1 INTRODUCTION

Obesity affects about 13% of the world's adult population (World Health Organization [WHO], 2020).1 According to the WHO, approximately 11% of men and 15% of women suffered from obesity in 2016. The prevalence of individuals with overweight and obesity has risen dramatically since 1975, and if the current tendency continues, by 2030, around 57% of the world's population will be affected by overweight or obesity.2 Globally, increased high-caloric food intake and reduced physical activity are two major hazards that promote obesity at a young age predisposing to the risk of numerous comorbid conditions and higher mortality in mid- and late-life.3, 4 Potential obesity-associated complications are cerebrovascular disease, type 2 diabetes mellitus, hypertension, several types of cancer, and dementia.5 Various neurological and psychiatric conditions have been consistently linked to structural alterations of the brain, which is also true for obesity. Multiple brain imaging studies showed altered activity in the reward circuitries that may predispose an individual to abnormal eating behavior and weight gain.6, 7 Although neuroimaging is a promising tool for detecting and examining brain abnormalities underlying obesity, causal links between brain alterations and obesity are still not clear.

1.1 Macrostructural gray and white matter changes in obesity

The most common measure used to classify the degree of obesity is body mass index (BMI, kg/m2). Numerous studies report associations between elevated BMI and volumetric brain changes in humans using volumetry and voxel-based morphometry (VBM) approaches.8, 9 Several studies reported volumetric gray matter (GM) reductions associated with increased BMI in the basal ganglia structures, as well as in the orbitofrontal cortex, insula, amygdala, hippocampus, and anterior cingulate cortex (ACC).10, 11 These brain areas have been previously associated with reward, motivation, and emotion processing, but also with cognitive mechanisms like memory, impulse control, and decision-making.5 Previous meta-analyses on GM alterations in obesity suggest that degeneration in these brain regions may underlie impaired internal feedback within the circuits associated with reward, emotion, and impulse control and, therefore, contribute to weight gain.12-14

In contrast to the cerebral cortex, the role of white matter (WM) has gained significant recognition only recently. Overall, compared with existing studies on obesity-related GM changes, there is less evidence for a comparable association with WM changes. Kennedy et al. showed reductions in WM volume associated with higher BMI in the anterior limb of the internal capsule and middle frontal gyrus.15 The anterior limb contains fibers projecting from the thalamus to the frontal pole and ACC, as well as subcortical connections between the caudate nucleus and putamen.16, 17 Those fiber tracts are involved in emotion processing, decision-making, and motivation. Moreover, reduced WM volume in the frontal lobe might explain a higher risk of developing dementia and cognitive decline in individuals with obesity.8 Inconsistencies in results and lack of research focusing on WM macrostructural abnormalities in healthy obesity limit the interpretation of existing findings. Besides, techniques investigating differences in brain volume are incapable of assessing the microstructural changes in WM that may underlie obesity-related pathology. Diffusion tensor imaging (DTI) is an increasingly popular neuroimaging method that is capable of detecting subtle alterations in WM fiber tracts on a microstructural level.

1.2 DTI technique

DTI is a magnetic resonance imaging (MRI) technique used to assess microstructural brain changes by measuring the diffusion of water molecules in WM tracts. The most frequently assessed DTI parameters are mean diffusivity (MD) also referred to as apparent diffusion coefficient (ADC) and fractional anisotropy (FA). MD or ADC reflects the mean water diffusion rate within a voxel, and FA indicates the preferred orientation of water diffusion in a voxel.18, 19 FA values vary between 0 and 1, where higher values indicate the preference of water to diffuse in one direction (anisotropic diffusion), whereas low FA values suggest that water molecules can diffuse equally in all directions (isotropic diffusion). Diffusion in WM tracts is anisotropic since it is restricted by axon fibers, while diffusion in GM and cerebrospinal fluid (CSF) is isotropic.18, 20 Additional parameters that can be derived include axial (AD) and radial diffusivity (RD) reflecting the rate of diffusion along axonal fibers and perpendicular to axonal fibers, respectively. Often in pathology anisotropy tends to decrease possibly resulting from increased RD or/and decreased AD. DTI is therefore a highly sensitive tool used to detect microstructural WM changes associated with pathological processes such as axonal damage, demyelination, and decreased fiber density.21, 22

1.3 DTI and obesity

A link between obesity and altered WM microstructure has been shown by several studies; however, the results are highly diverse.23-28 FA reductions in several major WM fiber tracts have been linked to numerous cognitive processes including reward-related behavior, cognitive and inhibitory control, or memory and decision-making.11, 24, 29-33 These findings are relevant for obesity research since cognitive impairment such as memory and learning deficits are often the case in people suffering from obesity.34 On the other hand, higher BMI or waist circumference (WC) were associated with increased FA in several studies.28, 35 Moreover, affected WM tracts vary across studies, which makes it challenging to interpret the pattern of changes.

Several factors could contribute to such mixed findings. One possible explanation is the type of sample examined. In particular, sex-dependent differences in WM microstructure using DTI were indicated by multiple studies. Shin et al. showed higher FA values in CC of healthy men compared with women.36 Other studies suggest that the degree of myelination, axonal fiber density, and axonal diameter as major modulators of FA vary in men and women.37-39 Another crucial factor is age. The majority of studies investigating age-related differences in DTI parameters report a negative association between FA and age accompanied by lower cognitive performance.40-42 Commonly reported areas affected by aging mainly include CC and prefrontal WM.43 Furthermore, a data analysis approach might contribute to the heterogeneity of results. DTI measures are commonly extracted by defining specific anatomical regions of interest (ROI) or by applying a whole-brain analysis. An ROI-based analysis is a hypothesis-driven approach and is based on the manual or automatic delineation of a priori specified brain regions. Whole-brain analyses can be performed by applying voxel-based analysis or tract-based spatial statistics. This approach is hypothesis-free assuming that structural changes can be observed anatomically anywhere in the brain. Since the choice of ROIs is subjective, interpretation of ROI studies results can be limited. Moreover, the examination of specific brain areas strongly constrains comparability across ROI and whole-brain studies.44, 45 Additionally, studies examining samples with a wider range of BMIs consistently demonstrate a relationship between the severity of obesity and DTI measures. Specifically, individuals suffering from morbid obesity display significantly lower FA compared with subjects with overweight.24, 46

Obesity-associated complications may further contribute to altered WM microstructure. Studies assessing microstructural brain alterations in individuals diagnosed with type 2 diabetes revealed WM abnormalities in several association fiber tracts including inferior and superior longitudinal and uncinate fasciculi (ILF, SLF, and UF).47 Although numerous DTI studies point out the importance of obesity-associated comorbid conditions, there is only limited evidence on WM microstructure in “healthy” obesity in comparison with the related disorders.

1.4 Aims

While the DTI method has been proven a powerful tool in detecting subtle microstructural WM abnormalities in obesity, existing results are inconsistent across studies due to the aforementioned factors. The current paper, therefore, aims to systematically review the DTI studies reporting structural alterations in individuals suffering from obesity with no history of neurological or psychiatric conditions compared with healthy lean control subjects.

2 METHODS 2.1 Search strategy

A systematic review was conducted on studies applying DTI in healthy lean controls and subjects with overweight/obesity following PRISMA 2020 guidelines. Full systematic searches on PubMed, Livivo, and Web of Science databases were performed using the following search items: “obesity” OR “obese” OR “overweight” OR “body mass index” OR “waist circumference” OR “waist-to-hip ratio” OR “body mass” AND “diffusion tensor imaging” AND “white matter.” The search included articles published between 1995 and 2020 and was restricted to English publications. The search and screening were performed independently by two investigators (LO and MH), and any disagreements were resolved via consensus Table 1.

TABLE 1. Demographic details of the studies selected for the current systematic review Study Number of subjects % of female subjects Mean age (SD) 1. Alarcon et al.61 152 44.1 14.1 (1.3) 2. Alosco et al.62 120 42 13.54 (2.9) 3. Augustijn et al.65 44 42.1 9.45 (1.1) 4. Bettcher et al.30 138 58.7 71.3 (6) 5. Birdsill et al.28 172 57 49.5 (6.4) 6. Bolzenius et al.55 62 67.7 62.4 (8.4) 7. Carbine et al.35 87 54.3 16.4 (2) 8. Chen et al.57 50 78 39.1 (9.3) 9. Dekkers et al.10 12,087 52.8 62 (7.3) 10. Figley et al.58 32 50 29.8 11. He et al.50 336 58.04 20.4 (1) 12. Karlsson et al.27 45 73.3 46.4 (9.2) 13. Kullmann et al.52 33 42.4 26.6 (3.4) 14. Lou et al.98 49 44.9 30.4 (7.7) 15. Marques-Itturia et al.53 63 60.3 25.3 (9.8) 16. Metzler-Baddeley et al.56 38 57.9 67.9 (8.6) 17. Mueller et al.46 49 46.9 26.4 (5) 18. Nouwen et al.63 53 81.1 15.8 (1.7) 19. Ou et al.64 24 50 9.5 (0.8) 20. Papageorgiou et al.23 268 57.1 43.3(16.1) 21. Repple et al.31 Dataset 1: 369 50.4 39.39 (11.34) Dataset 2: 1064 53.9 28.75 (3.68) 22. Ryan and Walther49 94 100 69.3(9.3) 23. Samara et al.59 46 76 29.8 (5.8) 24. Shott et al.11 42 100 28.05 (7.3) 25. Stanek et al.26 103 42.9 46.8 (16.5) 26. van Bloemendaal et al.54 48 45.8 58.5 (1.9) 27. Verstynen et al.25 28 60.7 30 28. Verstynen et al.32 155 49.7 40.7 (6.2) 29. Xu et al.24 51 41.2 29.6 (10) 30. Yau et al.60 60 60 17.3 (1.6) 31. Zhang et al.29 1255 50.7 55.43 (16.05) 2.2 Eligibility criteria and data synthesis

Studies matching the following inclusion criteria were included: (1) compared healthy controls and subjects with overweight/obesity; (2) used subjects with no metabolic or psychiatric disorders; (3) used DTI to assess WM; (4) was published as an original paper; (5) was published in English. The data from selected studies were extracted manually by two independent investigators. All available DTI parameters (FA, MD, AD, and RD) from each study were categorized based on the fiber tracts and summarized in Table 2.

TABLE 2. Summary of DTI results classified by fiber tracts Study Analysis method DTI measures Significant differences in DTI measures Association pathways Commissural pathways Projection and thalamic pathways Adults Verstynen et al.25 Whole-brain FA

↓ FA parahippocampal CG;

↑ FA ILF

↑ FA splenium of CC,

↓ FA MCP; SCP; bilateral ML; infundibulum; anterior limb of IC; CR;

↑ FA MCP; SCP

ROI ↓ FA hippocampal CG; SFOF ↓ FA body of CC ↓ FA pontine crossing tract; CT; ML; ICP; MCP; SCP; CP; anterior limb of IC; SCR Xu et al.24 Whole-brain FA, MD, RD, AD ↑ AD right SLF

↓ FA left and right body of CC

↑ left and right splenium of CC

↓ AD left body of CC

↑ right and left splenium of CC

↑ AD right CR

↑ MD fornix

↑ RD fornix

Dekkers et al.10 Whole-brain FA, MD

↑ FA (fibers not specified)

↓ MD

↓ MD

↑ FA (fibers not specified)

↓ MD

Figley et al.58 Whole-brain FA, MD

↓ FA right IFOF; left superior temporal gyrus

↑ MD right CG; left frontal WM

↓ FA genu of CC

↑ MD splenium of CC

↑ FA cerebellum;

↓ FA left anterior CR; left posterior TR

Ryan and Walther49 Whole-brain FA, AD, RD ↓ AD bilateral AF; right UF; left SLF ↓ AD genu of CC

↓ FA right CT and CBT

↓ AD anterior IC, bilateral superior CR; SCP; ICP

ROI

↓ FA bilateral temporal stem, right superior temporal WM

↓ AD right orbital, inferior and superior frontal WM, temporal brainstem

↑ RD right superior temporal and left medial temporoparietal WM

He et al.50 Whole-brain FA ↓ FA left and right CG Samara et al.59 Whole-brain FA, AD, RD

↓ FA (tracts not specified)

↑ AD, RD (tracts not specified)

Shott et al.11 Whole-brain FA, ADC

↓ FA IFOF; SLF; UF

↑ ADC right IFOF; ILF

↓ FA bilateral superior and anterior CR; EC; ATR;

↑ ADC superior CR

Repple et al.31 Whole-brain FA ↓ FA ILF; IFOF ↓ FA entire CC ↓ FA PTR, internal capsule; Papageorgiou et al.23 Whole-brain FA, MD, AD, RD ↓ FA right SLF; right ILF; right CG

↓ FA right CT; right

posterior TR

Verstynen et al.32 Whole-brain FA, AD, RD ↑ FA and ↓ RD (tracts not specified) ROI ↓ FA genu of CC ↓ FA left IFP; left and right SCP; left and right anterior CR Mueller et al.46 Whole-brain FA, AD, RD, ADC

↓ FA entire CC

↓ AD entire CC

↑ RD genu of CC

↓ ADC splenium of CC

ROI

↓ FA genu and splenium of CC

↓ ADC splenium of CC

↓ AD genu and splenium of CC

↑ RD genu and splenium of CC

Lou et al.98 Whole-brain FA ↓ FA bilateral CT; right brainstem Kullmann et al.52 Whole-brain FA, MD, AD, RD ↓ AD and MD right SLF

↓ FA right MCP

↓ RD right MCP

↓ AD and MD bilateral CT and anterior TR

Bettcher et al.30 Whole-brain FA ↓ FA cingulate section of CG ↓ FA genu, body and splenium of CC ↓ FA fornix Zhang et al.29 Whole-brain FA ↓ FA SLF; ILF; UF ↓ FA entire CC; ↓ FA CP; IC; CR; ATR; STR; EC; SCP Birdsill et al.28 Whole-brain FA, MD, AD, R ↑ FA SLF; CG ↑ FA body and splenium of CC; ↑ FA SCR; IC; EC; thalamic WM Marques-Iturria et al.53 Whole-brain FA No difference ROI ↓ FA in fibers connecting nucleus accumbens to other reward-network regions Van Bloemendaal et al.54 Whole-brain FA, MD, AD, RD ↓ AD forceps major Karlsson et al.27 Whole-brain FA, MD

↓ FA right IFOF

↓ MD IFOF; UF

↓ FA entire CC

↓ FA CT; posterior

TR;

Stanek et al.26 Whole-brain FA ↓ FA genu and splenium of CC ↓ FA fornix Bolzenius et al.55 ROI-A FA ↓ FA UF Metzler-Baddeley et al.56 ROI-M FA, MD, AD, RD ↑ AD, MD fornix Chen et al.57 Whole-brain FA, MD, AD, RD ↓ RD CC ↓ AD posterior IC Adolescents/children Carbine et al.35 ROI FA, MD, AD, RD

↑ FA left CG; IFOF;

↓ FA left IFOF; left uncinate fasciculus (UF), middle and temporal right UF

↑ MD right UF

↑ AD left CG; left UF

↑ RD right UF

↑ FA anterior frontal of CC; orbital CC;

↓ FA superior frontal CC

↑ AD anterior frontal and orbital CC

↑ RD superior and orbital frontal CC