Public Health Nutrition: 25(5), 1310–1320                                             doi:10.1017/S136898002100197X
                         Pattern analysis of vegan eating reveals healthy and unhealthy
                         patterns within the vegan diet
                         Catherine T Gallagher, Paul Hanley and Katie E Lane*
                         Research Institute for Sport and Exercise Sciences, I. M. Marsh Campus, Liverpool John Moores University, Barkhill
                         Road, Aigburth, Liverpool L17 6BD, UK
                         Submitted 7 September 2020: Final revision received 1 April 2021: Accepted 5 May 2021: First published online 11 May 2021
                         Abstract
                         Objective: This study aimed to identify the types of foods that constitute a vegan
                         diet and establish patterns within the diet. Dietary pattern analysis, a key instru-
                         mentforexploringthecorrelationbetweenhealthanddisease,wasusedtoidentify
                         patterns within the vegan diet.
                         Design: A modified version of the EPIC-Norfolk FFQ was created and validated to
                         include vegan foods and launched on social media.
                         Setting: UK participants, recruited online.
                         Participants:Aconveniencesampleof129vegansvoluntarilycompletedtheFFQ.
                         Collected data were converted to reflect weekly consumptionto enable factor and
                         cluster analyses.
                         Results: Factor analysis identified four distinct dietary patterns including: (1) con-
                         venience (22%); (2) health conscious (12%); (3) unhealthy (9%) and (4) tradi-
                         tional vegan (7%). Whilst two healthy patterns were defined, the convenience
                         patternwasthemostidentifiablepatternwithaprominenceofveganconvenience
                         mealsandsnacks,vegansweetsanddesserts,sauces,condimentsandfats.Cluster
                         analysis identified three clusters, cluster 1 ‘convenience’ (26·8%), cluster 2 ‘tradi-
                         tional’ (22 %)andcluster3‘healthconscious’(51·2%).Clusters1and2consistedof                                     Keywords
                         an array of ultraprocessed vegan food items. Together, both clusters represent                        Dietary pattern analysis
                         almost half of the participants and yielding similar results to the predominant                                 Vegan diet
                         dietary pattern, strengthens the factor analysis.                                                             Convenience
                         Conclusions:Thesenovelresultshighlighttheneedforfurtherdietarypatternstud-                                      Traditional
                         ies with full nutrition and blood metabolite analysis in larger samples of vegans to                              Healthy
                         enhance and ratify these results.                                                                       Ultraprocessed foods
               Over half a million people in the UK (≈1% of the popula-           Fe, Ca, iodine, n-3, Se and Zn in poorly adapted or non-
                                                                                                       (9)
               tion) follow a vegan diet where all animal sources are sub-        fortified vegan diets  . In dietary terms, a traditional vegan
               stituted withplant-basedalternatives.Veganismquadrupled            diet refers to a diet that omits all products derived wholly
                                                    (1)
               between 2014 and 2019 in the UK         with 600 000 vegans        orpartlyfromanimalorigin.Thedietfocusesmoreonwhole-
                                (2,3)                                                                                  (10)
               reportedin2019      , whilethepopularityinvegandietscon-           grains, pulses, fruit and vegetables    . It remains unclear if
                                         (4)
               tinuestogrowworldwide .Thefoodindustryisresponding                 modern vegan dietary adaptation methods can deliver the
               to this by producing more processed vegan food and drink           samehealthadvantagesastraditionalvegandiets.Forexam-
                                          (2,5)
               products than ever before     . In January 2021, ‘Veganuary’       ple, if vegans are choosing ultraprocessed vegan products
               saw over 440 000 people in the UK committing to a vegan            over more natural plant-based alternative sources, could
                   (6)                                                                                                                       (11)
               diet  , raising the profile of plant-based eating which has        this compromise the overall quality of the vegan diet?        .
                                                                 (7)
               been associated with a range of health benefits .                  Bywayofdefinition, ultraprocessed foods refer to products
                  It is reported that a well-planned vegan diet can meet all      mostly or entirely formulated from substances derived
                                                                 (8)                                                                         (12)
               the nutritional requirements necessary for health   .Thereis       from foods that typically contain little or no whole foods    .
               still some debate, however, about the nutritional quality of       These products are usually high in saturated fat, sugar and
               vegan diets and the risk of nutritional deficiencies, notably      salt. The majority of these food items are also considered
                                                                                                                                           (13,14)
               some key micronutrients such as vitamin B ,vitaminD,               poor sources of protein, fibre and micronutrients             .
                                                              12
               *Corresponding author: Email k.e.lane@ljmu.ac.uk
               ©TheAuthor(s),2021.PublishedbyCambridgeUniversityPressonbehalfofTheNutritionSociety.ThisisanOpenAccessarticle,distributedunder
               the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distri-
               bution, and reproduction in any medium, provided the original work is properly cited.
   https://doi.org/10.1017/S136898002100197X Published online by Cambridge University Press
                 Pattern analysis of vegan eating study                                                                                        1311
                                                                                                                                         (30)
                 Studies over the past two decades have provided important           Adaptation followed methods used by Dyett et al.        in their
                 information on the diet quality of various types of vegetar-        evaluation of a validated FFQ for self-defined vegans in the
                 ians, but no single study has addressed the quality of specific     USA.VeganfooditemsavailableintheUKwereidentified
                 vegan diets. Orlich et al.(7) reveal Adventist vegans con-          from mainstream UK supermarkets and vegan UK forums.
                 sumedthelowestamountsoffoodsandsnackshighinadded                    A collection of naturally vegan food products and newly
                 sugars and saturated fats, in comparison with non-vegetar-          emerging ultraprocessed vegan products were included in
                 iansandothervegetariangroups.Thisargumentisconsistent               the FFQ. Ten vegan volunteers in a UK university who met
                 with much of the literature surrounding vegan diets(15−18)          the study criteria took part in an initial pilot study.
                                                                               .
                 However, the main weakness with this research is that it is         Feedback from the volunteers was taken on board to further
                 outdated and perhaps not considering the increasing variety         modify the vegan FFQ. To further enhance validation of the
                 of processed food and drinks that are now available to              vegan-adapted FFQ, a focus group of Health and Care
                 vegans. In 2018, the UK developed more vegan products               Professions Council registered dietitians in the UK were then
                                       (5)                                           consulted. Modifications and additions to the food groups
                 thananyothernation .PopularUKsupermarketsarereact-
                 ing by producing vegan wines with a pledge to ensure their          were made accordingly based on the dietitians comments
                                                                         (19)        to generate the finalised version of the vegan-adapted FFQ
                 full range is suitable for vegans in the coming years      .In
                                                                            (20)     (see online supplemental material S1). Questionnaire instruc-
                 2019, Galaxy launched a vegan Mars bar in the UK              ,
                                                                            (21)     tions stated that the FFQ must reflect dietary habits over the
                 and in 2020, Mc Donald’s launched its first vegan meal        .
                 Thus, the production of vegan alternatives including vegan          past month, and therefore, participants must have been fol-
                 snacks and fast foods is prevalent and represents one of the        lowing a vegan diet for at least 1 month. Further questions
                 mainproductdevelopmenttrendswithinthefoodandretail                  were included such as motivations for adopting vegan life-
                 industry. However,manyofthesefooditemscanbehighin                   style, age, length of time vegan, cooking skills and supple-
                 saturated fats and sugars and if eaten regularly may pose a         ment use to ensure evaluation of factors influencing diet
                                                                                                                         (31)
                 risk to health. Therefore, a review of current vegan dietary        choice and nutritional knowledge      .
                 patterns is urgently required to address these uncertainties.
                    Several studies have evaluated the dietary patterns of           Recruitment
                 omnivores, pesco, lacto, ovo and semi-vegetarians in com-           Online social media accounts (Instagram and Facebook)
                                          (22−26)
                 parisonwithvegandiets          , but nonetodatehassubjected         wereusedtorecruitsubjects.TheFFQwaslaunchedonsocial
                 vegandietstodietarypatternanalysis.Itis important toestab-          media accounts in the UK. The recruitment team asked for
                 lish whether the increased availability of processed vegan          vegans in the UK to complete and share the FFQ. In order
                 replacementsforanimal-basedproductsisleadingtohabitual              toreducebias,participants’involvementinthisstudywasvol-
                 consumption of an array of ultraprocessed foods. The meth-          untary. Participants gave informed consent prior to complet-
                 odology for this unique study includes an innovative dietary        ing the voluntary FFQ. Inclusion criteria required participants
                 pattern analysis of vegan diets. Dietary pattern analysis offers    to be living in the UK and aged over 18 years, so only adults
                 aneffective way of understanding the diverse eating patterns        could take part. Participants were also required to have fol-
                 within vegan diets by evaluating methods of adaptation and          lowedavegandietforatleast1month.Thisallowedspecific
                 substitution(27). It was hypothesised that some vegan diets         dietary patterns to be captured.
                 wouldincorporate a range of food groups representing a tra-
                 ditional well-planned vegan diet. This was expected to be the       Statistical analysis
                 most common dietary pattern. The vegan food industry has
                 evolved; therefore, it was predicted that a convenience style       Statistical analyses were performed using IBM SPSS(version
                 eating pattern could also emerge, representing a small pro-         26.0; SPSS Inc.) and Microsoft Excel 2013. Data screening
                 portion of the participants.                                        and cleaning was conducted to check for any outliers and
                    This study aimed to identify patterns within the vegan           errors  on the categorical and continuous variables.
                 diet by establishing the everyday foods that vegans are             Descriptive statistics such as frequencies and percentages
                 choosing to consume enabling an evidence-based evalu-               were calculated for characterisation of the participants (i.e.
                 ation of the vegan diet.                                            gender,agegroupsandlengthoftimevegan).Statisticaltests
                                                                                     were used to calculate the significance of error.
                 Methods                                                             Data screening
                                                                                     SelectedfrequencyofconsumptionforeachfoodintheFFQ
                 FFQ                                                                 wascodedtoreflecthowofteneachitemwasconsumedper
                 AFFQwascreatedusingLJMU-approvedOnlinesurveytool,                   weekfordietarypatternanalysisasfollowed:NEVERorless
                 anonline food questionnaire creator, to enable the provision        than once/month 0, 1–3/month, once a week, 2–4/week,
                 of a validated interactive dietary assessment tool(28). The vali-   5–6/week, once a day, 2–3/d, 4–5/d, 6þ/d. This design
                 datedEPIC-NorfolkFFQ(29)wasmodifiedtoincludequestions               was taken from the validated EPIC-Norfolk FFQ, which
                                                                                                                         (32,33)
                 representative of foods and drinks suitable for vegans.             hasalsobeenusedinotherstudies             . Twomethodswere
   https://doi.org/10.1017/S136898002100197X Published online by Cambridge University Press
               1312                                                                                                              CT Gallagher et al.
               Table 1 Food groups and food items included in the analysis of the FFQ cohort
               Food groups 1                        Food groups 2 (variables)          Definition and content
               1. Legumes & nuts                    1. Protein alternatives to meat &  Soya, Tempeh, Tofu, silken tofu, lentils, pulses, nuts, falafel
               2. Meat alternatives                   fish
               3. Meat-free processed alternatives  2. Processed meat alternatives     Vegan nuggets, burgers, bacon, sausage, no fish fingers, ham
               4. Fish alternatives                                                      slices, turkey slices, chicken slices, meat-free mince, vegan
                                                                                         chorizo
               5. Vegan sandwiches                  3. Convenience meals & snacks      Garlic bread, pizza, sausage rolls, chips, ready prepared mash,
               6. Vegan wraps                                                            selection of pre-made vegan sandwiches & wraps, ready
               7. Ready-prepared foods                                                   meals, Not-zarrella sticks, French fries
               8. Fresh fruit                       4. Fruit                           Apples, pears, oranges, grapefruit, bananas, grapes, melon,
               9. Tinned fruit                                                           peaches, strawberries, avocado, tinned fruit, dried fruit
               10. Dried fruit
               11. Vegetables                       5. Vegetables                      Carrots, spinach, broccoli, Brussel sprouts, cabbage, peas,
               12. Soup                                                                  green beans, courgettes, cauliflower, parsnips, leeks, onions,
                                                                                         garlic, mushrooms, sweet peppers, beansprouts, green salad,
                                                                                         mixed vegetables, watercress, tomatoes, sweetcorn, beet-
                                                                                         root, coleslaw, vegetable soup, rainbow rice
               13. Starchy carbohydrates            6. Refined grains                  White bread, scones, crackers, pitta, sugary cereal, plain cer-
                                                                                         eal, white rice, pasta, tinned pasta, noodles, lasagne, cereals
                                                                                         (except high fibre options)
               14. High-fibre carbohydrates         7. Wholegrains                     Brown bread, wholemeal bread, porridge, all bran, wholegrain
                                                                                         cereals, brown rice, wholemeal pasta, wild rice
               15. White potatoes                   8. Potatoes                        Boiled potatoes, roast potatoes, sweet potatoes, home-made
               16. Sweet potatoes                                                        mash, baked potatoes, baby potatoes
               17. Plant-based milks                9. Dairy alternatives              Oat milk, soya milk, almond milk, rice milk, hazelnut milk, coco-
               18. Vegan cheese                                                          nut milk hemp, pea milk, Nutritional yeast, vegan hard
               19. Vegan yoghurts                                                        cheese, Yoghurt alternatives,
               20. Fats and oils                    10. Fats and oils                  Vegan butter spreads, pesto, peanut butter, olive oil, sunflower
                                                                                         oil, coconut oil, avocado oil, canola oil, sunflower ghee, rape-
                                                                                         seed oil, fry light
               21. Cakes & biscuits                 11. Vegan cakes & biscuits         Cookies, Digestive twists, bourbons, Lotus Biscoff, vegan
                                                                                         sponge cake, vegan cereal bars, party ring minis, granola
                                                                                         bars
               22. Sweets and desserts              12. Vegan sweets & desserts        Fudge, cheesecake pots, chocolate mousse pots, dark choco-
                                                                                         late, non-dairy ice cream, churros, star burst sweets
               23. Vegan crisps                     13. Vegan crisps                   Lentil Chips, Kettle chips, walkers, tortilla chips, vegetable
                                                                                         chips, pretzel bites
               24. Sauces & condiments              14. Sauces and condiments          BBQsauce,cheesesauce, Red lasagne sauce, free from
                                                                                         sauce, olive oil, vegetable oils, seeds, tahini, vegetable pates,
                                                                                         mayonnaise, hummus, chocolate spread, coleslaw, potato
                                                                                         salad
               25. Salt                             15. Salt                           All added salts
               26. Alcohol                          16. Alcohol                        Vegan friendly alcohols
               27. Vegan takeaway                   17. Vegan takeaway                 From fast-food outlets providing vegan options
               28. Cooking                          18. Cooking from scratch           Additional question to help with establishing vegan patterns
               29. Recipes used                     19. Creating own recipes           Additional question to help with establishing vegan patterns
               30. Use of vegan brands              20. Purchasing vegan brands        Additional question to help with establishing vegan patterns
               used to classify the individual food items before applying            used factor analysis as a statistical method to reduce large
               factor and cluster analyses. In the first instance, the food          sets of dietary intake variables into smaller sets of variables
                                                                                                                       (36,37)
               and drink items were combined and collapsed into thirty               that represent eating patterns         . The smaller sets of
               food groups and in the second twenty food groups                      composite variables derived through the principal compo-
               (Table 1), respectively, with similar nutrient profiles, simi-        nent method are referred to as ‘components’, and the
                                                                      (34)
               larly to previous research by Ashby-Mitchell et al.       .           variables within these are referred to as ‘factors’.The
                                                                                     Kaiser–Mayer–OlkinmeasureandBartlett’stestofsphericity
                                                                                     were undertaken before applying the principal component
               Factor analysis                                                       method, to ensure the data were suitable for factor analy-
               Factor analysis with the principal component method was                 (38)
                                                                                     sis  . The twenty foodvariables fromfoodgroups2shown
               performed in SPSS, with the procedure ‘dimension reduc-               inTable 1wereenteredintothefactoranalysis.Obliminand
               tion’ and ‘FACTOR’ on both sets of food groups to identify            Varimax rotations were applied. The components derived
               the primary components, which accounted for variation in              from the Oblimin rotation were selected similar to previous
               dietary intake. However, the smaller set of food groups                                                                    (39,40)
                                                                                     work by researchers exploring dietary patterns            .The
               (n20)wasdeemedmoreappropriateduetothesmallsam-                        rotation redistributes the variance of each component
                       (35)                                                                                             (41)
               plesize    . Themethodsfollowedpreviousstudiesthathave                allowing for a simpler structure      . Oblimin rotation was
   https://doi.org/10.1017/S136898002100197X Published online by Cambridge University Press
                Pattern analysis of vegan eating study                                                                                     1313
                Fig. 1 Scree plot to show eigenvalues of each component number
                chosenasthepreferredmethodof‘rotation’asithasarange                   Most participants were female (87%), and most were
                                                                       (42)
                of advantages compared with other types of rotation       .       aged18–24(36%)years.Themostcommonreasonselected
                   The number of components selected was based on the             for following a vegan lifestyle was ‘Health, Environment &
                assessmentofthescreeplot,withvalues>1deemedappro-                 Animalwelfare’(43%).Healthbenefitswereintheminority
                priate to establish the patterns that explain the largest         with only 3% following the vegan lifestyle primarily for
                                        (36)
                proportion of variance     . Six components had an eigen-         ‘health’. It is important to note that on the questionnaire,
                value>1,buttherewasagradualbreakinthescreeplotafter               these were presented as separate reasons and not a single
                the fourth component (Fig. 1); therefore, four components         reason. Participants were able to select more than one rea-
                were retained. The dietary patterns were characterised by         son.Mostvegans(41%)hadbeenfollowingavegandietfor
                highandlowintakesofveganfoodanddrinks.Thepatterns                 1–3years.Someparticipants(17%)wereeatingavegandiet
                werelabelledbasedonthetypesoffactorsrepresentingthe               for <6 months; (8%) 6–12 months; (23%) 4–10 years and
                component and explanations in the literature.                     (11%) over 10 years. From those taking nutritional supple-
                                                                                  ments,themajoritytookvitaminB (68%).Almosthalftook
                                                                                                                     12
                Cluster analysis                                                  vitaminD(42%).Amoderatenumber(26%)weretakingFe
                Two-factor cluster analysis identifies groupings by running       supplements and 19% took Ca supplements. A small num-
                pre-clustering first and then by running hierarchical methods     berofparticipants(15%,12%,14%and7%)consumedZn,
                                                                          (35)
                to enable automatic selection of the number of clusters      .    iodine, n-3 and Se supplements, respectively. Again these
                Two-factor cluster analysis was performed to order the twenty     micronutrientswerepresentedinalistonthequestionnaire,
                food groups in a dendrogram, where food groups with the           and participants were able to select more than one
                highest correlations were further grouped together, while         supplement.
                samples with small correlations were widely separated. In
                particular,thetwofoodgroupswiththelargestcorrelationwere
                identified and merged into a single ‘synthetic’ sample. The       Factor analysis
                remaining food groups were then searched for the largest          Inspection of the correlation matrix revealed the presence of
                correlation with the synthetic sample. This process was           manycoefficients of 0·3 and above. The Kaiser–Meyer–Olkin
                                                                                                                                             (38)
                repeated until all samples were merged into a single sample,      value was 0·727, reaching the recommended value of 0·5
                                                                                                                  (44)
                and the correlations among samples were then expressed as         The Barlett’sTestofSphericity       reached statistical signifi-
                                  (43)                                            cance,supportingthefactorabilityofthecorrelationmatrix(35).
                a hierarchical tree  .
                   Thedietarypatternswerecharacterisedbyhighandlow                    Factor analysis with the principal component method
                intakes of vegan food and drinks. The clusters were               revealed the presence of six components with eigenvalues
                labelledbasedonthetypesofinputsrepresentingthecom-                exceeding1,explaining22%,12%,9%,7%,7%and5%of
                ponent and explanations in the literature.                        the variance. However, inspection of the scree plot (Fig. 1)
                                                                                  revealed a gradual break after the fourth component.
                                                                                  Therefore,thefirst fourcomponentsexplainthelargestpro-
                Results                                                           portion of variance in the dietary intake data and were
                                                                                  retained as ‘dietary patterns’. Together these components
                Participant characteristics                                       represent a cumulative percentage of 50% of the inter-
                Data collection took place from Monday 2 March 2020               individual variability. To aid the interpretation of these four
                throughFriday3April2020.Therewere129fullycompleted                components, oblimin rotation was performed, representing
                FFQ. Sample characteristics are presented in Table 2.             fourdefinite dietary patterns (Table 3). The first component
   https://doi.org/10.1017/S136898002100197X Published online by Cambridge University Press