Print Close

Multivariate association between risk factors for non-communicable diseases and cortical thickness

Poster No:

2086

Submission Type:

Abstract Submission

Authors:

Eliana Nicolaisen-Sobesky¹^,2, Somayeh Maleki Balajoo¹^,2, Mostafa Mahdipour¹^,2, Agoston Mihalik³^,4^,5, Felix Hoffstaedter¹^,2, Janaina Mourao-Miranda³^,4, Masoud Tahmasian¹^,2, Simon Eickhoff¹^,2, Sarah Genon⁶^,2

Institutions:

¹Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich, Jülich, NRW, ²Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf, Düsseldorf, NRW, Germany, ³Centre for Medical Image Computing, Department of Computer Science, University College London, London, ⁴Max Planck University College London Centre for Computational Psychiatry and Ageing Research, University College London, London, United Kingdom, ⁵Department of Psychiatry, University of Cambridge, Cambridge, United Kingdom, ⁶Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich, Jülich, Germany

First Author:

Eliana Nicolaisen-Sobesky
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf
Jülich, NRW|Düsseldorf, NRW, Germany

Co-Author(s):

Somayeh Maleki Balajoo
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf
Jülich, NRW|Düsseldorf, NRW, Germany

Mostafa Mahdipour
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf
Jülich, NRW|Düsseldorf, NRW, Germany

Agoston Mihalik
Centre for Medical Image Computing, Department of Computer Science, University College London|Max Planck University College London Centre for Computational Psychiatry and Ageing Research, University College London|Department of Psychiatry, University of Cambridge
London|London, United Kingdom|Cambridge, United Kingdom

Felix Hoffstaedter
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf
Jülich, NRW|Düsseldorf, NRW, Germany

Janaina Mourao-Miranda
Centre for Medical Image Computing, Department of Computer Science, University College London|Max Planck University College London Centre for Computational Psychiatry and Ageing Research, University College London
London|London, United Kingdom

Masoud Tahmasian
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf
Jülich, NRW|Düsseldorf, NRW, Germany

Simon Eickhoff
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf
Jülich, NRW|Düsseldorf, NRW, Germany

Sarah Genon
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich|Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, NRW, Germany

E-Poster

Introduction:

Non-communicable diseases and brain health have a common set of risk factors, including unhealthy diet, hypertension, smoking, excessive alcohol drinking and physical inactivity. A biomarker used to study brain health is cortical thickness (CT). Studies linking risk factors to CT have mostly used univariate/bivariate approaches and yielded inconsistent results1–3. Moreover, bivariate/univariate approaches provide a partial view of such association, while multivariate approaches can link a wide set of risk factors to whole-brain CT. In addition, to gain a comprehensive understanding of the underlying neurobiology, multiple brain features should be analyzed, such as brain function and neurotransmitter systems. Finally, there is a male bias in biomedicine, and consequently findings might not generalize to women. Hence, gender/sex-specific analyses are needed.

Methods:

We analyzed women (n=3685, 46-81 years) and age-matched men (n=3685; 46-81 years) from UK Biobank, without self-reported non-cancer illnesses. Risk factors included 70 variables, spanning body size and metabolism, physical activity, sleep, diet, smoking and alcohol consumption, among others. CT was parceled in 148 cortical parcels (Destrieux atlas). We used a multivariate method (regularized canonical correlation analysis, RCCA) to find risk factors that correlated to CT (i.e., latent dimensions)4,5. The RCCA was embedded in a multiple-holdouts machine learning framework to optimize the generalizability and stability of the latent dimensions. We used two consecutive splits of the dataset4–6: 5 outer splits (model evaluation and statistical testing with 1000 permutations), each with 5 inner splits (hyperparameter optimisation with out-of-sample generalisability criteria). We used raw (absolute), proportional, and corrected (regressing out brain size) measures of CT. We ran 6 independent RCCAs (one per sex/gender and CT measure) and compared the loadings with Pearson's correlation and spin test7. Finally, we compared the brain loadings with an extensive set of brain features8 including functional and neurotransmitter patterns.

Results:

We found one latent dimension linking inter-individual variability in risk factors to inter-individual variability in raw CT (Fig. 1-2) (women r_...range=0.25-0.31, p=0.005-0.005; men rr_...ange=0.28-0.32, p=0.005-0.005). Of note, this latent dimension was similar across sexes/genders, and across CT measures, on both risk factors loadings (r>0.96, p<0.001) and brain loadings (r>0.90, p<0.001). This dimension captured variability in sedentarism/physical activity, as well as body morphology and metabolism. The CT variability captured by the latent dimension described an axis from insula and anterior cingulate cortex to occipital lobe and superior parietal areas. Interestingly, this brain pattern was associated with binding potentials of neurotransmitter receptors 5-HT1a (r>0.55, p<0.001), D2 (r>0.53, p<0.001) and VAChT (r>0.50, p<0.001), as well as transporter DAT (r>0.48, p<0.001, except for men-raw CT).

Conclusions:

metabolism to CT variability ranging from insula and anterior cingulate cortex to occipital and parietal areas. These results highlight the adipose-tissue-brain axis that has gained attention recently. In addition, our results indicated that the cortical structural pattern related to risk factors was associated with serotoninergic, dopaminergic, and cholinergic systems. Interestingly, these molecular systems have been linked before to phenotypes associated to body morphometry and activation, such as physical activity, satiation, feeding behavior, obesity, or anorexia nervosa9,10. Hence, this work suggests that cortical structure related to these neurotransmitter systems may be associated with risk factors for non-communicable diseases, highlighting the multivariate and multi-level association between brain and body phenotypes.

Lifespan Development:

Aging

Modeling and Analysis Methods:

Multivariate Approaches ²

Neuroanatomy, Physiology, Metabolism and Neurotransmission:

Anatomy and Functional Systems ¹

Cortical Anatomy and Brain Mapping

Transmitter Systems

Keywords:

Aging

Cortex

Machine Learning

Multivariate

Neurotransmitter

RECEPTORS

STRUCTURAL MRI

Systems

Other - brain-body associations, interindividual variability

^1|2Indicates the priority used for review

·First latent dimension from regularized canonical correlation analysis between risk factors for non-communicable diseases and raw cortical thickness

·First latent dimension from regularized canonical correlation analysis between risk factors for non-communicable diseases and raw cortical thickness

Provide references using author date format

1. Gurholt, T. P. et al. Population-based body–brain mapping links brain morphology with anthropometrics and body composition. Transl. Psychiatry 11, (2021).
2. Kim, H. et al. Association between body mass index and cortical thickness: Among elderly cognitively normal men and women. Int. Psychogeriatrics 27, 121–130 (2015).
3. Medic, N. et al. Increased body mass index is associated with specific regional alterations in brain structure. Int. J. Obes. 40, 1177–1182 (2016).
4. Mihalik, A. et al. Multiple Holdouts With Stability: Improving the Generalizability of Machine Learning Analyses of Brain–Behavior Relationships. Biol. Psychiatry 87, 368–376 (2020).
5. Mihalik, A. et al. Canonical Correlation Analysis and Partial Least Squares for Identifying Brain–Behavior Associations: A Tutorial and a Comparative Study. Biol. Psychiatry Cogn. Neurosci. Neuroimaging 7, 1055–1067 (2022).
6. Monteiro, J. M., Rao, A., Shawe-Taylor, J. & Mourão-Miranda, J. A multiple hold-out framework for Sparse Partial Least Squares. J. Neurosci. Methods 271, 182–194 (2016).
7. Alexander-Bloch, A. F. et al. On testing for spatial correspondence between maps of human brain structure and function. Neuroimage 178, 540–551 (2018).
8. Markello, R. D. et al. neuromaps : structural and functional interpretation of brain maps. Nat. Methods (2022) doi:10.1038/s41592-022-01625-w.
9. Voigt, J. P. & Fink, H. Serotonin controlling feeding and satiety. Behav. Brain Res. 277, 14–31 (2015).
10. Frank, G. K. W. Advances from neuroimaging studies in eating disorders. CNS Spectr. 20, 391–400 (2015).