Uncovering Neuroanatomical Networks Responsible for Abnormal Eating Behavior in Frontotemporal Dementia

Jennifer L. Whitwell, PhD¹

Magnetic resonance imaging (MRI) is a powerful tool that allows us to not only assess atrophic patterns associated with disease but to start to unravel the complex associations between regional tissue loss and specific clinical symptoms. Patterns of atrophy associated with frontotemporal dementia (FTD) and its clinical variants have been well described and studies have begun to demonstrate that different behavioral symptoms observed across the FTD spectrum have different anatomical loci.1^- 6 Changes in one’s eating behavior is a behavioral feature that is particularly hard to measure and, hence, to study accurately with neuroimaging. This could involve the tendency to overeat or cram food in the mouth, indiscriminate eating, oral exploration of inanimate objects, increased selectivity in food choices, or a preference for sweet foods.7 These behaviors are typically best captured with caregiver questionnaires since a loss of insight results in an underestimation of the presence and severity of the behaviors by the patient. These questionnaires are not perfect, however, and are limited by problems with overestimation or underestimation. In addition, the reduced quantitative scale often inherent to these questionnaires (eg, 4-point scale of normal, mild, moderate, severe) can also make neuroimaging analyses difficult.

In this issue of JAMA Neurology, Ahmed and colleagues8 aimed to overcome these difficulties by performing prospective well-controlled experiments to assess caloric intake and sucrose preference in patients with FTD, with the additional aim of assessing MRI correlates of performance on these tests. An ad libitum test meal was used to assess caloric intake and an experiment involving tasting of desserts of varying sugar content was used to assess sucrose preference. These experiments were performed in 19 patients with the behavioral variant of FTD (bvFTD) and 15 patients with semantic dementia (SD), as well as in 2 control groups consisting of 15 patients with Alzheimer dementia and 25 healthy normal controls. All participants also underwent a 3-T MRI. The authors found that caloric intake was only increased in the patients with bvFTD, while an increased preference for the sweetest dessert was observed in both bvFTD and SD. This increased preference for the sweetest dessert was not due to difficulty in perceiving sweetness since all groups could correctly perceive which desert was the sweetest. The MRI analysis was performed using voxel-based morphometry and voxelwise general linear model statistics to assess correlations between performance on these 2 tests and gray matter volume. A large number of overlapping regions were found to correlate with performance on both tests. Within the bvFTD group, increased caloric intake was associated with regions of loss extending from the temporal lobe (including inferior and middle temporal gyri and fusiform) to the occipital lobe, as well as the thalamus, hippocampus, parahippocampal gyrus, cingulate cortex, and cerebellum, particularly in the right hemisphere. These temporal regions and occipital lobe were also found to correlate with caloric intake in SD, although this analysis also identified a number of frontal regions, including the orbitofrontal cortex, and was more predominantly left sided. The correlation analysis for sucrose preference was performed across patients with bvFTD and SD and showed that sucrose preference was associated predominantly with regions in the temporal poles, frontal lobe, insula, striatum, and nucleus accumbens, but also the occipital lobe and cerebellum. The authors point out that the networks of regions identified share structures controlling cognitive reward, autonomic, neuroendocrine, and visual modulation of eating behavior. They also conclude that differing neural networks control eating behavior in bvFTD and SD and that a similar network controls sucrose preference.

This study should be applauded for performing such detailed and prospective assessments of eating behaviors in FTD. The findings extend and quantify previous observations of overeating in bvFTD4^,7 and show that a preference for sweet foods is observed in both bvFTD and SD, confirming results from this same group collected using caregiver questionnaires.9 These experiments provide quantitative data that are ideal for use in neuroimaging correlations. This is also the first study that specifically assessed MRI correlates of eating behavior separately within patients with bvFTD and SD. Previous studies either assessed only patients with bvFTD,2 a combination of patients with bvFTD and SD,3 or a combination of neurodegenerative diseases.4 These previous studies emphasized the involvement of the right orbitofrontal cortex, insula, and striatum in both binge eating and sweet preference,2^- 4 regions that have been related to reward-seeking behaviors.2These regions were identified in the Ahmed et al study8 but were observed as part of a much more distributed network. In addition, the orbitofrontal cortex was only observed in the correlations that included the patients with SD suggesting that patients with SD may have particular problems with decision making. This also begs the question of whether the patients with SD would have shown a different brain network associated with sucrose preference if they had been assessed separately from the patients with bvFTD. Behaviors in FTD have been typically associated with involvement of the right hemisphere,10 which was confirmed by the analyses within bvFTD. However, regions associated with caloric intake in SD were observed more in the left hemisphere, consistent with the fact that SD targets the left hemisphere in most patients. Findings from the current study that were somewhat surprising were the strong associations between both caloric intake and sucrose preference and the occipital lobe and cerebellum.8 This is surprising since these regions are not typically associated with bvFTD and SD, except in some specific genetic variants.11 The authors may have uncovered a previously unrecognized role for these structures in FTD, suggesting autonomic and visual contributions to problems with eating. However, one should be cautious in the interpretation of voxel-based morphometry results, particularly when the number of subjects is relatively small and when the threshold for atrophy determination is relatively lenient. The authors did perform a correction for multiple comparisons using the false discovery rate correction, which is perfectly acceptable and adds confidence to the findings, although this correction is a less stringent control of type I errors compared with the familywise error correction, which is also regularly applied in voxel-based morphometry studies. Another issue to consider when interpreting these findings is that the caloric intake and sucrose preference scores correlated with disease severity measures, and hence, it is difficult to be sure that the results are specific to these eating behaviors vs how much they reflect correlates of disease severity.

The concept that behavioral abnormalities in FTD result from a network of brain regions, rather than 1 or 2 specific structures, is certainly an attractive, although more complicated, neurobiological explanation for these behavioral manifestations. There is currently a great deal of interest in network theory within the field. The results from this study by Ahmed and colleagues provide the most rigorous assessment of this issue to date and will now need to be validated. An important future direction to help validate these results would be to directly assess functional connectivity in the brain. Task-free functional MRI could be used to investigate whether the regions implicated with each of these eating behaviors are indeed functionally connected and allow the assessment of how disruptions in these networks are associated with behavior.

JAMA Neurol. 2016;73(3):267-268. doi:10.1001/jamaneurol.2015.4496.