Study design and participants

This investigation is part of the Childhood Health, Activity, and Motor Performance School Study Denmark (CHAMPS study-DK), a large-scale, quasi-experimental study based on a natural experiment. At baseline in 2008, all 19 schools in the municipality of Svendborg, Denmark, were invited to participate in the study and ten schools agreed. The study and the methods used have been described in detail elsewhere [29], and only measurements pertinent to this paper are included below. Data for the present investigation are solely from the most recent follow-up conducted in 2015 and this investigation is therefore cross-sectional. For this follow-up parents/legal guardians of 745 adolescents (from a total of 1457 approached) provided written informed consent, and 705 adolescents participated in the testing. Of these, 558 had complete data and were included in the present analysis. The CHAMPS study–DK was approved by the Regional Committees on Health Research Ethics for Southern Denmark, Region of Southern Denmark (Project number: S-20080047 and S-20140105).

Measurements

All blood samples, anthropometric, and physiological measurements were collected on one day and cognitive tests were collected on a separate day.

Statistical analyses.

Differences between boys and girls and between participants included and excluded in the main analyses were analyzed using unpaired t-tests for continuous and chi-squared test tests for categorical variables.

The associations between waist circumference (standardized by sex and age), metabolic risk factors (standardized by sex and age) and cognitive performance outcomes were assessed using linear regression models adjusting for potential confounders (sex, pubertal status, SES, and age). Structural equation modeling (SEM) was used to further explore the hypothesized link between waist-circumference and cognition by assessing potential mediation by the metabolic risk factors. Separate models including each of the individual metabolic risk factors and the MetS-cluster score as mediating factors in the association between waist circumference and cognitive performance adjusting for the potential confounders were estimated. Fig 1 shows the hypothesized path model with the direct and indirect associations between waist circumference and cognitive outcomes through the metabolic risk factors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Schematic illustration of the path diagrams of the structural equation model.

Schematic illustration of the path diagrams of the structural equation model; waist circumference →cognition, with MET cluster score as the mediating factor. Directs associations = c’, indirect associations = ab, total associations = c’ + ab. All associations adjusted for sex, pubertal status, parental education and age. MET cluster score: sum of z-scores of HOMA, HDL cholesterol (inverse), systolic blood pressure and triglyceride.

https://doi.org/10.1371/journal.pone.0199281.g001

Direct (c’ in Fig 1) and indirect (through the risk factors, ab in Fig 1) associations between the standardized waist circumference and the (unstandardized) cognitive performance outcomes were calculated by maximum likelihood estimation of the SEM model with Satorra-Bentler standard error estimates. To evaluate the goodness-of-fit of the SEM models we evaluated the coefficient of determination of the model, as well as the Bentler-Raykov squared multiple correlation coefficient for each equation in the models. Analyses were re-run including flanker task RT outliers to investigate the effect of this exclusion criterion. Residuals were checked for normality across all analyses, and variables that did not result in normally distributed residuals were transformed by the natural logarithm. All analyses were carried out in Stata/SE 14.1.

Executive function and metabolic risk factors

Previous studies have indicated heterogeneous results regarding which individual metabolic risk factors are associated with executive functions. Similar to the present results, Scudder et al. (2015) found that unfavorable levels of HDLc and TG were associated with longer flanker task RT. However, after adjusting for putative confounders, the association with TG did not remain significant [20]. On the contrary, data from the NHANES study showed no association between cognitive performance measured with the digit span task (a measure of working memory) and exceeding the MetS cut point or being in the highest or lowest quartiles for HDLc or TG, respectively [22, 24]. Likewise, studies in adults are also conflicting (for review see e.g. [12]). However, a meta-analysis showed a consistent association between total cholesterol in midlife and future risk for cognitive impairment and dementia [36]. Interestingly, a similar relationship was not observed in older age-groups. Exploring the underlying mechanisms, different studies have found that dyslipidemia contributes to microvascular disease and vascular damage in the brain [12, 37]. However, the exact mechanisms by which dyslipidemia influence cognitive functioning are still unclear, mainly because dyslipidemia rarely occurs in isolation, rather is usually accompanied by hypertension and impaired glucose metabolism [37]. Future studies are needed both to address possible differences in the association between dyslipidemia and cognitive impairment across the lifespan, and the biological mechanisms driving this association.

In the present study, no association between the flanker task variables and insulin resistance measured by HOMA-IR was found. Previous studies in youth have used fasting glucose as a measure of glucose metabolism, and have likewise not observed any associations with executive functions [20, 22, 24]. In adults, an association between diabetes and cognitive impairment (including dementia) has been well established (e.g. [38, 39]). Insulin resistance has been proposed as a key factor linking the association between MetS and cognitive impairment [40], and hyperglycemia has been found to be an important contributor of the association between MetS and cognition [41]. The discrepancy between studies in youth and adults suggests that dysregulation of glucose metabolism in the early stages might not affect cognition, whereas more severe, later stage insulin resistance and concomitant severe hyperglycemia might. The mechanisms by which type 2 diabetes and insulin resistance affect cognition and brain function includes disrupted white matter integrity, vascular micro- and macro-abnormalities in the brain and brain atrophy (for review see e.g. [12, 13]). These conditions are only seen in late stages type 2 diabetes and after prolonged hyperglycemia [37], and not typically observed in youth populations. This might explain the discrepancies in results from different age groups.

The present study does not find any associations between sysBP and executive functions, which corroborates several studies in youth [20, 23]. However, others found that children and adolescents with sysBP ≥ 90th percentile performed poorer on a working memory task compared to their normotensive peers [22, 42]. As such, it has been suggested that hypertension may cause impaired cognitive function via impairment of vascular reserve and microvascular disease [43], which are consequences of severe and prolonged hypertension and therefore, to our knowledge, not found in healthy youth.

Taken together, there is no agreement between studies on which of the metabolic risk factors are associated with executive functions, and it is possible that individual differences in health status, adiposity, fitness level etc. across study populations and the different age groups included contribute to the observed heterogeneity. Despite the lack of consistency among studies, current evidence indicates that optimal cognitive function may be compromised by metabolic dysregulation even in the first decades of life.

Results from our study did not show any associations between the MetS-cluster score and measures of executive function. To date, only a few studies have investigated the associations between clustered metabolic risk and executive functions in children or adolescents [20–23]. Yau et al. (2012) found that adolescents with MetS performed worse on measures of mental flexibility compared to adolescents without symptoms of MetS. However, none of the other tests of executive function were associated with MetS status [21]. These findings were combined with reduced structural integrity in the brains of adolescents with MetS compared to their healthy counterparts [21]. Similarly, Rubens et al. (2016) found that adolescents with MetS had lower working memory and attentional performance as well as lower reading scores, whereas no differences were observed for perceptional and visuospatial reasoning skills using the Wechsler Intelligence Scale for Children Revised. Additionally, Scudder and colleagues (2015) performed a study in children (mean age 7.5 years) and found that those not meeting any MetS risk-factor criteria had shorter RTs on a modified flanker task and performed better during the more difficult conditions of the task compared to children presenting with one or more risk-factors for MetS. Accordingly, current evidence suggests an association between MetS and some measures of executive function, but highlights the possibility that MetS may have differential effects on the various executive function subdomains.

Possible mechanisms driving the association between adiposity and executive functions

Most studies, both in youth and adults, agree on a negative relationship between weight status and cognitive control [10, 15–17]. Many possible mechanisms linking adiposity and cognition have been proposed including all the metabolic risk factors composing MetS, but also factors like vascular damage, low grade inflammation and hyper glucocorticoids have been proposed as possible contributing candidates [13]. In the present study waist circumference displayed an indirect positive association with incongruent RT through a higher MetS-cluster score and an indirect positive association with both congruent and incongruent RT through lower HDLc. This suggests the negative impact of excess adiposity on cognition seen in some studies (e.g. [15–17]) may be mediated, at least in part, by the effect of adipose tissue on these metabolic risk factors. However, in mediation analyses performed on cross-sectional data controlling for prior level of the mediators or the outcome (to reduce potential reverse causality bias) is not possible, which renders it impossible to conclude anything on causality [44]. Therefore, future studies should verify this relationship, optimally in a prospective study design and including a larger array of adiposity-related metabolic markers. One previous cross-sectional study in adults found that insulin sensitivity mediated the relationship between weight status, measured by BMI, and brain activation during a working memory task [14], which is in contrast to our findings on insulin resistance. These discrepancies could be due to different measurements of exposure, mediator and outcome, and also to the differences in age groups studied. Our findings of no direct association between adiposity and cognitive control is raising the intriguing question that it might not be excessive adiposity per se, but the concomitant metabolic disturbances in traditional metabolic risk factors that drive the association with impaired cognition. It has been shown that these biological markers can be improved by lifestyle changes even without substantial weight loss [45, 46], which is encouraging given the knowledge that weight loss can be difficult to maintain.

Limitations

A clear limitation in the present study is the cross-sectional study design making it impossible to conclude causality. Also, it is possible that lower executive functions could predispose weight-gain associated with impulsive behavior such as e.g. snacking or low physical activity levels (i.e. reverse causality). Further, residual or unmeasured confounding cannot be rejected. Inclusion of a more comprehensive cognitive test battery could have extended our results to other aspects of executive function, memory and cognition in general, allowing for an understanding of the general versus selective nature of the relationship of MetS on cognition. Also, it is possible that other biologic consequences of excess adiposity, not measured in this study, could mediate the association between adiposity and executive control, and future studies could include e.g. measures of vascular damage or subclinical atherosclerosis. Finally, the sampled population in this study is, according to the biological characteristics, a metabolic healthy and relatively normal weight cohort of young Danish people, and might not be representative for adolescents in an international context. The variation in waist-circumference for age is therefore likely to be smaller than observed elsewhere, potentially explaining the absence of association between waist circumference and cognitive outcomes.