The Effect of A 4-Week Vegan Diet on Adipokine Levels in Healthy Male and Female Participants: A Randomized-Controlled Trial

Background Vegan diet (VD) is reported to show benecial health effects including cardiovascular and anti-inammatory protection, but the underlying mechanisms remain unclear. We hypothesized that adipokines, a special type of cytokine produced by the white adipose tissue with known effects on metabolism and the immune system, may contribute to the observed anti-inammatory effects of VD. Methods A parallel group interventional trial was designed to evaluate the effect of VD compared to meat-rich diet (MD) on serum levels of two central adipokines, leptin and adiponectin. Fifty-three healthy, omnivore participants (62% female, average age 31 years and BMI 23.1 kg/m 2 ) were randomly assigned to VD or MD for 4 weeks. Results End value comparison between VD group and MD group showed a signicantly lower level of adiponectin in the MD group (11.6 vs. 15.5 µg/mL, p = 0.025) indicating a moderate effect size (Cohen’s d = 0.524). Participants’ sex affected adipokine levels requiring a separate analysis of male and female participants. Leptin was increased by MD only in male participants (p = 0.019) whereas adiponectin remained stable. Female participants in VD group showed higher adiponectin levels at the end of trial (compared to VD-baseline, p = 0.023, as well as compared to MD group, p = 0.015). The end concentration of adiponectin depended on diet in female participants (p = 0.010). Conclusion The results of our trial suggest that plasma concentration of leptin and adiponectin do not explain the immunomodulatory potential of VD in healthy participants, but it appears that diet modies adipokine levels in a sex-specic manner.

Furthermore, low adiponectin concentrations are associated with metabolic syndrome, obesity-linked insulin resistance and atherosclerosis [4,9]. Several publications examined the in uence of dietary intake on the concentration of leptin and adiponectin in plasma, yet these works did not investigate the effect of diet composition except for caloric intake [10][11][12]. One recent study indicated that long-term vegan and vegetarian diets might affect leptin concentrations of normal-weight individuals, speci cally leading to lower leptin concentrations than average [13]. The comparison of two 12-week lasting protein-rich vegan diet with a control diet in rats revealed higher levels of adiponectin in the vegan group [14]. Plant-based diets have been reported to mediate anti-in ammatory effects [15][16][17]. Despite the established interplay of metabolism and immune function, the underlying mechanism of these diet-induced anti-in ammatory effects have been insu ciently studied. In 2017, we started a clinical trial aiming to broadly map the in uence of a vegan diet on the immune system to clarify the underlying mechanism leading to an antiin ammatory effect of a vegan diet [18,19]. Based on previous studies by us and others and the essential role of metabolism in immune function, we addressed this question in a randomized clinical trial and hypothesized that the introduction of a vegan diet might lead to lower levels of leptin and higher levels of adiponectin, a combination that would be favorable for adequate immune function.

Methods
A monocentric, controlled, randomized trial with parallel group design with healthy participants was performed at the Center for Complementary Medicine, University Medical of Freiburg, Germany, between April and June 2017. Methods have been described previously in detail [18,19]. The analysis of adipokines is a subgroup analysis of a clinical trial evaluating the effect of VD on immune system, Vitamin B12 metabolism and gut microbiome.
Criteria of inclusion were: Healthy, omnivorous, normal weight subjects, between 18 and 60 years of age, no regular intake of a medication as well as no clinically relevant allergies. Participants declared having no history of an eating disorder, participation in another clinical trial and blood donation in the last 4 weeks before the start of this trial. Abuse of drugs, nicotine or alcohol was not allowed. Participants had to be able to speak and understand German and to complete a nutritional protocol.
Newspaper announcement and bulletins were used for recruitment. After phone call, eligible subjects were invited for a personal visit to check eligibility criteria in detail. Precondition for inclusion was a written informed consent. After inclusion, each participant received an extensive training to keep his/her own balanced mixed diet according to the recommendations of the German Nutrition Association (DGE) [20]. After this one-week-lasting run-in phase, fasting baseline parameters were taken early in the morning. Afterwards, participants were randomly assigned to either a meat-rich (> 150 g of meat per day; any meat of their choice) or a strict vegan diet for four weeks. Every subject received another extensive training on the assigned diet and detailed written information and a recipe book. Fasting parameters were taken again after four weeks early in the morning. Participants nished the trial after blood sampling. Participants were free to choose their food within their assigned diet, but lling out a weekly nutritional protocol was mandatory for all participants. Results of nutritional protocols were used to evaluate diet

Statistical analysis
For the primary purpose of the study it was calculated that 48 participants (24 of each group) would be necessary to detect a statistical difference of p < 0.05 considering a statistical power of 80% and a hypothesized large effect size [18,19]. Analysis of adipokines was an exploratory aim. Therefore, to evaluate the effect of Time, Diet and Sex and their interactions on the levels of adipokines, with an estimated large effect size, at a signi cant alpha level of 0.05, and number of participants equal to 53 (males: 8 VD and 12 MD; females: 18 VD and 15 VD) a post hoc statistical power of 81% was calculated.
Data was entered and analyzed blinded for diet assignation via IBM SPSS (version 25.0). Baseline characteristics were evaluated by t-test, Mann-Whitney-U-Test and Fisher's exact test. Because some biochemical markers were not normally distributed, Mann-Whitney-U-test was used for comparison of group differences. Adjustment for baseline was considered by using ANCOVA. Multiple linear regression was performed to evaluate dependency effects of age, BMI (end concentration), weight changing (weight change of more than one SD (± 1.3 kg) compared to baseline weight), sex and diet on end concentration of adipokines. A three-way mixed ANOVA was used for evaluating between-subject effects of diet (VD, MD) and sex (male, female), within-subject effect of time (baseline, end), and their interactions. Correlation analysis was calculated using Spearman-Rho. Bonferroni correction was performed for all tests. Power analysis was calculated using G*Power (version 3.1.9.7).

Results
The ow of the study and baseline characteristics of the chosen participants are herein described in a condensed form as they were previously published [18,19]. Out of 150 interested persons 103 were invited for an interview after checking eligibility criteria by phone-call. Sixty-one interested persons ful lled all criteria for inclusion and started the run-in phase. Eight of these had to be excluded before randomization because of consent withdrawal or acute illness. Overall, 26 participants were allocated to VD and 27 participants were allocated to MD for four weeks (Table 1). All participants completed the study as per protocol. At both time points (baseline and end), BMI did not differ signi cantly between the groups. The intake of energy was similar in both groups (Table 1) and was within the recommendations set forth by the DGE for healthy adults [18][19][20]. The results of nutritional protocols were previously published in detail [18]. Comparison of baseline and end concentrations of leptin as well as of adiponectin between VD group and MD group did not reach statistical signi cance (Table 2). However, end value comparison between VD group and MD group showed a signi cantly lower level of adiponectin in the MD group (11.6 vs. 15.5 µg/mL, p = 0.025, Table 2). The effect size was moderate (Cohen's d = 0.524).
The results of mixed ANOVA presenting main and interaction effects are shown in Table 2. Results of leptin and adiponectin measurement were signi cantly affected by participants' sex ( Table 2). Table 2. Concentration of leptin and adiponectin before and at the end of the trial in and between both groups as well as main and interaction effects (MD = Meat-rich diet, VD = Vegan diet, SD = Standard deviation, *Wilcoxon signed-rank-test, + Mann-Whitney-U-Test, °Mixed ANOVA, † adjusted for baseline).

Results of multiple linear regression
The effect of multiple predictors on end concentration of adipokines is shown in Table 3. The end concentration of leptin depended on participants' end BMI.
Participants' sex affected the concentration of leptin and adiponectin as male participants had signi cantly lower levels of both adipokines compared to female participants (Table 3 and 4). (*"Changing of weight" is de ned as weight change of more than one SD (± 1.3 kg) compared to baseline weight; β = standardized beta coe cient ). In uence of diet on adipokine levels of female participants Comparison of baseline and end concentration of leptin between VD and MD female subgroups did not reach statistical signi cance (Table 5 and Figure 1, Panel A). In female participants, the end concentration of leptin correlated signi cantly with the end BMI (r = 0.742, p < 0.001, Figure 2, Panel A).
The end concentration of leptin depended on female participants' end BMI (p < 0.001, Table 5). Comparison of baseline and end concentration of adiponectin showed a signi cantly higher concentration in the female VD subgroup at the end of the trial (p = 0.023, Table 5 and Figure 1, Panel B). Furthermore, the end concentration of adiponectin was signi cantly higher in VD than in MD (18.8 vs. 12.7 µg/mL, p = 0.001). The end concentration of adiponectin depended on diet in female participants (p = 0.010, Table 6). The results of mixed ANOVA revealed an interaction effect between time and diet (time x diet p = 0.004) regarding the adiponectin concentration (Table 5). In female participants, the end concentration of adiponectin did not correlate with the end BMI (r = -0.240, p = 0.178, Figure 2, Panel B). Leptin concentration increased signi cantly during the trial in MD group (p = 0.019, Table 7 and Figure 3, The results of mixed ANOVA did not show any main effects of time or diet as well as interaction effects between diet and time on leptin concentration in male participants (Table 7). In male participants, the end concentration of leptin correlated signi cantly with the end BMI (r = 0.596, p = 0.006, Figure 4, Panel A), but multiple linear regression did not reveal any dependency on end concentration of leptin (Table 8).
Adiponectin concentration remained stable during the trial in male participants (Table 7 and Figure 3, The results of mixed ANOVA did not show any main effects of time or diet as well as interaction effects between diet and time on adiponectin concentration in male participants (Table 7). In male participants, the end concentration of adiponectin did not correlate with the end BMI (r = -0.293, p = 0.211, Figure 4, Panel B). Multiple linear regression did not reveal any dependency on end concentration of adiponectin (Table 8), but the results are limited due to the poor test quality.   In uence of age on concentration of leptin and adiponectin The average age of participants was 31 years, and the median age of participants was 27 years. In order to be able to divide the comparison groups evenly, we initially used median value for further calculation.
As mentioned before (Tables 3, 6 and 8) multiple linear regression did not reveal any dependency of age and end concentration of adipokines. There was no correlation between age and adipokine concentration in male and female participants ( Figure 5 and 6).

Correlation of leptin and adiponectin with nutritional intake
All results of nutritional protocols were previously published [18,19]. Leptin concentration at the end of the trial did not show any correlation with nutritional intake of saturated, mono-or polyunsaturated fatty acids as wells as with total fat, cholesterol or total energy. We found a signi cant correlation of end adiponectin concentration with saturated fat intake (r = -0.364, p = 0.017), but not of mono-or polyunsaturated fatty acids as well as total fat and total energy. The intake of cholesterol was negatively correlated with the end concentration of adiponectin (r = -0.474, p = 0.001) showing differences in VD group (r = -0.535, p = 0.010) and MD group (r = -0.036, p = 0.876). We found no sex-speci c differences.

Correlation of adipokines with immune parameters and branched-chain amino acids
As we were able to previously show signi cantly lower levels of immune cells (leukocytes and monocytes) in VD compared to MD [19], we correlated the concentration of adipokines with these immune cells. Due to the immunomodulatory potential of adipokines we also performed a correlation of adipokines with C-reactive protein

Discussion
Adipokines are factors derived from white adipose tissue that regulate metabolism and immune function [22]. Leptin acts as a modulator of appetite whereas adiponectin sensitizes cells to insulin. The direct control of central metabolism by adipokines results in pro-and anti-in ammatory effects [22]. This study was driven by the hypothesis that the adipokines leptin and adiponectin contribute to the immunomodulatory effects observed in an earlier study upon intervention of healthy subjects with a balanced vegan diet. Our results suggest that leptin concentration is not modi able by short-term VD. Instead, a comparison of baseline and end concentration of leptin in MD participants revealed a signi cant increase of this adipokine in male participants. Adiponectin was higher in the VD group after the trial and this was not dependent on initial baseline concentrations. Further analysis of adiponectin changes within the VD group showed sex-dependency as only adiponectin of female participants differed signi cantly at the start and end of the trial. Female participants in VD group had a slight weight-loss, which might confound the increase in adiponectin after the trial since the difference might not be caused by diet, but by weight loss. Weight-loss is discussed to be related to changing of adipokine levels.
Published reports indicate effects of weight-loss on serum leptin, but the same has not been unequivocally seen with respect to serum adiponectin concentration [12]. Therefore, we performed a subgroup analysis excluding patients with a weight change of more than 1.3 kg. Interestingly, the difference of adiponectin between VD and MD remained statistically signi cant suggesting a not-weight related effect of diet on adiponectin levels in female participants. In male participants, who had no weight-loss during the trial, adiponectin remained stable, but comparability of male and female participants is limited.
Sex-dependent differences in adipokine levels are known for a long time [23][24][25], and the results of our trial also show sex-dependent differences leading to the performed subgroup analyses. The results of our and other studies emphasize not only the previously known fundamental difference of adipokine levels in males and females, but also a supposed sex-difference in alterability by diet or speci c nutritional components [26]. This is also supported by the results of Vučić Lovrenčić et al. as they observed differences of adipokine concentration in female vegetarians compared to female omnivores but not in male vegetarians compared to male omnivores [27]. Apart from sex-related differences, the general impact of speci c diets or of speci c nutritional components on adipokines levels of normal-weight human beings remains still unclear, since the majority of publications dealt with adipokine changes in weight-losing subjects or in animals [28]. A systematic review by Eichelmann et al. indicated no pronounced effect of plant-based diets on adipokines levels, but they added that the analysis was restricted by number and quality of available studies suggesting necessity for more research [29]. Some other clinical trials indicated that speci c nutritional components such as soups, vegetables, vegetable oils or dietary ber might have an increasing effect on leptin concentration in weight-stable healthy participants [26,[30][31][32]. Leptin is considered a pro-in ammatory cytokine being increased in patients with rheumatic diseases [7,33]. Consumption of meat is hypothesized to be one of the major contributors to promote in ammation in rheumatic diseases being supported by clinically observed improvement of disease activity by plant-based diets [16,34,35]. The increase of meat intake in our male MD participants led to an increase of leptin that supports a relationship between higher meat intake and in ammation. It remains unclear whether an increase in leptin leads to pathological increased in ammation of healthy participants as none of the participants showed adverse effects clinically or in laboratory tests [19].
Nevertheless, numerous in uencing factors such as sex, age and sexual hormone status have to be taken into account and may bias results of studies with adipokines. Signi cance of our age-related analyses is limited due to the younger age of our participants (median 27 years of age). Most of other publications reported age-related effects on adipokine levels in participants older than 60 years [36]. Furthermore, affection by sexual hormone status and menstrual cycle was not captured in our trial, which might imply a bias of our results concerning especially female participants [37]. The relatively short duration of our trial, 4-weeks, may be a factor contributing to the apparent diffuse association between diet and adipokine pro les. Interestingly, leptin responds quickly to changes in dietary composition. More than 20 years ago Havel et al. performed a trial comparing a just 24-hours-lasting high fat/low carb vs. a low fat/high carb diet in healthy women, and found a decrease of the physiologically circadian rhythmic leptin concentration in participants with high fat/low carb diet [38]. This nding of leptin is supported by other studies indicating possibility of rapid modulation of human adipokine levels [39,40]. Furthermore, Havel et al. hypothesized that greater intake of fat might lower leptin concentration whereas low fat/high carb diet induces higher levels of leptin. Comparability of those earlier studies with our results is limited as the composition of VD and MD were not comparable to low fat/high carb and high fat/low carb diet [18]. Although, intake of cholesterol and saturated fatty acids differed signi cantly between VD and MD in our cohort, we found no correlation of these parameters with leptin, which might be attributable to a similar intake of carbohydrates in VD and MD not-inducing a smaller insulin glucose response as seen in high fat/low carb diets. Leptin and insulin are known to be able to affect each other's serum concentration [41,42]. Furthermore, we found that adiponectin was negatively correlated with cholesterol and saturated fatty acids intake. Rodent models have shown an effect of dietary fatty acids on serum adiponectin indicating lower levels of adiponectin in high fat diet supporting our results, but clinical data in humans are lacking [43]. In our study, the type of diet affected the correlation of cholesterol and adiponectin as correlation was only found in VD, which is nearly free of dietary cholesterol. The observed correlation might, therefore, not be caused by the amount of cholesterol intake, but by other elements of the diet itself.
In a recent publication we reported about an association of lower levels of BCAA with lower in ammatory parameters in VD [19]. The previously described immunomodulatory potential of adipokines motivated us to investigate possible associations between lower levels of BCAA and lower levels of leptin and higher levels of adiponectin, respectively. The concentration of BCAA correlated inversely with the concentration of adiponectin in VD group. Regardless of diet, a correlation between BCAA and adiponectin was also reported by others [44,45]. Interestingly, we found a correlation of BCAA and adiponectin only in VD group, potentially indicating the pre-described anti-in ammatory effect of VD. However, we found no correlation of adiponectin with immune parameters underlining the controversial immunomodulatory potential of adiponectin. Recent literature suggests pro-and anti-in ammatory effects of adiponectin depending on pre-existing diseases [46]. In participants not suffering from in ammatory diseases, adiponectin appears to have anti-in ammatory potential being an insulin-sensitizing, vascular-protective and antiin ammatory protein [7,46,47]. Similar effects are described for VD underlining a potential relation between higher levels of adiponectin in VD [17,48]. Finally, while leptin and adiponectin are considered 'key cytokines' that have dedicated cellular receptors and well-described roles in human physiology [49], recent studies suggest that novel cytokines such as GDF15, CXCL14, S100A4 and Meteorin-like may also be important in the regulation of glucose and lipid metabolism and immune cell function [50][51][52][53]. Further research incorporating comprehensive pro ling of cytokines in clinical trials is necessary and desirable to uncover whether the concentration of these cytokines is modi able by dietary intervention.

Conclusions
In summary, serum levels of the most commonly known adipokines, leptin and adiponectin, do not fully explain the immunomodulatory potential of VD in healthy participants. However, the results of our trial suggests that the effect of VD and MD on adipokines levels might depend on participants' sex, as male and female participants showed different response on the nutritional change. Elucidating whether the observed sex-speci c differences emerge from the in ammatory potential of diets requires further investigation, ideally in long-term trials.

Consent for publication
Not applicable Availability of data and materials The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
None. The authors have no con icts of interest or nancial ties to disclose.

Funding for research
This research did not receive any speci c grant from any funding agency in the public, commercial or not-      Correlation of adipokine concentration with male participants' age. A. Correlation between baseline leptin and age (Spearman's r = -0.071, p = 0.767). B. Correlation between end leptin and age (r = 0.038, p = 0.874). C. Correlation between baseline adiponectin and age (r = 0.317, p = 0.174). D. Correlation between end adiponectin and age (r = 0.335, p = 0.149).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. CONSORT2010Checklist.doc