We recruited the vegan participants (n = 22) through an advertisement published in the Finnish Vegan Association’s monthly newspaper and via an online discussion forum. Vegans were self-defined. The selection criteria were as follows: the participants had to (1) have followed a vegan diet for at least a year, [2] be between 18 and 50 years of age, [3] be apparently healthy, and [4] be nonusers of regular medications (except oral contraceptives or hormone replacement therapy). Non-vegetarian participants (n = 19) were recruited using the same media and criteria, except the dietary consumption. Further, we matched the non-vegetarian participants by age and sex. The study received the approval of the Ethical Committee of Kuopio University Hospital (69//2011). All participants provided their written consent to participate this study.

We assessed the consumption of foods at the baseline with a three-day food record using household measures. The food record sheets were blank and consisted of only an example menu. The participants received instructions on how to complete this record, and a dietician checked the completed food records. The participants also filled out a questionnaire concerning their long-term eating habits, including nutrient supplementation. The question concerning the consumption of a special diet included three vegetarian options (vegetarian, lacto-vegetarian and vegan) where vegetarian option was regarded as a semi-vegetarian option. The duration of the present diet was asked on a separate question.

We calculated the food and nutrient intakes using the Finnish Nutrica® software version 2.5, which is compiled mainly using Finnish values for the nutrient composition of foods. These calculations take into account the loss of vitamins during food preparation. This software was developed at the Research Centre of the Social Insurance Institution of Finland [27]. We obtained the nutrient compositions of the foods in the Nutrica® software from analyses carried out in the 1990s. The Nutrica® software contains the latest data on the vitamin contents of fruits and vegetables; furthermore, it provides a comprehensive database comprising more than 1,300 food items and dishes and 30 nutrients.

Moreover, we grouped the food items as follows: whole grain products (including rye products), rye products, rice, pasta, vegetables, root vegetables, potatoes, pulses, nuts, fruits, berries, fruit juices, berry juices, coffee, and tea. Specifically, the whole grain products included different breads, flakes, wheat bran, germ, and muesli products and excluded refined flour products. The rye products included different rye breads, rye flour, flakes, bran, and malt. The rice and pasta groups included both whole grain and refined products, which we did not include in the whole grain products. The vegetable group included all fresh and frozen vegetables and excluded pickled and canned vegetables. The root vegetables (the most commonly consumed root vegetables in Finland are carrots, Swedish turnip [swede], turnip, and beetroot) included all roots, except potatoes, which constitute a separate variable. The fruits included fresh, canned, and dried fruits and fruit nectars, while juices belonged to fruit juices. The berries included all fresh and frozen berries, crushed lingonberries, and lingonberry jam, which, in Finland, are usually prepared without sugar. The coffee and tea variables included all coffee and tea drinks.

We collected fasting peripheral venous blood samples and centrifuged them to separate the plasma and the serum, and we stored them at -70°C before the analysis. Furthermore, we collected urine samples 24 h prior to phlebotomy and divided them into aliquots. For one of the subjects, only a spot urine sample was available. We carried out the hematological and chemical analyses at Vitalab (hematological data), Helsinki University Central Hospital (folate, vitamin B12, vitamin D, and homocysteine), and National Institute for Health and Welfare (α-tocopherol, β-carotene, selenium, iodine, fatty acid profile, and polyphenols).

We used Siemens Advia 2120i and Advia 120 automatic cell counters to conduct the hematological analyses, which also involved the use of photometry to assess the participants’ Hb levels. The serum total cholesterol, HDL-cholesterol, and triglycerides were determined enzymatically using an Abbott Architect c8000 using commercial reagents from Abbott Laboratories (Abbott Park, IL, USA). LDL-cholesterol was calculated using the Friedewald formula.

To analyze the serum vitamin B12 and folate concentrations, we employed Abbot Architect i2000 and i2000SR analyzers (Abbott Diagnostics, Abbott Park, IL, USA) using two-step assays with automated sample pretreatment. Furthermore, we utilized Liquid Stable (LS) 2-part homocysteine reagent (Axis-Shield Diagnostics Ltd., Dundee, UK) with a Hitachi Modular P800 Chemistry analyzer (Hitachi Ltd., Tokyo, Japan) to analyze plasma homocysteine. We conducted all analyses according to the instructions of the manufacturers.

To analyze vitamin D, we used a previously unpublished method. 25-OH-D2 and 25-OH-D3 were obtained from Fluka; deuterium labelled internal standards (ISs) and 6,19,19-D3-25-OH-vitamin D3, from IsoSciences; and 26,27-D6-25-OH-vitamin D2, from Medical Isotopes Inc. We prepared stock solutions of 25-OH-D2, 25-OH-D3, and ISs in ethanol and diluted into methanol/water (70% v/v). We checked the standard solution concentrations by spectrophotometry (molar absorption coefficient: 18200 for both metabolites in ethanol at 265 nm). The imprecision for 25-OH-D2 and 25-OH-D3 was not higher than 7.0 and 7.1%CV, respectively.

To 150 μl of calibrator, serum, or quality control sample, we added 30 μl of IS working solution. After precipitating the proteins with 150 μl of methanol, we extracted the analytes with 3 mL of hexane. We dried the upper organic phase under nitrogen and dissolved the residue in 150 μl of 700 mL/L methanol/water. We injected 25 μl on an LC-MS/MS system comprising an API 4000 triple quadrupole mass spectrometer (AB Sciex) and an Agilent series 1200 HPLC system with a binary pump. To perform the separation, we employed a Discovery HS F5 column (2.1 × 100 mm, 3 μm; Waters, 30°C, flow rate of 300 μl/min). We employed a linear methanol/water gradient (0 min, 70% methanol; from 6.7 min to 10 min, 90% methanol; and from 10.2 min to 15 min, 70% methanol). The column was directly connected to the electrospray ionization probe.

We detected 25-OH-D₂, 25-OH-D₃, and ISs in the multiple reaction monitoring mode (25-OH-D₂, m/z 413 to m/z 395; 25-OH-D₃, m/z 401 to m/z 383; and ISs, m/z 405 to m/z 386 and m/z 419 to m/z 401). We obtained the data and processed it using the Analyst Software (Ver. 1.4, Sciex). All results were generated in the positive-ion mode (entrance potential: 10 V, declustering potential: 65 V, collision cell exit potential: 11 V, collision energy potential: 14 V). The front-end electrospray settings for the MS/MS ionization source were as follows: curtain gas, 25; GS1, 30; GS2, 30; CAD, 6; probe temperature, 350°C; and ion spray voltage, 5500 V. For all MS/MS experiments, we optimized all mass calibration and resolution adjustments on both the resolving quadrupoles using a polypropylene glycol solution with an infusion pump. We performed collisionally activated decomposition MS/MS with nitrogen as a collision gas.

We analyzed plasma α-tocopherol and β-carotene by reverse-phase HPLC, as previously described [28]. The imprecision between series for α-tocopherol was 3.7% and that for β-carotene, 5.2%. We analyzed plasma selenium using an electrothermal atomic absorption spectrometric method modified from [29] and [30]. We diluted a 50 μL aliquot of plasma fivefold with a solution containing 530 mg ascorbic acid and 0.2% Triton X-100. Then, we injected 15 μL of the diluted sample + 4 μL of 0.79 mmol/L Pd in 120 mmol HCl into a graphite tube. The instrument used was Perkin-Elmer 600 with Zeeman background correction. The imprecision between series was 4.4%, and the mean inaccuracy from analyzing serum standards (ClinCheck, Recipe, Munich,Germany) was 2.3%. We analyzed the urinary iodine concentrations using a microplate method [31]. The between series imprecision was 9%. Participation in the EQUIP (CDC, Atlanta) urinary iodine quality program showed a mean bias of -5.7%.

We utilized gas chromatography to determine serum fatty acids. We extracted the serum using dichloromethane-methanol (2:1, v:v) [32] and transesterified the fat yield with dry methanol (5% H₂SO₄) [33]. We analyzed the methylated fatty acid composition using an Agilent 6890 gas chromatograph (Hewlett Packard, Palo Alto, CA, USA) with a split injector and hydrogen as the carrier gas. We employed a DB-225 column (length: 30 m, I.D.: 0.32 mm, phase layer: 0.25 mm; Agilent J&W GC, Palo Alto, CA, USA). The temperature was varied from 160°C to 230°C. We normalized the percentage composition of fatty acid methyl esters to 100%. The between series variability of this method is 2%–7% for fatty acid peaks over 1% and 6%–18% for smaller peaks.

We determined the serum polyphenols, including genistein and daidzein, by gas chromatography and mass spectrometry after enzymatic hydrolysis using a modification of a previously described method [34]. The between series imprecision of polyphenols was 5%–9%.

We expressed the data as mean ± SD (min-max); only the biochemical data was expressed as medians (with 25^th and 75^th percentiles). Differences between the groups were analyzed using the Mann-Whitney U test. Due to possible bias in statistical significances between the groups we made Bonferroni corrections for each group of parameters. Corrected threshold for statistical significance was calculated separately for food [30], nutrient [33], biochemical [30] and fatty acid [20] parameter group. We also made power calculations to find out how sensitive our data is to notice the possible difference between the groups in key nutrients such as vitamin B12, vitamin D, and iodine. We performed all statistical analyses using PASW Statistics 18 (2009, SPSS Inc.).

The vegans had higher daily intakes of legumes (p ≤ 0.001), tofu and soy flour (p<0.001), and margarine (p<0.001) than the non-vegetarians did. According to Table 2, vegans consumed no meat and fish, and little if any amount of dairy or other animal-derived products. There were no observable statistically significant differences in the consumption of any other foods between the groups after adjusting the data with the new threshold for statistical significance obtained using Bonferroni correction for multiple comparisons (p<0.0015), Table 2. The vegans’ daily mean consumption of protein sources was as follows: soy milk (70 ± 125 g), soy yogurt (101 ± 187 g), soy groats (68 ± 70 g), wheat protein seitan (19 ± 36 g), and falafel (81 ± 90 g). Most vegan subjects consumed enriched plant food items such as soy and oat drinks daily.

There was no statistically significant difference in total energy intake between the vegans and the non-vegetarians; however, there were differences in nutrient intakes and essential nutrients. Specifically, the vegans obtained less, saturated fat, cholesterol, niacin, selenium, vitamin D, and vitamin B12 (p<0.001). Also the percent amounts of protein, carbohydrates, and SAFA intakes differed significantly between the groups (p<0.001), Table 3.

The participants were healthy according to routine clinical chemistry measurements, except for three vegans who suffered from borderline anemia. The vegans also had lower ferritin stores, as shown in Table 4.

The vegans exhibited lower serum concentrations of vitamin B12, vitamin D (25-hydroxyvitamin D2 and D3) total cholesterol and LDL cholesterol, as well as plasma concentrations of selenium, β-carotene, and α-tocopherol compared to the non-vegetarians, as shown in Table 4. However, when adjusted to the new threshold of statistical significance (p<0.0016) only serum concentrations of vitamin D (25-hydroxyvitamin D2 and D3), and plasma concentrations of β-carotene, and selenium, and urinary output of iodine were significantly different between the groups. The median urine iodine concentrations were under the reference concentration in both study groups. We observed marked inter-individual variation in the intakes of energy and protein and in the concentrations of nutritional biomarkers.

The serum fatty acid profiles were markedly different in the vegans as compared to the non-vegetarians, as shown in Table 5. The vegans had lower proportions of saturated fatty acids and higher proportions of (n-6) PUFA (P < 0.001). In particular, the groups were differentiated by the absence of conjugated linoleic acid (CLA) in the serum of the vegans. The proportions of (n-9) monoenes (p<0.01) and linolenic acid (18:3 n-3) were higher in the vegans’ serum. The proportions of EPA (p<0.001) and docosahexaenoic acid (DHA) (p<0.001) were clearly lower in the vegans’ serum than in that of the non-vegetarians; however, the groups did not differ with regard to docosapentaenoic acid (DPA) (22:5 n-3). The serum concentrations of the isoflavones genistein (p<0.001) and daidzein (p<0.001) were considerably higher in the vegans than in the non-vegetarians.

Some health-related and nutritional measures were more favorable in vegans than in non-vegetarians. Most importantly, the serum total cholesterol was 20% and LDL cholesterol was 25% lower in the vegan group than in the non-vegetarian group. Furthermore, vegans showed a more favorable fatty acid profile and higher serum concentrations of certain polyphenols compared with the non-vegetarians. These findings were likely the result of high consumption of rapeseed oil and margarines as well as soy and rye products.

The vegans consumed relatively small amounts of fruit, berries, nuts, and root vegetables, which was the likely cause of the lower serum concentrations of β-carotene (p = 0.001) and α-tocopherol (p = 0.003) compared to the non-vegetarians. However, after calculating the ratio of serum β-carotene and α-tocopherol to cholesterol concentration, and when adjusted to the new threshold of statistical significance obtained in Bonferroni calculation for multiple comparisons (p<0.0016) the differences in these antioxidant nutrients were not statistically significant. The poorer antioxidant vitamin status of vegans disagrees with the findings of earlier studies [35]. This is likely because the non-vegetarian subjects of this study were health conscious, as shown by their high consumption of different vitamin and mineral supplements as well as fruits and berries.

Despite the use of nutritional supplements, the serum vitamin B12 concentrations in the vegans were lower compared to the non-vegetarians (p = 0.002); however, only 5% of vegans had serum vitamin B12 concentration below 140 pmol/L. It therefore appears that the consumption of vitamin B12 supplements, which 91% of the vegan subjects consumed, maintained their serum vitamin B12 concentrations within the reference limits. The onset of deficiency symptoms such as neuropsychiatric disorders and megaloblastic anemia usually occurs in 5–10 years when the serum vitamin B12 concentration is below 150 pmol/L [36].

The serum total concentration of vitamin D (25-hydroxyvitamin D2 and D3) was 34% lower in the vegans than in the non-vegetarians. However, the vegans had higher concentrations of 25-hydroxyvitamin D2 (p<0.001). The fraction of subjects having serum vitamin D concentration >75 nmol/L, which is the level proposed by some researchers to be optimal for preventing adverse health conditions [37], was 10% in vegans and 78% in non-vegetarians. In addition, more vegans had a serum vitamin D concentration ≤50 nmol/L as compared to the non-vegetarians (24% vs. 6%). The reasons for the marginal vitamin D status are presumably neglecting supplementation (23% of vegans), irregular supplementation, and, possibly, the time of sampling. 25-hydroxyvitamin D3 (calcidiol) concentrations are typically lowest during the winter [38]. Similar lower calcidiol concentrations were reported in Finnish, British, and Vietnamese vegans. [8, 9, 20, 21]

All vegan subjects and 91% of the non-vegetarian subjects had iodine concentrations lower than the WHO’s limit for mild iodine deficiency (<100 μg/L urine). These data indicate that iodine intake may be insufficient in the Finnish population but particularly so in vegans, who do not consume milk products, the main source of iodine in many countries. Previously, goiter caused by iodine deficiency was common in Finland. However, after the fortification of table salt and cattle feed with iodine started some fifty years ago, iodine-deficiency-related goiter was eradicated. Today, the consumption of iodized table salt has decreased, partly because the food industry does not use iodized salt. Therefore, recommendations regarding iodine intake are not met by the general population [12]. Previous studies in vegans have also reported low urinary excretion of iodine. [6, 22, 23]

The serum selenium concentrations were lower in vegans than in non-vegans, however, on the whole, the values were similar to those found in countries that do not add selenium to fertilizers. It should be noted that Finland is the only country in the world that uses this strategy for supplementing the population with selenium [39]. The selenium intake was above the current nutrition recommendations [10] in both groups. The difference between groups is likely because dietary selenium is mainly obtained from animal products, which make up over 70% of the selenium intake in Finland [39].

Compared to the non-vegetarians, the proportions of C15:0, C17:0, and CLA, obtained mainly from milk products, were negligible in the vegans, indicating strict compliance with the vegan diet. The percentages of EPA and DHA of all fatty acids were respectively 0.6% and 0.9% in the vegans, and they were clearly lower than in the non-vegetarians. These differences were expected because vegans do not consume fish or fish oil products. However, the observed proportion of EPA in the vegans was still higher than expected. These results support the view that linolenic acid (LNA) is converted to EPA in humans. One may regard the vegans in this study as a high LNA population, as they consumed relatively high amounts of rapeseed oil, a common vegetable oil in the Nordic countries and a rich source of LNA. In countries consuming other types of vegetable oils, vegans would likely show even lower proportions of EPA and DHA in plasma. On the other hand, it should be noted that this issue is not straightforward, because linoleic acid (LA) and LNA compete for enzymes involved in fatty acid metabolism. A previous study among Kenyan Maasai [40] showed that despite a negligible intake of EPA and DHA, the proportion of DHA in red blood cells (RBCs) was no less than half that of a German sub-cohort. The authors speculated that a low intake of LA could also be advantageous and favor the endogenous conversion of LNA to DHA at a state of competition between n-3 and n-6 fatty acids.

Few studies have investigated the nutritional status of long-term vegans; therefore, our study gives important new information about these issues. However, this study has some limitations. Most importantly, the sample size was small and the results should therefore be confirmed in a bigger and more representative sample. Vegans were also self-defined, however, their compliance was confirmed by analyzing blood fatty acid profile, and no traces of foods of animal origin were noticed. This sample represented rather well educated young adults whose primary reasons for adopting a vegan diet were animal welfare and environmental concerns rather than beneficial health effects. The dietary intake was assessed by food records, which are relatively accurate to population mean dietary intakes. In addition, we studied various nutritional biomarkers. The nutrient intake was calculated only from foods and drinks excluding dietary supplements; however, this was compensated when relevant nutritional biomarkers were analyzed. In addition to dietary habits, many other lifestyle factors such as physical activity, alcohol consumption, and smoking may have their impact on nutrition, and hence confound the data. However, no differences in these parameters were noticed between the groups. From a statistical viewpoint, performing multiple tests may result biased significances. In order to eliminate the possible bias we made Bonferroni calculations and used more specific thresholds for statistical significances in each group of parameters. Taking multiple comparisons problem into account, the observed power between group differences were 0.47 for iodine, 0.43 for vitamin B12 and 0.91 for vitamin D.