Introduction

Vegetarian (excluding meat and fish/seafood) and vegan (excluding all foods of animal origin) diets are becoming increasingly popular [1, 2], also among children and adolescents. In a representative German survey, the proportion of vegetarian adolescents (12–17 years old) increased from 1.6% (2006) to 5.0% (2015–2017) [3]. Plant-based diets, such as vegetarian or vegan diets, are associated with various health benefits, particularly regarding the prevention of cardiovascular diseases [4,5,6,7,8,9,10]. Compared to an omnivore diet, a more favorable dietary fatty acid profile was found in vegans (lower intake of saturated fatty acids, higher intake of polyunsaturated fatty acids) as well as a higher intake of dietary fiber, vitamin C, vitamin E, folate, magnesium and iron [11, 12]. However, limiting food choices by excluding foods of animal origin is also discussed in terms of increased risk of certain nutrient deficiencies, in particular in the pediatric population [13, 14]. The growth and development processes during this vulnerable phase of life are associated with an increased energy and nutrient requirement in relation to body weight. Therefore, an adequate nutrient intake during childhood and adolescence is of particular importance for normal physical and mental development [15,16,17]. Potentially critical nutrients in a vegetarian child diet are iodine and vitamin D, just as in omnivore diets, as well as iron, zinc, and long-chain omega-3 fatty acids. In vegan diets, in addition to the same critical nutrients for vegetarian diets, the intake of protein, vitamin B12, vitamin B2 and calcium also need attention [18]. To ensure an adequate intake of these nutrients careful planning is required with a special focus on nutrient-dense foods, such as whole grains, legumes, nuts and seeds, and on fortified foods [18, 19]. Particularly in vegan diets, attention should also be paid to a targeted intake of fortified foods [19].

One way to describe the overall diet of a population in its complexity is through dietary pattern analysis [20]. Exploratory, a posteriori, pattern analyses are statistical methods that generate population-specific dietary patterns from the collected dietary intake data of a population [21]. A number of studies have investigated the exploratory dietary patterns of children and adolescents in Western countries [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37]. In these studies, different dietary patterns could be identified, which not only showed significant differences in the quality of food choices, but were also associated to different extents with energy and nutrient intake [22, 25, 30, 37]. Among adult vegans, it has been shown that in addition to traditional vegan dietary patterns focusing on unprocessed foods, there are also unhealthier dietary patterns characterized by higher intake of processed foods, i.e., vegan sweets, snacks, fast food, and convenience meals [38, 39]. To the best of our knowledge, there are no studies on dietary patterns among vegetarian or vegan children and adolescents and potential associations with nutrient intake during this vulnerable phase of life. Thus, the aim of the present study was to identify dietary patterns within the vegetarian, vegan, and omnivore diets of children and adolescents using data from the Vegetarian and Vegan Children and Youth Study (VeChi Youth Study). In addition, we explored the correlations between the extracted dietary patterns and nutrient intake.

Methods

The VeChi Youth Study

The VeChi Youth Study is a cross-sectional study that collected data on diet, lifestyle, anthropometry, and nutritional status of vegetarian, vegan, and omnivore children and adolescents in Germany. All examinations were performed with parental and older participants’ (≥ 14 years) written consent. Examinations included a 3-day weighed dietary record, anthropometric measurements, a spot urine, a blood sample as well as questionnaires on socio-demographic factors and lifestyle. Examinations and interviews were performed by trained staff. Further details about the study have been described elsewhere [40,41,42].

The VeChi Youth Study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of Witten-Herdecke University (139/2017). The study has been registered at the German Clinical Trials Register (DRKS00012835).

Study sample

The VeChi Youth Study included 401 healthy children and adolescents aged 6–18 years who were examined in three study centers in Germany. One participant age 5.5 years and one participant age 19.1 years were also included into the study sample. Due to the low prevalence of vegans in Germany [2, 3], the collective had to be recruited specifically. Exclusion criteria were as follows: (a) diagnosed diseases that could affect the studied variables (e.g., enteropathy, pancreatic diseases, metabolic disorders like phenylketonuria or fructose malabsorption), and (b) special diets other than vegan or vegetarian diet, e.g., predominantly (70%) raw food diet (according to [43]). Of the 401 participants included, 390 (girls n = 221; 57%) compleed a three-day weighed food record. Accordingly, data from 110 vegans (28%), 145 vegetarians (37%) and 135 omnivores (35%) were available for the present analysis (Supplementary Fig. 1). The power calculation for the VeChi Youth Study was carried out for serum ferritin as the primary outcome, described in detail elsewhere [41].

Questionnaires

An online questionnaire was used to record socio-demographic variables, such as nationality, monthly income, or parental education, as well as nutritional variables, such as duration of vegetarian or vegan diet, motivation for the diet or intake of dietary supplements. The Winkler Index was used to describe socioeconomic status [44]. This index is made up of three scores on school education, training, and net household income (1–7 points each). If the mother’s and father’s scores differed, the higher score was used and categorized as the family index for socioeconomic status. Using the Winkler index, the families were divided into three categories: low (3–8 points), medium (9–14 points) or high (15–21 points) social status. Physical activity was assessed using the validated Adolescent Physical Activity Recall Questionnaire [45]. The duration of physical activity (hours/week) and MET minutes per week were calculated using the information provided on organized (e.g. training in a sports club) or unorganized sports activities (recreational sports, e.g. football matches with friends or jogging) and using databases on the metabolic intensity of the activities [46, 47].

Anthropometric measurements

Body weight (kg) was measured on the day of the examination in underwear and without shoes using an electronic column scale (Seca 799, graduation 100 g, up to 150 kg body weight). Height (cm) was determined using a stadiometer (Seca 222, graduation 1 mm). The body mass index (BMI, kg/m2) was calculated from the quotient of body weight (kg) and the squared height (m). The gender- and age-independent BMI standard deviation scores (BMI-SDS) were calculated using the national reference data from Kromeyer-Hauschild [48]. The BMI-SDS indicates how many standard deviations an individual BMI is above or below the BMI median of the reference population for a given age and gender.

Dietary assessment

Dietary intake of study participants was assessed using 3-day weighed dietary records as described in detail elsewhere [40]. In short, over a period of three consecutive, freely selectable days, participants and/or their parents weighed and recorded all foods and beverages consumed as well as leftovers using electronic scales. If weighing was not possible (e.g., in case of eating out) participants were asked for semi-quantitative recording using household measures (e.g., tablespoons, pieces). For commercial food products (e.g., ready-to-eat meals and meat or dairy alternatives), the exact brand name was also recorded. Missing data were assessed by the study staff by requesting the information from the parents via email.

Energy, nutrient and food group intakes were calculated using the LEBTAB food composition database (Version III, DONALD Study, University of Bonn) [49] which contains nutritional information on staple foods based on standard German food composition tables (Bundeslebensmittelschlüssel BLS 3.02) as well as brand-specific foods and dietary supplements. The energy and nutrient composition of commercial food products were calculated by recipe simulation based on the nutrient and ingredient declaration.

To calculate food group intakes, each food item consumed was assigned to one of 22 food groups (Supplementary Table 1). Food groups were formed based on similarities in nutrient profile, ingredients, or culinary use. Mean intake of each food group (in g/day) was determined by summing daily intake amounts and then averaging, i.e., dividing by the 3 dietary record days. Food group intake was calculated as consumption amount per 1000 kcal total energy intake per day (g/1000 kcal/day) to account for differences in energy requirements and total quantitative intake per day between boys and girls as well within the wide age range of the study population [40]. For dietary pattern analysis, food group intakes (g/1000 kcal) were standardized (mean 0, standard deviation 1) using z-transformation.

Diet group classification

Vegetarian, vegan or omnivore diets were categorized according to the following question during recruitment:

Do you/does your child eat.

  • a vegetarian diet (no meat, sausage, fish, but dairy and/or eggs)

  • a vegan diet (no meat, sausage, fish, dairy and eggs)

  • an omnivore diet (including meat and/or fish)?

In addition, in the online questionnaire some crosscheck questions were asked, whether there are exceptions in food intake (e.g., occasional intake of dairy products in vegan diets or occasional intake of meat or fish/seafood in vegetarian diets). Based on these control questions, 24 study participants were reclassified [40].

Statistical analysis

All statistical analyses were performed using SAS® 9.4. The significance level was set at P < 0.05. Data were checked for plausibility and outliers.

A separate principal component analysis (PCA) was conducted for each diet group (vegetarian, vegan, omnivore). The suitability of the data for PCA was assed using the Kaiser-Meyer-Olkin (KMO) and Bartlett’s sphericity test. With values < 0.5, the KMO test indicated insufficient data adequacy (vegetarian group: 0.39, vegan group: 0.18, omnivore group: 0.31) whereas Bartlett’s test confirmed the data suitability for PCA (all diet groups P < 0.0001).

Due to the exclusion of animal source food groups in the vegetarian and vegan diets, the number of food groups analyzed differed depending on the diet group (omnivore: 22 food groups, vegetarian: 21 food groups, vegan group: 17 food groups).

Principal component analyses (SAS® procedure PROC FACTOR (method = prin)) were performed using the following criteria to identify the number of significant dietary patterns [50]: Eigenvalue > 1; Scree plot evaluation (identification of a “break” in the graphical presentation of eigenvalues; Interpretability (adequate number of food groups with high factor loadings in the dietary pattern: according to Hatcher [51], at least 3 factor loadings ≥ |0.4|). The components were rotated using orthogonal transformation (varimax rotation, SAS® option rotate = varimax) to obtain uncorrelated patterns with a simpler structure and better interpretability. The number of patterns to be extracted was selected using the “nfact” option [50].

Food groups with factor loadings ≥ |0.2| were considered to contribute significantly to a dietary pattern [31, 33, 52] and were therefore used for the description of dietary patterns. Factor loadings ≥ |0.4| were considered high. The identified dietary patterns were labeled according to the food groups with high positive factor loadings or qualitative characteristics of the diet. Although the chosen labels do not comprehensively describe the underlying dietary patterns, they facilitate the presentation and discussion of the results.

Each study participant was assigned a score for each dietary pattern, which was the sum of the individual standardized food group intakes (g/1000 kcal) weighted by the corresponding factor loadings of the food groups in each dietary pattern. These pattern scores indicate the degree of adherence to the respective dietary pattern. Thus, a high individual pattern score indicates that the diet of the study participant strongly matches the dietary pattern under consideration.

Spearman correlation coefficients were calculated to examine the association between dietary pattern scores and energy and nutrient intakes within the diet groups. A positive correlation between a dietary pattern and a nutrient indicates that a high dietary pattern score (high correspondence between a participant’s diet and this pattern) is associated with a higher nutrient intake. A negative correlation, on the other hand, means that a high score in the pattern is associated with a lower nutrient intake. Because nutrient intake is strongly correlated with energy intake, the nutrient density approach was used to adjust for energy intake [53]. For this purpose, macronutrients were analyzed as %E and micronutrients and dietary fiber as intakes per 1000 kcal [53, 54]. According to Cohen’s established categorization [55] and in line with a previous study on correlations between children’s dietary patterns and nutrient intake [37], |r| > 0.5 was interpreted as a strong correlation.

Results

Sample characteristics

The characteristics of the study population, stratified by diet group, are shown in Table 1. The median age of the study population was 12.5 years and age ranged from 5.5 to 19.1 years. The median BMI-SDS was below zero in all diet groups. Most of the participants (n = 278; 71.3%) were from families with a high socioeconomic status. Vegetarian subjects stated that they had been following this diet for a median of 4.1 years and vegan subjects for a median of 3.4 years. In all three diet groups, the participants completed the 3-day weighed dietary records predominantly on two weekdays as well as one weekend day and only minor differences were observed regarding the seasons in which the records were completed (Supplementary Table 2). Further characteristics of the study population and data on food group intake have been described in detail elsewhere [41, 42].

Table 1 Sample characteristics of VeChi Youth Study participants (n = 390, 6–18 years old) stratified by diet group
Full size table

Dietary patterns

Three predominant components (dietary patterns) were identified in the vegetarian diet group (Supplementary Table 3), four in the vegan group (Supplementary Table 4), and five in the omnivore group (Supplementary Table 5). The identified components together explained 32.7% of the variance in food group intake in the vegetarian group, 43.2% in the vegan group, and 43.0% in the omnivore group. The dietary patterns and factor loadings are summarized in Fig. 1 and described in detail below.

Fig. 1
figure 1

Spider diagram of factor loadings for the dietary patterns identified by principal components analysis among (a) vegetarian, (b) vegan, and (c) omnivore children and adolescents in the German VeChi Youth Study

Full size image

Vegetarian diet group

The first identified dietary pattern “Animal foods” was characterized by high positive factor loadings for unfermented and fermented dairy products, animal fats, and eggs or egg dishes. In contrast, high negative factor loadings were found for plant-based dairy and meat alternatives and for legumes. The second component “Vegetables and fruits” had high positive factor loadings for vegetables and fruits as well as high negative factor loadings for convenience foods and sugar-sweetened beverages. The third pattern “Meat alternatives and potatoes” was associated with high positive factor loadings for potatoes and potato products, vegetables, plant-based meat alternatives, and juices and had a high negative factor loading for sweet pastries.

Vegan diet group

The first identified pattern was labeled “Vegetables and legumes” because it had high positive factor loadings for vegetables, potatoes and potato products, legumes, and water. The second component “Refined carbohydrates” was characterized by high positive factor loadings for sweets and salty snack foods and refined grains and had high negative factor loadings for whole grains and plant-based dairy alternatives. The third component “Meat alternatives and juices” was associated with high positive factor loadings for plant-based meat alternatives and juices and high negative factor loadings for nuts and seeds. The fourth pattern “Fruits and convenience foods” had high positive factor loadings for fruits and convenience foods as well as high negative factor loadings for sweet pastries and vegetable oils and fats.

Omnivore diet group

The first component called “Flexitarian” in the omnivore diet group was characterized by high positive factor loadings for whole grains, plant-based spreads, dairy alternatives, and meat alternatives, and had high negative factor loadings for sweets and salty snack foods. The second component “Vegetables and fruits” had high positive factor loadings for vegetables, water, and fruits and high negative factor loadings for sweet pastries and sugar-sweetened beverages. The third component “Dairy Products” had high positive factor loadings for animal fats and unfermented dairy products. High negative factor loadings were also found for vegetable oils and fats and legumes in this pattern. The fourth component “meat and convenience foods” in the omnivore group was associated with high positive factor loadings for meat and fish as well as convenience foods and with high negative factor loadings for potatoes and potato products. The fifth dietary pattern “Refined grains and juices” was characterized by high positive factor loadings for refined grains and juices. A high negative factor loading was found in this pattern for fermented dairy products.

Association between dietary patterns and energy and nutrient intake

For each dietary pattern, several significant associations were found with energy and nutrient intakes (Tables 2, 3 and 4), but correlation coefficients |r| were often < 0.5.

Table 2 Spearman correlation between dietary pattern scores and energy and nutrient density in vegetarian participants of the German VeChi Youth Study (n = 145)
Full size table
Table 3 Spearman correlation between dietary pattern scores and energy and nutrient density in vegan participants of the German VeChi Youth Study (n = 110)
Full size table

Vegetarian diet group

Table 2 presents the correlations between dietary pattern scores and energy and nutrient density in the vegetarian diet group. Both the “Vegetables and fruits” and the “Meat alternatives and potatoes” patterns showed moderate negative correlations with energy density (r = -0.32 and r = − 0,37). No significant correlation was found between the “Animal foods” pattern and energy density. Higher scores on the vegetarian “Animal foods” pattern were strongly correlated with fat quality (higher saturated fatty acids, r = 0.60; lower polyunsaturated fatty acids, r = -0.54 and linoleic acid r = -0.57). Strong negative correlations were also seen for nutrient density of dietary fiber (r = -0.65) and iron (r = -0.50) and a strong positive correlation for vitamin B12 density (r = 0.73). The “Vegetables and fruits” pattern showed strong positive correlations with dietary fiber (r = 0.69), vitamin B1 (r = 0.53) and magnesium (r = 0.60). Higher scores on the ‘Meat alternatives and potatoes’ pattern were associated with greater tocopherol-equivalents (r = 0.51) and vitamin C density (r = 0.61) in the diet of the vegetarian children and adolescents.

Vegan diet group

Table 3 shows the correlations between dietary pattern scores and energy and nutrient density in the vegetarian diet group. Strong negative correlations with energy density were seen for the dietary patterns “Vegetables and legumes” (r= -0.64) and “Fruits and convenience food” (r= -0.53). The “Refined carbohydrates” and “Meat alternatives and juices” patterns showed no significant correlations with energy density. Higher scores on the vegan “Vegetables and legumes” pattern were strongly correlated with vitamin C density (r = 0.51). The pattern „Refined carbohydrates” was strongly negatively correlated with dietary fiber (r= -0.60), vitamin B1 (r= -0.68), magnesium (r= -0.64), iron (r= -0.66) and zinc (r= -0.69). The “fruits and convenience food” pattern showed strong positive correlations with dietary fiber (r = 0.67) and magnesium (r = 0.55) and strong negative correlations with total fat (r= -0.52) and saturated fatty acids (r= -0.53). No strong correlations were found between the “Meat alternatives and juices” pattern and any of the nutrients, but moderate negative correlations were found with dietary fiber (r= -0.36), vitamin B1 (r= -0.34), magnesium (r= -0.45), iron (r= -0.31) and zinc (r= -0.46).

Table 4 Spearman correlation between dietary pattern scores and energy and nutrient density in omnivore participants of the German VeChi Youth Study (n = 135)
Full size table

Omnivore diet group

Table 4 presents the correlations between dietary pattern scores and energy and nutrient density in the omnivore diet group. The “Vegetables and fruits” pattern showed strong negative correlations with energy density (r= -0.63), whereas the “Dairy products” showed moderate positive correlations with energy density (r = 0.31). Higher scores on the “Flexitarian” dietary pattern were strongly correlated with dietary fiber (r = 0.62), magnesium (r = 0.63) and iron density (r = 0.57). The “Vegetables and fruits” pattern showed strong positive correlations with dietary fiber (r = 0.67), folate-equivalents (r = 0.52) and magnesium density (r = 0.53) as well as strong negative correlations with added sugars (r= -0.57). The “Meat and convenience food” correlated strongly with protein density (r = 0.53). No strong correlations were found between the “Refined grains and juices” pattern and any of the nutrients.

Discussion

In the cross-sectional VeChi Youth Study, different dietary patterns were identified not only among omnivore children and adolescents, but also within the vegetarian and vegan diet groups. These dietary patterns were differentially associated with nutrient intake. Regardless of diet group, we observed both, more health-conscious patterns as well as less favorable patterns. The dietary patterns “Vegetables and fruits” (vegetarian and omnivore), “Meat alternatives and potatoes” (vegetarian), “Vegetables and legumes” (vegan), “Fruits and convenience food” (vegan) and “Flexitarian” (omnivore) correlated with higher micronutrient density and therefore seem more health-conscious. In contrast, the patterns “Animal foods” (vegetarian), “Refined carbohydrates” (vegan), “Meat alternatives and juices” (vegan), “Dairy products” (omnivore), “Meat and convenience food” (omnivore) and “Refined grains and juices” (omnivore) were associated with a less favorable micronutrient profile and/or lower intakes of potentially critical nutrients.

Many studies using a posteriori dietary pattern analysis found various dietary patterns among omnivore children and adolescents in Western countries [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37]. Similar to the present study, these studies were able to identify at least one beneficial pattern considered as healthy, which was characterized by high consumption of unprocessed or minimally processed plant-based foods, such as vegetables, fruits, legumes, whole grains, or nuts [22,23,24,25, 27,28,29,30,31,32,33,34]. In addition, consistent with the VeChi Youth Study, less favorable dietary patterns were identified, that focus on energy-dense and highly processed foods, such as fast food, convenience foods, snacks, sweets or sugar sweetened beverages [23,24,25,26,27, 29, 30, 32, 34, 35].

The two predominant dietary patterns in the omnivore diet group, “Flexitarian” and “Vegetables and fruits”, were both characterized by a high intake of plant-based foods, including meat- and dairy alternatives. This suggests that plant-based dietary patterns are also common among children and adolescents classified as “omnivore”. This would be consistent with the results of several studies that found increasing popularity of flexitarian diets in European countries [2, 56, 57]. Since current research [58] as well as our results show substantial differences between flexitarians and traditional omnivores in terms of food selection, nutrient intake and nutrition status [58], a differentiation between flexitarian and traditional omnivore diets seems to be useful.

To date, a posteriori dietary patterns within vegetarian or vegan diets have only been studied among adult vegans [38, 39], but the dietary patterns in these studies show certain similarities with the vegan dietary patterns found in the VeChi Youth study. In a British study with 129 adult vegans (87% female, age 18–64 years) Gallagher et al. were able to identify four different vegan dietary patterns using PCA (based on 20 food groups), the patterns „Convenience“, „Health conscious“, „Unhealthy“ and „Traditional vegan“ [39]. In a recent Austrian study with 516 vegans (mean age (SD) 28.0 (7.7), 85% female), Haider et al. identified two dietary patterns using PCA (based on 18 food groups), a “Convenience” and a “Health-conscious” pattern [38]. The “Vegetables and legumes” pattern in the VeChi Youth study shows clear parallels with the “Traditional vegan” pattern found in British vegans [39] and the “Health Conscious” pattern in Austrian vegans [38]. All three dietary patterns focus on unprocessed plant foods, especially vegetables, fruits, and potatoes. In accordance with the VeChi Youth Study, Gallagher et al. and Haider et al. also extracted more unfavorable dietary patterns [38, 39]. In particular, the “Refined carbohydrates” pattern in the VeChi Youth study shows parallels with the unhealthier “Convenience” patterns found by Gallagher et al. [39] and Haider et al. [38], as all of these patterns are characterized by higher intake of vegan convenience foods, snacks, sweets and ultra-processed food items. These similarities between our extracted dietary patterns and the vegan dietary patterns in other studies confirm the results of our pattern analysis and show that comparable vegan dietary patterns can be found in vegan children/adolescents and adults.

Unfavorable vegan dietary patterns are associated with a lower intake of legumes, whole grains, nuts, and seeds. However, sufficient intake of these food groups is of particular importance to ensure adequate supply of potentially critical nutrients (such as protein, iron, zinc, calcium, and Vitamin B2) in a vegan diet [18, 19]. In the VeChi Youth Study, the vegan subgroup was found to have the highest overall consumption of these food groups [42]. However, our analysis showed that legumes did not occur equally in all vegan dietary patterns. Only the “Vegetables and legumes” pattern was associated with higher legume consumption, whereas the “Meat alternatives and potatoes” pattern was characterized by lower consumption. Thus, with increasing consumption of meat alternatives, legumes appear to be neglected as a protein source. Higher consumption of whole grains, nuts, and seeds did not distinguish any of the identified patterns, but the “Refined carbohydrates” and “Meat alternatives and juices” patterns were associated with lower consumption of the two food groups. More unfavorable nutrient correlations were also found for these two dietary patterns, once again highlighting the importance of these food groups in providing critical nutrients.

Many of the associations between dietary patterns and energy and nutrient intake in the VeChi Youth study can also be found in other studies with omnivore children and adolescents [22, 25, 30, 37], supporting the findings of our analysis. Across these studies a posteriori dietary patterns considered healthier were associated with a more beneficial nutrient profile [22, 25, 30, 37]. As the vegetable and fruit-rich patterns in the VeChi Youth Study, the “Health conscious” or “Traditional/health-conscious” dietary patterns identified in 3-9-year-old children [30] and 13-year-old adolescents [37] in the Avon Longitudinal Study of Parents and Children (ALSPAC), were correlated with higher energy-adjusted intakes of dietary fiber and most micronutrients [30, 37]. Ambrosini et al. [22] also found these correlations in 13-year-old adolescents in the Australian Raine study. In addition, there was an inverse association between the “Healthy” pattern and energy-adjusted saturated fat intake [22], as was also shown in all three dietary groups in the VeChi Youth study. The same correlations between the above-mentioned nutrients and the identified “Healthy” patterns were also shown by Richter et al. [25] in 12- to 17-year-old girls and boys in the EsKiMo.

Similarly to other studies, the dietary patterns in the VeChi Youth study considered as unhealthier were associated with a more unfavorable nutrient profile. The vegan pattern “Refined carbohydrates” in the VeChi Youth Study showed particularly high agreement in nutrient correlations with the unfavorable dietary patterns “Snacks/sugared drinks” in the ALSPAC [37] and “Western” in the Raine study [22]. The three patterns were similar not only in the high factor loadings for sweets, snack foods, and sugar-sweetened beverages, but also correlated with higher total sugar intake and lower intakes of protein, dietary fiber, and most micronutrients (especially magnesium, iron, and zinc) [22, 37].

Patterns with individual food groups, which are classified as unfavorable due to their high degree of processing, however, cannot be clearly classified as unhealthy. In the present work, e.g. the dietary pattern “Fruits and convenience food” was the only vegan pattern to show significant positive correlations with the intake of the critical nutrients protein, zinc and calcium. This may due to the fact, that several plant-based dairy alternatives are fortified with minerals, e.g. calcium (27). In addition, negative correlations with the intake of energy, added sugars and saturated fatty acids were observed. This underlines that vegan convenience foods or more highly processed plant-based alternative products are not to be classified as unfavorable per se, but, depending on their ingredients and nutrient enrichment, can contribute to an adequate intake of critical nutrients. With increasing consumption of meat alternatives, legumes appear to be neglected as a protein source. Thus, our results can contribute to the debate about the composition of meat- and dairy alternatives and whether they should be equated to other highly processed foods [59,60,61,62].

The strengths of the VeChi Youth Study include the detailed dietary survey using 3-day weighed dietary records, in which fortified foods as well as plant-based convenience foods and alternative products were taken into account. Another strength is the large study population of vegetarian and vegan children and adolescents, including an omnivore control group, with equal age distribution and matching sociodemographic characteristics of the three diet groups.

However, the analysis also has some limitations. Due to the above-average socioeconomic status of the study participants and the regional focus in North Rhine-Westphalia, the VeChi Youth Study is not representative and the results can therefore only be generalized to a limited extent. However, the high socioeconomic status as well as the high proportion of urban residents in the study population correspond to the known sociodemographic characteristics of vegetarians and vegans [3]. In contrast, the rather high socioeconomic status in the omnivore group is probably due to selection bias [42]. Thus, in a representative study, potentially larger differences between dietary patterns and associations with health-related outcomes would be detectable.

The sample size for the pattern analysis was relatively small, as the power calculation was carried out for a different outcome, which may also affect the representativeness of our results. Overall measure of sampling adequacy according to KMO indicated insufficient suitability of the data for PCA, however the observed patterns seem plausible for vegan, vegetarian and omnivore diets. In addition, in order to obtain reliable results in the PCA, it is recommended that the number of study participants should be at least five times as large as the number of analyzed variables (food groups) [50, 63], which we were able to fulfil in our analyses. Furthermore, Bartlett test for sphericity showed a significant result in all three diet groups (P < 0.001), which indicates the adequacy of the data for PCA. Due to the small sample size, it was not possible to perform stratified analyses in addition to stratification of the diet group and thus to determine age- or sex-specific differences in dietary patterns. Therefore, we could not rule out the possibility that different age- and sex-distributions in the diet groups may have influenced our results. However, prospective studies with omnivore populations have found consistent dietary patterns from childhood to adolescence [24, 30, 64] and very similar patterns in younger girls and boys [27, 28, 32]. In adolescents, on the other hand, different dietary patterns have been found according to sex [25, 26, 65]. The extent to which vegetarian or vegan children and adolescents show age- or sex-specific differences in their dietary patterns is not known and should therefore be investigated in future studies.

With regard to the dietary assessment, it should be mentioned that the record duration of 3 days might not be sufficient for recording the participants’ habitual diet. However, more records days are discussed to may cause a decline in the quality of information [66]. As the participants were free to choose the days of the dietary records, it cannot be ruled out that the records were also completed on uncommon, unrepresentative days, such as holidays. Moreover, the dietary records were completed on three consecutive days, but at different times of the year. It is therefore possible that seasonal differences in the participants’ diet al.so had an impact on the results.

A general limitation of PCA is subjectivity, especially when aggregating food groups, determining the number of patterns to be derived, and labelling the components [67]. The population specificity also makes the identified dietary patterns difficult to reproduce. Therefore, to increase comparability with other studies, a methodological approach analogous to that used in other studies was adopted in extracting the dietary patterns [25, 26]. Furthermore, dietary patterns extracted by PCA are only able to explain a small proportion of variance in diets [68]. In the present evaluation, the individual dietary patterns extracted based on 17, 21, and 22 food groups could each explain only 7.4–13.0% of the variance in food group intake. However, these proportions of variance are consistent with the results of other studies in children [27, 35] or adolescents [34, 36].

Concerning the investigated nutrient correlations, it should be mentioned that not all potentially critical nutrients in vegetarian or vegan diets could be considered in the analysis and thus, for example, information on the intake of iodine, selenium and long-chain omega-3 fatty acids was missing. For iodine, this is due to the fact that table salt as the main source of iodine was not quantitatively assessed. Data on selenium are not available in German food composition tables and databases from other countries are not useful as selenium content in foods depends on geography. Sources of long-chain omega-3 fatty acids in vegan diets are fortified foods and supplements, but the latter are not included here. Supplementary studies with investigations on these nutrients are therefore desirable. Moreover, the sole investigation of nutrient intake does not allow any conclusion to be drawn on the actual nutrient status of the study participants. The inclusion of biomarkers for nutrient status and longitudinal studies could provide further insights into the relationship between dietary patterns and nutritional status.

Conclusion

The present study shows different dietary patterns that vary in their contribution to nutrient intake, not only in omnivores but also in vegetarian and vegan children and adolescents. Regardless of diet type, there appear to be both more nutritionally favorable plant-based dietary patterns associated with higher micronutrient density and more nutritionally unfavorable dietary patterns with a greater focus on energy-dense and/or more highly processed foods. In the vegetarian and vegan diet group, the nutrient correlations suggest that certain dietary patterns characterized by a lower intake of legumes, whole grains and/or nuts and seeds are associated with a lower intake of critical nutrients. This emphasizes the importance of these food groups for adequate nutrient intake in vegetarian and vegan diets.

There are many similarities between the vegetarian, vegan and omnivore diet groups in terms of the composition and nutritional properties of the patterns identified. This indicates that certain dietary patterns occur independently of the form of diet. A general nutritional evaluation of the diet based on the conventional classification into “vegetarian”, “vegan” or “omnivore” is therefore only of limited use. Rather, a differentiation between nutritionally more favorable and less favorable dietary patterns within the form of diet seems appropriate. Prospective studies are desirable to verify the consistency of the identified dietary patterns and to assess their long-term preventive relevance.