Nova food groups measured as the
suggest that compared with regular red meat eaters, all amount in grams/calories consumed
other diet types had a significantly higher dietary energy The analyses of absolute food intake showed closely
sourced from minimally processed foods except for consistent results with that of the relative measures
vegetarians (Fig. 5). Dietary energy sourced from UPFs presented above except for the consumption of UPFs in

8 www.thelancet.com Vol 78 December, 2024

 Articles

Fig. 4: Mean percentage points difference between diet types for the consumption of each Nova food group as measured by daily food
intake. Abbreviations: Coeff, coefficient; CI, confidence interval. Nova 1 includes unprocessed and minimally processed foods, Nova 2 includes
processed culinary ingredients, Nova 3 represents processed foods, Nova 4 represents ultra-processed foods. All linear regression models were
fully adjusted for age, sex, ethnicity, smoking status, physical activity, body mass index category, highest education attainment, average
household income, Index of Multiple Derivation quintile, and total daily energy intake.

vegans that was found significantly higher on average by minimally processed food consumption according to
81.9 g/day (95% CI: 33.4, 130.4) compared with regular increasing levels of animal-sourced food avoidance.
red meat eaters (Appendix Table S3). The absolute daily Consumption of UPFs was high in all diet types and
energy intake from minimally processed foods were represented more than 20% of daily food intake and
significantly higher among low red meat eaters but more than 46% of daily energy intake among study
lower among flexitarians, pescatarians, and vegetarians participants. Vegans consumed on average 3.2 percent-
compared with regular red meat eaters. Moreover, the age points higher amount of minimally processed foods
absolute daily energy sourced from UPFs were found in their diets, but their UPF consumption was not found
significantly lower among all diet types compared with significantly different from those of regular red meat
regular red meat eaters. eaters. Furthermore, vegetarians consumed on average
1.3 percentage points higher amount of UPFs in their
Sensitivity analysis diet while low red meat eaters, flexitarians, and pesca-
The results of sensitivity analysis showed largely tarians consumed more than 0.8 percentage points
consistent findings particularly for the consumption of lower amount of UPFs compared with regular red meat
minimally processed foods and UPFs (Appendix eaters.
Tables S4 and S5). There was an exception where ana- Only two previous smaller studies assessed associa-
lyses were restricted to participants with >1 dietary re- tions of diet types and UPF consumption.24,25 One study
calls (Appendix Table S4), where vegans were found analysed consumption of a few selected UPFs in a
consuming a similar proportion of minimally processed German sample and found vegetarians were more likely
foods but a significantly higher proportion of UPFs than to consume plant-based meat alternatives but less likely
regular red meat eaters. to consume fast food, sweet and salty snacks, and ultra-
processed beverages compared with high meat eaters.25
The other study of a French sample analysed data
Discussion collected using the same dietary assessment method as
This large cross-sectional analysis of the UK Biobank used in our study (24-h recall).24 The authors found
study found no clear patterns for either UPF or higher UPF consumption in the diets of vegetarians and

www.thelancet.com Vol 78 December, 2024 9

 Articles

Fig. 5: Mean percentage points difference between diet types for the consumption of each Nova food group as measured by daily
energy intake. Abbreviations: Coeff, coefficient; CI, confidence interval. Nova 1 includes unprocessed and minimally processed foods, Nova 2
includes processed culinary ingredients, Nova 3 represents processed foods, Nova 4 represents ultra-processed foods. All linear regression
models were fully adjusted for age, sex, ethnicity, smoking status, physical activity, body mass index category, highest education attainment,
average household income, and Index of Multiple Derivation quintile.

vegans than meat eaters (37.0%, 39.5%, and 33.0% of Health Organization, and may be reflective of the high
daily energy intake, respectively).24 By contrast, our UPF consumption among all diet types.26,27
findings did not show a consistent pattern of higher Dietary patterns represent a key opportunity for
UPF consumption with greater levels of animal-sourced intervention on environmental sustainability given that
food avoidance. Our study found that compared with the food system contributes to one third of global
regular red meat eaters, low red meat eaters, flexitarians, greenhouse gas emissions and consumes substantial
and pescatarians ate a significantly lower amount of natural resources.7 The planetary and human health co-
UPFs while vegetarians had a significantly higher UPF benefits of a balanced diet based on fruit, vegetables,
consumption. This was consistent across proportional whole grains, legumes, and nuts are well-understood.28
consumption as measured by total food weight and total However, plant-based diets vary in dietary quality
energy intake. when food ultra-processing is considered. UPFs are
Although lower nutritional quality has been linked to manufactured predominantly using cheap ingredients
ultra-processed dietary patterns,8 our study has shown derived from a few high-yielding plants (e.g., corn,
higher fibre density and lower total energy, saturated wheat, rice) or selected animal species fed on these
fatty acids and sodium density among vegans. This may crops in the case of animal-sourced UPFs.8,29 Any
possibly be explained by the avoidance of meat and further increase in UPF consumption could worsen the
dairy, and higher consumptions of fruits, vegetables, already imbalanced food system and may accelerate
legumes and nuts compared with other diet types. biodiversity loss.7,29
However, mean free sugar intake was universally high Plant-based and low meat diets are often perceived as
and above 13% of total energy intake for all diet types healthy and environmentally sustainable, and the
regardless of their level of animal-sourced food avoid- increasing per capita consumption of meat and dairy
ance. This is more than double the maximum daily alternatives in the UK and globally is reflective of the
intake recommended in the UK and by the World growing demand of plant-based alternatives.5 The UPF

10 www.thelancet.com Vol 78 December, 2024

 Articles

industry has responded to these demands as evident reflect the increasing availability and sales of plant-
from the increased production and marketing of new sourced UPF products on the market.5 Sixth, the self-
ranges of plant-based UPF products every year.17 How- completed FFQ and 24-h recall data could be subject
ever, novel evidence from literature has suggested that to potential mis-reporting, social desirability and recall
the degree and purpose of food processing might be bias. Finally, owing to the observational nature of the
important when considering the healthfulness of plant- study, any residual confounding may bias the findings
based foods.6,16 In addition to evidence from well- although a wide range of socio-demographic and life-
conducted prospective cohort studies that has shown style factors were adjusted for.
links between UPF consumption and obesity, type 2 This large UK-based study found similarly high UPF
diabetes, and various negative health outcomes, consumption among diet types regardless of their levels
emerging research has demonstrated that plant-sourced of animal-sourced food avoidance. Policies that
UPFs were detrimental for cardiovascular health.11–16 In encourage the urgently needed transition to more sus-
this study, we found that UPF consumption were uni- tainable and plant-based dietary patterns should pro-
versally high in the diets of UK Biobank participants, mote the consumption of minimally processed foods,
and many UPFs, including mass-produced bakery and rebalancing diets away from UPFs.
goods, carbonated drinks, and ready-to-eat foods, were
Contributors
commonly consumed by all diet types. Furthermore, our
KC, JCP, and EPV conceptualised the study. KC compiled the data and
findings highlight that pescatarians, vegetarians, and performed statistical analyses. KC and EPV had access and verified the
vegans were more likely to include plant-based milk and underlying data used in this study. All authors contributed to the
meat alternatives in their diet. This is concerning as finalisation of statistical models and interpretation of findings. KC and
UPFs produced purely from plant-derived substances EPV wrote the first draft of the manuscript, and JCP, FR, RBL, IH, MJG,
and CM critically reviewed and edited the manuscript. All authors had
are increasingly promoted by the UPF industry as full access to all the data in the study, approved the final manuscript, and
healthy and sustainable alternatives to mobilise con- accept responsibility for the decision to submit for publication.
sumers’ transition away from meat-based diets. It is,
therefore, important that urgently needed policies that Data sharing statement
UK Biobank data are available through application to the database
address food system sustainability also promote reba-
https://www.ukbiobank.ac.uk/.
lancing diets towards minimally processed foods away
from UPFs. Declaration of interests
Our study has many strengths. We present a All authors declare no conflict of interest.
comprehensive analysis of dietary consumption based
on both weight and energy content of all foods and Acknowledgements
This study is supported by the National Institute for Health and Care
drinks consumed. The large sample size enables com-
Research (NIHR) School for Public Health Research (SPHR) (Grant
parison between six dietary patterns, including by the Reference Number NIHR 204000). The views expressed are those of the
type and level of animal-sourced food avoidance. The authors and not necessarily those of the NIHR or the Department of
use of both FFQ and 24-h recall data facilitates more Health and Social Care. Funding for grant IIG_FULL_2020_033 was
accurate categorization of diet types. The validated 24-h obtained from World Cancer Research Fund (WCRF UK), as part of the
World Cancer Research Fund International grant programme. CM is
recalls allowed for more detailed dietary consumption supported by the NIHR Global Health Research Centre on NCDs and
data being captured than many FFQs. Environmental Change (NIHR203247). The funders had no role in
There are important limitations to acknowledge. study design, data collection, data analysis, interpretation of study
First, the study population was not nationally repre- findings, writing of the report, or decision to submit the paper for
publication.
sentative and may over-represent populations with white
Where authors are identified as personnel of the International
ethnicity, and the mean UPF consumption were lower Agency for Research on Cancer/World Health Organization, the authors
than UK average.9 Second, the relatively small number alone are responsible for the views expressed in this article and they do
of vegans may have contributed to the large un- not necessarily represent the decisions, policy or views of the Interna-
certainties in the regression coefficients estimated for tional Agency for Research on Cancer/World Health Organization.
This research has been conducted using the UK Biobank Resource
this diet type. Third, misclassification bias may be pre- under Application Number 29239. The authors would like to thank the
sent for a few food items due to limited information participants of the UK Biobank study.
provided on food processing. We have taken the
approach assigning them to the most probable food Appendix A. Supplementary data
group based on published findings of common foods Supplementary data related to this article can be found at https://doi.
org/10.1016/j.eclinm.2024.102931.
and drinks consumed in the UK.9 Fourth, the study is
limited by 39.6% of the sample with only one 24-h
recall, but our sensitivity analysis restricting to sam- References
ples with >1 dietary recalls showed largely similar 1 Climate Action Tracker. The CAT thermometer. Available from:
findings. Fifth, the 24-h recall data were completed by https://climateactiontracker.org/global/cat-thermometer/. Accessed
January 27, 2024.
the UK Biobank participants during 2009–2012. Thus, 2 Mbow C, Rosenzweig C, Barioni LG, et al. Food security. In:
the dietary patterns compared in this study may not fully Shukla PR, Skea J, Calvo Buendia E, et al., eds. Climate Change and

www.thelancet.com Vol 78 December, 2024 11

 Articles

Land: an IPCC special report on climate change, desertification, land 16 Rauber F, da Costa Louzada ML, Chang K, et al. Implications of
degradation, sustainable land management, food security, and green- food ultra-processing on cardiovascular risk considering plant
house gas fluxes in terrestrial ecosystems. 2019. https://doi.org/10. origin foods: an analysis of the UK Biobank cohort. Lancet Reg
1017/9781009157988.007. Health Eur. 2024;43:100948.
3 Alves R, Perelman J, Chang K, et al. Environmental impact of di- 17 Good Food Institute. State of the industry report. In: Plant-based:
etary patterns in 10 European countries; a cross-sectional analysis meat, seafood, eggs, and dairy; 2023. Available from: https://gfi.
of nationally representative dietary surveys. Eur J Publ Health. org/wp-content/uploads/2023/01/2022-Plant-Based-State-of-the-
2024;34(5):992–1000. Industry-Report.pdf. Accessed July 16, 2024.
4 Food Standards Agency. Food and you 2: wave 6 data tables - En- 18 UK Biobank. UK Biobank. Available from: https://www.ukbiobank.
gland, Wales and northern Ireland combined. Available from: Food ac.uk/. Accessed August 18, 2023.
Standards Agency Website: https://www.food.gov.uk/research/ 19 Liu B, Young H, Crowe FL, et al. Development and evaluation of
food-and-you-2; 2023. the Oxford WebQ, a low-cost, web-based method for assessment of
5 Enromonitor International. Passport. Available from: https://www. previous 24 h dietary intakes in large-scale prospective studies. Publ
euromonitor.com/our-expertise/passport. Accessed October 26, Health Nutr. 2011;14(11):1998–2005.
2023. 20 UK Biobank. UK Biobank touch-screen questionnaire: final
6 Wickramasinghe K, Breda J, Berdzuli N, et al. The shift to plant- version. Available from: https://biobank.ndph.ox.ac.uk/showcase/
based diets: are we missing the point? Global Food Secur. refer.cgi?id=113241. Accessed October 30, 2023.
2021;29:100530. 21 Cassidy S, Chau JY, Catt M, et al. Cross-sectional study of diet,
7 Food and Agriculture Organization & World Health Organization. physical activity, television viewing and sleep duration in 233 110
Sustainable healthy diets: guiding principles. Available from: adults from the UK Biobank; the behavioural phenotype of car-
https://www.who.int/publications/i/item/97892415166482019. diovascular disease and type 2 diabetes. BMJ Open. 2016;6(3):
8 Monteiro CA, Cannon G, Levy RB, et al. Ultra-processed foods: e010038.
what they are and how to identify them. Publ Health Nutr. 22 UK Biobank. Index of multiple deprivation scores. Version 1.0.
2019;22(5):936–941. Available from: https://biobank.ndph.ox.ac.uk/ukb/ukb/docs/imd_
9 Rauber F, Steele EM, da Costa Louzada ML, et al. Ultra-processed baseline.pdf. Accessed August 18, 2023.
food consumption and indicators of obesity in the United Kingdom 23 UK Biobank. UK Biobank research ethics approval. Available from:
population (2008-2016). PLoS One. 2020;15(5):e0232676. https://www.ukbiobank.ac.uk/learn-more-about-uk-biobank/about-
10 Juul F, Parekh N, Martinez-Steele E, et al. Ultra-processed food us/ethics. Accessed August 18, 2023.
consumption among US adults from 2001 to 2018. Am J Clin Nutr. 24 Gehring J, Touvier M, Baudry J, et al. Consumption of ultra-
2022;115(1):211–221. processed foods by pesco-vegetarians, vegetarians, and vegans: as-
11 Hall KD, Ayuketah A, Brychta R, et al. Ultra-processed diets cause sociations with duration and age at diet initiation. J Nutr.
excess calorie intake and weight gain: an inpatient randomized 2021;151(1):120–131.
controlled trial of ad libitum food intake. Cell Metabol. 25 Ohlau M, Spiller A, Risius A. Plant-based diets are not enough?
2019;30(1):67–77.e3. Understanding the consumption of plant-based meat alternatives
12 Rauber F, Chang K, Vamos EP, et al. Ultra-processed food con- along ultra-processed foods in different dietary patterns in Ger-
sumption and risk of obesity: a prospective cohort study of UK many. Front Nutr. 2022;9:852936.
Biobank. Eur J Nutr. 2020;60:2169–2180. 26 Scientific Advisory Committee on Nutrition. SACN carbohydrates
13 Levy RB, Rauber F, Chang K, et al. Ultra-processed food con- and health report. Available from:https://www.gov.uk/government/
sumption and type 2 diabetes incidence: a prospective cohort study. publications/sacn-carbohydrates-and-health-report ; 2015. Accessed
Clin Nutr. 2020;40(5):3608–3614. October 1, 2024.
14 Chang K, Gunter MJ, Rauber F, et al. Ultra-processed food con- 27 World Health Organization. Sugars intake for adults and children.
sumption, cancer risk and cancer mortality: a large-scale prospec- Geneva: World Health Organization; 2015:1–59.
tive analysis within the UK Biobank. Eclinicalmedicine. 2023;56: 28 Willett W, Rockström J, Loken B, et al. Food in the Anthropocene:
101840. the EAT–Lancet Commission on healthy diets from sustainable
15 Yuan L, Hu H, Li T, et al. Dose–response meta-analysis of ultra- food systems. Lancet. 2019;393(10170):447–492.
processed food with the risk of cardiovascular events and all- 29 Leite FHM, Khandpur N, Andrade GC, et al. Ultra-processed foods
cause mortality: evidence from prospective cohort studies. Food should be central to global food systems dialogue and action on
Funct. 2023;14(6):2586–2596. biodiversity. BMJ Glob Health. 2022;7(3):e008269.

12 www.thelancet.com Vol 78 December, 2024