Bloody olfaction? Confounding associations of sex and age on the influence of blood parameters and body weight on odor identification performance in healthy adults

Olfactory function and nutrition are closely related and may influence each other via metabolic parameters. However, the relationship between nutritional blood parameters and olfactory performance is still unclear. Inconclusive findings exist for specific blood parameters. In this extensive analysis, we examined the relationship between olfactory performance, measured with MONEX-40, as well as intensity and pleasantness ratings with 38 metabolic blood parameters, age, sex, and the anthropometric measurements body mass index (BMI) and body fat percentage (BFP). Therefore, we included data of 418 healthy, well-phenotyped Caucasians of the Enable cohort. We replicated age-dependent olfactory identification scores ( p < 0.001) and found slight evidence for a body fat dependence measured with BFP (BF 10 = 10.466). We further identified a sex difference only in middle-aged adults ( p < 0.001) that could be explained by environmental factors. Several blood parameters correlated significantly with the MONEX-40 score ( p < 0.05 – p < 0.001). However, these effects diminished after adjusting for sex and age ( p > 0.9) that were identified as confounders. The same applies for BFP. In addition, no pa- rameters were identified to correlate significantly with perceived olfactory intensity or pleasantness score if controlled for sex and age ( p > 0.08). Our results suggest that metabolic blood parameters are not related to olfactory identification performance in a relevant manner and highlight the importance of controlling for sex and age in chemosensory research.


Introduction
Olfactory perception is responsible for preferences in everyday life, especially with respect to food choice, nutritional habits, and appetite. The human sense of smell develops already prenatally and odors perceived by the fetus influence liking and attention towards aromas after birth [1]. Associated with this is the acceptance and pleasantness of corresponding food items. However, humans do not only prefer food due to the odor but can also smell if food is rotten or poisonous. Therefore, the sense of smell is essential for safe nutrition and influences eating behavior [2][3][4]. Food consumption, in turn, is reflected in metabolic parameters as well as anthropometric measures.
On the other hand, evidence exists that fasting enhances olfactory perception, whereas in a satiated state, odor sensitivity decreases. Therefore, various metabolic processes seem to regulate olfactory perception [5], resulting in a complex interaction. This leads to the research question, whether correlations between metabolic parameters and olfactory performance exist. mass index (BMI) that is often used to classify participants of studies in an obese and non-obese group. Several studies found a negative association between olfactory function and BMI [6][7][8][9][10]. However, other studies established that BMI is not related to olfactory dysfunction overall [11,12] or among healthy people [13] or that the relationship is age-and sex-dependent [14]. In some studies, olfactory function was measured only with identification tests; here, it is not clear whether associations with the BMI exist. Whereas in one study, a correlation was established, other studies did not support this [15] or only supported this finding for smokers [16]. As the BMI is a measure of a person's relative weight, it is closely related to metabolic parameters and nutrition. A more valid predictor for excessive body fat, especially for active people, is body fat percentage (BFP), as this also takes into account muscle mass [17]. Neither BMI nor BFP are sex-or age-independent [14,[18][19][20][21]. However, using BFP instead of the BMI could provide a more stable outcome if associations between body weight and smelling ability are examined.

Sex and age and the relation to smelling ability
Sex and age, on the other hand, have been broadly shown to correlate with olfactory performance. A decreased olfactory function can be found in older adults [2,7,11,[22][23][24][25][26]. That can be led back to structural changes in the nose and an aging olfactory system as well as a decline in cognitive abilities and memory function [27,28]. Most studies suggest a gender effect: women outperform men with regards to smelling abilities [25,26,[29][30][31][32]].

Metabolic parameters and the relation to smelling ability
Several metabolic parameters have already been examined for a relationship with our sense of smell. A high odor identification score was found to be significantly associated with high levels of free triiodothyronine (FT 3 ), however not with thyroid-stimulating hormone (TSH) or free thyroxine (FT 4 ) [33]. Also, in other studies, a diminished olfactory function for adults with hypothyroidism has been established [34,35]. However, a significant association was reported not in all studies [36]. Cholesterol was also in the interest of several studies. Reduced HDL cholesterol and elevated triglycerides, but not LDL or total cholesterol, were significantly associated with olfactory dysfunction in Korean women [37]. However, neither HDL nor LDL cholesterol or triglycerides individually were found in another study to be significantly associated with smell dysfunction or thresholds [2]. Nevertheless, in one of these studies, a significant association with higher blood concentrations of total cholesterol was established [38]. Yet other studies proved different outcomes based on age and sex [14], the use of adjustment [11] or between healthy people and patients with type 2 diabetes mellitus [39], which in turn is associated with olfactory performance [39,40], but not in others [36,41]. However, diabetes mellitus type 1 or HbA1c were not found to be related to olfactory function [42]. Another blood parameter related to nutrition and sensory perception is the blood concentration of insulin. Higher blood levels of insulin and poor insulin sensitivity were shown to decrease olfactory performance and mediate the effect of body weight on olfactory performance [43,44]. However, no association between blood insulin and self-reported olfactory dysfunction was reported [11]. The effect of intranasal insulin was found to decrease [45][46][47] as well as enhance [43,48,49] olfactory performance in different studies. Glucose levels were established to be associated with olfactory performance in older people dependent on age and sex [14].
In addition to the above-mentioned metabolic parameters, creatinine, urea, and phosphorus but not albuminmolecules that are related to renal functionwere found to be associated with odor threshold [50]. Albumin, though, was associated with odor identification [51]. This study also established associations with total cholesterol and LDL cholesterol for a study cohort, in which patients with chronic kidney disease and kidney failure were included. In addition, patients with liver cirrhosis report olfactory loss, although, neither serum levels of bilirubin nor zinc were shown to correlate with olfactory function [52].
As stated above, former research provides first evidence that metabolic parameters might affect olfactory performance. Inversely, olfaction might influence metabolic parameters via changes of eating behavior (see Fig. 1). Due to several interventions, cross-sectional or patient studies, several nutrition-related parameters have been identified. The results of the studies often differ, partly depending on the sex, age, or diseases of participants or other confounders. In large studies in the US (NHANES), Korea (KNHANES), and China, representative population samples were investigated [11,14,37,38]. In these studies, the examined blood values were limited or no objective olfactory tests were conducted. Therefore, we performed a broad analysis for well-phenotyped healthy Caucasians from 18 to 85 years of the Enable cohort [20]. We did not focus on special parameters but explored 38 blood parameters in addition to sex, age, and body weight. We aimed to establish whether and which metabolic parameters correlate with olfactory identification performance, more accurately the 40-item Monell Extended Sniffin' Sticks Identification Test (MONEX-40) score [53]. We hypothesized that age and sex are important confounders of potential relationships. In addition, we hypothesized that BFP instead of BMI provides a better measurement for the evaluation of the relation between body weight and olfactory performance.

Ethics statement
We used data of the Enable study [20] The study protocol for this study was approved by the ethical committee of the Faculty of Medicine of the Technical University of Munich (approval no. 452/15) and Friedrich-Alexander-Universität Erlangen-Nürnberg (approval no. 291/15B). The guidelines of the International Conference on Harmonization of Good Clinical Practice and the World Medical Association Declaration of Helsinki (in the revised version of Seoul, South Korea 2008) were considered. All study participants had provided written informed consent.

Participants
To address our research question, we used data of the wellphenotyped Enable cohort that was described in detail elsewhere [20]. This cohort covers healthy participants (but partially with overweight) out of four age groups (children, young adults, middle agers, older adults). For our analysis, we included participants from the young (18 - To get valid smelling scores, we excluded participants reporting respiratory illnesses, acute allergic reactions or blocked noses (n = 28). In total, we included 418 participants in our analysis (see Table 1). Participants' BMI was between 18.6 and 34.9 kg/m2 (mean 25.6, SD 4.2, 49.3% had a BMI of < 25). BFP was in the range of 3.0 to 53.5 (mean 31.6, SD 10.0) (measured with Seca mBCA 515 device (Seca GmbH & Co KG, Hamburg, Germany)).

Olfactory performance
Participants' olfactory performance was tested using the 40-item Monell Extended Sniffin' Sticks Identification Test (MONEX-40) [53]. Participants were presented 40 odors with the help of Sniffin' Sticks and were asked to identify each odor based on 4 descriptors (3 wrong, 1 correct) that were provided per odor. Therefore, participants could achieve a score from 0 (all sticks were identified false) to 40 (all sticks were identified correctly). In addition, participants rated the perceived intensity and pleasantness of each odor on Likert scales ranging from 0 to 10 or − 5 to 5, respectively. We calculated the average intensity and pleasantness scores for each participant by taking the means of the 40 respective participant's ratings.

Statistical data analysis
Using a Chi-square test, we examined the association between age and sex with the MONEX-40 score. To account for equality of variances, we squared the MONEX-40 score. A two-way ANOVA with the factors age (young, middle-aged, and old adults) and sex (female, male) on the dependent variable of the squared MONEX-40 score was used to detect effects of sex and age group on the ability to smell. The Shapiro-Wilk test revealed that not all blood parameters were normally distributed. Therefore, we performed a non-parametric correlation using Kendall's tau-b. In addition to the 38 blood parameters, we added BMI, BFP, sex, and age as variables. To detect nutritional parameters that correlate with the MONEX-40 score after controlling for sex and age effects, we additionally performed partial correlations with sex and age as covariates. All reported p-values were adjusted using Holm correction [55].
In addition to the above-mentioned statistical methods, we used Bayesian non-parametric correlation with non-informative prior to gain more information about the relation between the MONEX-40 score and blood parameters. This approach focuses on both the null and the alternative hypothesis at the same time and therefore can quantify evidence for the null hypothesis [56], in our case that there exists no linear correlation. To control for sex and age, we separated the participants into groups based on age and sex in accordance with the results of the ANOVA. As the Bayesian statistics are not affected by multiple comparisons, we do not have to adjust for multiple comparisons [57]. To simplify the language, we used in the following the word "significant" not only for frequentist correlations with Holm corrected significance levels below 0.05 but also if the Bayes factor BF 10 was above 10, Table 1 Overview of participants' descriptives. We present the number of participants divided into age and sex groups as well as the median, mean and standard deviation (SD) of the age groups. indicating strong evidence for a non-zero correlation. To examine if sex and age are confounding factors, we performed confounding analysis [58] (that is statistically the same as mediation analysis based on [59]) for blood parameters for which the significant relationship disappeared after controlling for sex and age. The blood parameter of interest was the independent variable; the MONEX-40 score was the dependent variable, and sex and age were considered as potential confounders. We performed separate analyses for each blood parameter. We used bias-corrected bootstrap with 1000 replications to calculate confidence intervals and account for non-parametric data [60]. According to Rucker and Zhao [61,62], we interpreted the indirect effect as a requirement for a confounding effect. Hypotheses were confirmed if the 95% confidence interval (CI) did not include zero.
To investigate other olfactory performance characteristics, we also analyzed the relationship between intensity and pleasantness scores of odors with age, sex, and nutritional parameters using partial Kendall's tau correlation analysis with conditions on sex and age. Results are presented in the supplementary material.

Influence of sex and age group on MONEX-40 score
A Chi-square test showed that there was no significant association between age and sex with the MONEX-40 score (X 2 (2, N = 418) = 0.676, p = 0.713). Out of 418 participants, 77 (18.4%) reached a MONEX-40 score lower than 27 that is often used as a cut-off value to identify normosmic subjects (Fig. 2) [47,65]. 76.6% of the hyposmic subjects belonged to the older age group, 19,5% to the middle-agers, and only 3.9% were young adults (percentage of hyposmic subjects per age group: young 3.4%, middle-aged 8.4%, old 38.6%).
A two-way ANOVA with the factors age (young, middle-agers and older adults) and sex (female, male) on the dependent variable of the squared MONEX-40 score revealed a significant interaction effect (F(2, 412) = 3.744, p = 0.024, η 2 = 0.014, Levene's test for equality of variances: F(5, 412) = 1.786, p = 0.115). Post-hoc analyses revealed a significant decrease of the smelling ability from young or middle-agers to older adults for both sexes. A significant sex difference was only found for middle-agers with a better score for females (Fig. 3, Table 2).

Correlation between blood parameters and MONEX-40 score
Next, we examined the associations between the blood parameters as well as BMI and BFP on the MONEX-40 score. As sex and age were found to have a significant influence on the MONEX-40 score, we conducted in addition to zero-order correlations partial correlations, either by conditioning on these factors (frequentist) or by building subgroups (frequentist and Bayesian). As sex was not found to have a significant effect for young and old adults, we did not split these cohorts based on sex. This resulted in four different subgroups (young adults, female middleagers, male middle-agers, and older adults). Coefficients, p-values, and BF 10 values for all correlations are shown in the supplementary material.
The frequentist zero-order correlation between each blood parameter and the MONEX-40 score revealed weak (|τ| < 0.2) negative correlations for RDW, γGT, LDH, urea, uric acid, and ferritin as well as a weak positive correlation for phosphorus and moderate (|τ| < 0.3) positive for GFR. In addition, the correlations of the MONEX-40 score with sex and age were statistically significant. All parameters showing a significant correlation with smell ability were also significantly associated with sex and/or age. A partial correlation for age and sex revealed no significant factors (see Table 3), indicating a confounding effect of sex and/or age (see Fig. 4).
Bayesian zero-order correlation provides evidence that in addition to the above-mentioned parameters, creatinine (τ = − 0.105, BF 10 = 10.677) and BFP (τ = − 0.105, BF 10 = 10.466) correlated negatively with the MONEX-40 score. However, the significant variables again showed a significant relationship with sex or age (see supplementary material).

Young adults
Neither age nor sex correlated significantly with the MONEX-40 score in the group of young adults. That resembles the results of the ANOVA. Therefore, we did not control for sex and age. Regarding blood parameters, there was a weak negative correlation between TSH and MONEX-40 score that was statistically significant (τ = − 0.288, p = 0.006, BF 10 = 302.008). No other variables reached significance (see supplementary material).

Middle-agers
In the group of middle-agers, sex correlated significantly with the MONEX-40 score (τ = 0.309, p < 0.001, BF 10 = 1.159e+7) but not with age. This fits with the results of the ANOVA. Therefore, we controlled for sex. No blood parameters reached significant correlations with the MONEX-40 score, neither using frequentist nor Bayesian correlation (see supplementary material).  Table 1. For understanding that is more intuitive the unsquared MONEX-40 scores are shown. ***: p < 0.001 in post-hoc comparisons of sex and age group in two-way ANOVA with factors age and sex on squared MONEX-40 score (see Table 2). * p < 0.05, ** p < 0.01, *** p < 0.001, Holm adjusted for comparing a family of 6.

Older adults
In the group of older adults, neither sex nor age correlated significantly with MONEX-40 score in accordance with the ANOVA results. Therefore, we did not separate the senior cohort. No blood parameters correlated significantly with the MONEX-40 score in frequentist or Bayesian correlation (see supplementary material).

Table 3
Kendall's tau correlation between selected parameters 1 and MONEX-40 score, sex, and age. Results of zero-order correlation as well as partial correlation with condition on sex and age are illustrated.  Fig. 4. Example for a confounding effect of age on metabolic blood parameters: Age, GFR, and MONEX-40 score correlated with each other. However, the correlation between GFR and MONEX-40 score became insignificant after controlling for age, indicating a confounding effect of age. In this case, sex did not correlate significantly with GFR.

Confounding analysis of nutritional parameters on MONEX-40 score through age and sex
Several blood parameters showed significant zero-order correlations with the MONEX-40 score; however, all partial correlations became insignificant. To provide evidence for a confounding effect of age and sex on these relationships, we conducted confounding analyses for the blood parameters RDW, γGT, LDH, phosphorus, creatinine, GFR, urea, uric acid, ferritin, TSH as well as BFP. Table 4 summarizes the results necessary to detect confounders, further statistics can be found in the supplementary material. Our data indicated indirect-only confounding [62].

Discussion
Whether and which blood parameters may be associated with olfactory performance is still unclear, as several studies lead to different outcomes (e.g. hypothyroidism [34][35][36], cholesterol [2,11,14,37,38], type 2 diabetes mellitus [36,[39][40][41], insulin [11,43,44], see Introduction). In this broad study, we examined the relationship between olfactory performance and 38 blood parameters as well as sex, age, and the anthropometric measurements BMI and BFP of 418 healthy, well-phenotyped Caucasians. Therefore, we did not focusas other studieson olfactory performance effects for patients suffering from illnesses like diabetes [37,38,40,43,47], thyroid [31][32][33], or kidney [48][49][50]62] diseases but examined a whole bandwidth of metabolic parameters equally for healthy participants (including overweight subjects). By using the MONEX-40 test (51), we were able to objectively examine participants' olfactory identification scores. Within a total of 18.4% of subjects that can be classified as hyposmic subjects using a typical cut-off value of 27, we have a representative sample if it is taken into account that over one-third of our participants were older than 75 years (63, 64).

Age and sex
It is well established that olfactory performance is age-and sexdependent. From youth to adulthood, olfactory performance typically increases greatly. Around the retirement age, a pronounced decline can be observed [25]. Whereas the increase can be explained by learning effects, the later decrease is most likely due to age-related structural changes in the nose and an aging olfactory system. In addition, a decrease in cognitive abilities and memory decline can contribute to the performance decline [27,28]. We were able to replicate these results, as there were significant negative differences between our young and middle-aged participants compared to older subjects. No significant differences were established between the young and middle-aged subjects in accordance with the flatting olfactory performance curve for subjects within these age range [25]. Within our three pre-defined age groups, age did not correlate significantly with olfactory performance. This indicates that within an age group, individual differences are more important than a smaller age difference.
In addition to age effects, previous studies reported a sex difference with better olfactory performances for women versus men. Whereas some studies found such effects over a broad age range [25,26,31], some only found these effects for young adults [22,32]. Interestingly, our female participants outperformed the male participants only in the middle-aged cohort, whereas no differences were established for young and old participants. Similarly, only in the middle-aged cohort sex correlated significantly with the MONEX-40 score. This could be due to the larger sample size of the middle-aged compared to the young and old age groups as sex-related differences were shown to be relatively small and therefore need a larger sample size to become significant [25,31]. However, the difference in mean MONEX-40 scores was 2.4 for our middle-aged participants and less than half a point if males and females in the young and older age group were compared. This reveals a sample-size independent effect of a more enhanced sex difference in middle-aged adults. According to our results, discussed explanations based on hormones [32] also seem insufficient as they cannot explain the non-significant sex difference in our young cohort. In addition to anatomical and/or physiological as well as higher-level cognitive differences between males and females, a difference in odor experience or learning was discussed to cause the differences [31,66]. It was shown that women are more likely to take main responsibility for meal planning or preparing and that older men are more likely to take no responsibility for these activities [67]. That corresponds to the trend of an increasing proportion of men, especially men of younger age, that are cooking [68]. Therefore, the differences in exposure to and sensibilization for odors due to cooking and meal preparation could be one explanation that differences were only found in middle-aged adults. With a diminished sense of smell for older people, this possible effect seems to lose its importance. Further research should investigate this potential relationship.

BMI and BFP
We did not find a significant correlation between olfactory identification performance and BMI. However, we established evidence in favor of a negative correlation between BFP and the MONEX-40 score. As several studies suggest a significant relationship between body weight or obesity and olfactory performance [9], we recommend using BFP instead of BMI as a measure of overweight. An explanation could be that BFP in contrast to BMI can better differentiate between muscle and fat mass [17]. Stratified analysis of our data did not establish significant Table 4 Total indirect and direct effects of given blood parameters and BFP on MONEX-40 score through age and sex.

Blood parameters
There are already existing large-scale studies in Korean, Chinese, and US populations. KNHANES (Korea) data revealed that blood parameters were not associated with olfactory self-reported dysfunction if the effects were adjusted, although t-tests revealed some significant blood parameters [11]. Among the Chinese population, only significant associations after adjusting for total cholesterol were reported [38]. In the NHANES study (US), however, it was revealed that the direction is opposite for men and women and no blood parameters were found with not varying association by sex and age group [14]. Both latter studies also relied on self-reported olfactory function. Our study adds to these previous ones for a Caucasian population with the advantage that we objectively assessed olfactory ability using the MONEX-40 test.
We were able to show that several blood parameters relevant in metabolism (namely RDW, γGT, LDH, phosphorus, creatinine, GFR, urea, uric acid, and ferritin) correlated significantly with odor identification ability. All these parameters also correlated significantly with sex and/or age in accordance with the literature [69][70][71][72][73][74]. Controlling for these effects diminished the correlations with the MONEX-40 score. In addition, no parameters were identified to correlate significantly with perceived olfactory intensity or pleasantness score if controlled for sex and age (see supplementary material). We confirmed the non-correlation using Bayesian hypothesis testing, in fact, for nearly all blood parameters evidence favored a zero-correlation (see supplementary material). This indicates that age and sex should be of particular interest if examining relationships between blood parameters and olfactory function.
However, not all effects disappeared after controlling for sex and/or age. In stratified correlation analysis, we identified that TSH correlated negatively with the MONEX-40 score for young adults. That fits to other studies that found a diminished olfactory function for adults with hypothyroidism and therefore with low levels of FT3 and FT4 and high levels of TSH [34,35]. Similarly, the authors of another study found that a high odor identification score is significantly associated with high levels of FT3. However, in this study, no significant associations with TSH or FT4 were established [33]. Whereas in these studies, young adults or subjects over a broad age range were investigated, in a study with subjects older than 55 years no significant association between odor identification and thyroid disorder was found [36]. This corresponds to our result that the relation between TSH and MONEX-40 is age-dependent.
Besides TSH, if removing sex and age effects, we were not able to find evidence for a relationship between metabolic parameters -partially resulting from eating behavior -and odor identification performance. In fact, we were able to show evidence in favor of a zero-correlation for cholesterol (HDL, LDL, or total) or triglycerides. These parameters were identified in some studies to be related to olfactory function [11,37,44,50]; however, in other studies no significant associations were reported [2,11,38]. Serum insulin also did not affect olfactory identification performance in our study as found in a previous study [11], even if intranasal insulin levels seem to affect olfactory performance, either negatively [45][46][47] or positively [43,48,49]. Phosphorus, urea, and creatinine were also shown previously to be related to olfactory function, more specifically to olfactory thresholds [50]. This study cohort included patients with renal insufficiency and in the analysis, authors did not control for sex and age. We did establish significant associations for zero-order correlations, too, but for creatinine and phosphorus in other directions than the previous paper. We found negative correlations for creatinine and positive correlations for phosphorus while they established the complete opposite effects. In addition, we showed that no direct but only indirect effects existed if sex and age were considered. Taken together, the discrepancy in the directionality could be explained by differences in study populations. Our broad analysis, therefore, highlights the importance of controlling for sex, age, and health status of the study population.
Nevertheless, we identified blood parameters of which the correlating effect with the odor identification ability could be explained through (I) a confounding effect by sex and/or age. Another possible explanation could be that (II) these blood parameters mediate the age and sex effect on odor identification (see Fig. 5). Explanation I entails that the significant correlation is due to the fact that people within a given age range or with a given sex (a) have a special level of a blood parameter and (b) have a special prevalence of smell ability, but that (a) and (b) do not (totally) influence each other. Explanation II claims that age and sex do influence the smelling ability via blood parameters as mediators. That means that special blood parameters that change during aging or that are different between males and females influence olfaction and, therefore, can partially explain possible sex and age differences.
As we were interested in the correlation between blood parameters reflecting nutritional behavior and the metabolic state, we performed confounding analysis to test explanation I. Indeed, age and sex were identified as confounders. However, although we provided evidence for a confounding effect, a mediating effect could also occur. Mediation   Fig. 5. Two possible explanations for the significant zero-order but non-significant correlation if a blood parameter BP is controlled for sex and / or age.
analysis to clarify this question should be based on thorough literature research to justify the underlying theory and precisely include relevant blood parameters in the model. In addition, targeted studies should be performed, for example through interventions (compare [44] that showed a mediating effect of insulin on the relationship between BMI and olfactory sensitivity).

Conclusion
Nutrition and olfaction are closely related and may influence each other via metabolic parameters. Using data of the well-phenotyped Enable cohort, we examined the relationship between 38 blood parameters as well as age, sex, and body weight/composition on olfactory identification performance measured with MONEX-40. We replicated age-dependent olfactory identification scores and found slight evidence for a body fat dependence measured with BFP. Our data suggest a sex difference only in middle-aged adults that could be explained by environmental factors. In addition, we were able to show that several blood parameters relevant to metabolism correlate significantly with odor identification ability. However, removing the effects of sex and age diminished these correlations, and sex and age seem to confound the correlations in healthy participants. To summarize, our results suggest that metabolic parameters resulting from nutrition are not related to olfactory identification performance in a relevant manner and highlight the importance of controlling for sex, age, and health status of study population in chemosensory research.

Declaration of Competing Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Data Availability
Data will be made available upon request.

Author contributions
BB, TS, DV, HH, JF provided data of Enable cohort. DS performed statistical analysis and wrote the manuscript. BK and JF revised the manuscript. All authors critically read the manuscript and approved the final version. The present work was performed in partial fulfillment of the requirements for obtaining the degree "Dr. rer. biol. hum.".

Funding
This work was supported by a grant of the German Ministry for Education and Research (BMBF, grant no. 01EA1409A), the German Academic Scholarship Foundation, as well as the Initiative Campus of the Senses, a project with financial support of the Bavarian Ministry of Economic Affairs, Regional Development and Energy (StMWi) and the Fraunhofer Society. No influence on the design, conduct, analyses, interpretation, or publication of the trial exists. The preparation of this paper was supported by the Enable Cluster and is cataloged by the Enable Steering Committee as Enable 41 (http://enable-cluster.de).

Data availability statement
All analyzed data is part of the Enable dataset [20].

Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.physbeh.2022.113907.