One man smothers everything his wife cooks in honey. Another takes three extra helpings at mealtimes. Yet another refuses to eat anything but McDonald’s and Kentucky Fried Chicken. These represent a smattering of the eating disorders seen in frontotemporal dementia patients at the research clinic of Neuroscience Research Australia in Sydney. Such behavior can lead exasperated caregivers to hide food and lock refrigerator doors. What machinations in the brain can explain these cravings? Rebekah Ahmed and John Hodges of the University of New South Wales, Sydney, measured dietary preferences and regional neurodegeneration in the brains of healthy controls and patients with various forms of dementia. As reported online January 25 in JAMA Neurology, they found that while dietary preferences among FTD patients correlated with atrophy of brain regions that control eating behavior, some of those regions are not typically associated with FTD.   

“Although changes in eating behavior and food preferences are salient features of frontotemporal dementia, the biology of these changes has not been well defined,” wrote Jason Warren, University College London, to Alzforum (see full comment below). “The work takes significant steps in studying the eating behaviors of FTD under carefully controlled conditions … [and is] one of the most comprehensive delineations to date of the brain networks that mediate abnormal eating behavior.”

Eton Mess.

Volunteers rated three increasingly sugary versions of this traditional English dessert made with whipped cream, meringue, and strawberries. [Photo by Vidya Crawley.]

Until now, insight into food disturbances in FTD came mostly from caregiver surveys (Ahmed et al., 2014). These suggested that at least 60 percent of patients with the behavioral variant of FTD (bvFTD) change their eating habits (Piguet et al., 2009). They overeat, mistake inedible items such as soap for food, or prefer sweets. These predilections help scientists distinguish bvFTD from other dementias such as Alzheimer’s disease. People with another form of FTD, semantic dementia, seem to be even pickier and eat unusual foods, though researchers haven’t studied this group carefully. Researchers also know that surveys capture just part of the picture because caregivers may downplay the problem and patients may hide their penchant from caregivers.

To get a handle on the extent of these eating problems, Ahmed and colleagues adapted methods validated in obesity research to observe patients eating in real time. The researchers recruited 19 patients with bvFTD, 15 with semantic dementia, and 15 with AD. Twenty-five healthy volunteers served as controls, and were matched for age, sex, and body mass index. A full day of tests started when participants who had been fasting for 10 hours were offered a buffet-style breakfast at the clinic. Foods on offer ranged from healthy to sugar-laden, including toast, eggs, muesli, orange juice, tea, milk, frosted cereals, and donuts. Participants had 30 minutes in a room alone to eat as much as they wanted. Since the researchers had weighed everything beforehand, they could calculate precisely how many calories each person consumed.

Later that afternoon, volunteers sampled three versions of Eton mess, an English dessert made with whipped cream, crushed meringue, and sliced strawberries, and traditionally served at that boarding school’s annual cricket match against rival Harrow (see image above). Three versions of the yummy treat boasted 26, 39, or 60 percent sugar. Participants ranked the samples on sweetness and taste, and then were left alone for 15 minutes with a large bowl of each to eat as much as they liked.

That same day, participants underwent structural magnetic resonance imaging. Using voxel-based morphometry, the researchers determined whether any regions of gray-matter loss correlated with caloric intake and a preference for sweets.

The results painted a striking picture of eating disorders in these patients. At breakfast, people with bvFTD consumed double the calories of any other group—an average of 1,344 versus 573, 710, and 603 for patients with semantic dementia, AD, and controls, respectively. All bvFTD patients overate. Though some semantic dementia patients also ate more than controls, many of them didn’t like the food choices, and some refused to eat at all. One patient, finding none of the fresh fruit she usually had for breakfast, ate only dried cranberries she picked from the muesli. Ahmed said the researchers were taken aback by how rigid the food preferences were among people with semantic dementia. Survey results suggest that if presented with their favorite foods, these patients may eat as much as people with bvFTD, she said.

All FTD patients preferred sweet tastes, not just people with bvFTD, as previous surveys had indicated. Most preferred the high- or medium-sugared dessert. Since these patients correctly ranked the desserts’ sweetness, the researchers ruled out inability to taste sucrose as a reason for gravitating to the sweetest option. In contrast to FTD patients, controls and AD patients preferred the least-sweet or medium-sweet option. Left alone to eat their fill, bvFTD patients once again ate more than anyone. Five of the SD patients turned up their noses on this course entirely.

Did neurodegeneration in any part of the brain correlate with this behavior? Rather than pinpointing discrete regions, MRI analysis implicated complex networks. Patients with bvFTD who consumed more calories had more atrophy in the anterior and posterior cingulate gyri, lateral occipital cortex, thalamus, and right cerebellum than did those who consumed fewer calories. These areas of the brain belong to networks that control decision-making and reward-based modulation of behavior, suggesting that the patients may struggle to evaluate options and don’t know when they are sated. The thalamus is well known to control appetite.

Atrophy of similar regions, mostly in the left hemisphere, associated with caloric intake among semantic dementia patients, as did atrophy of orbitofrontal cortices, the nucleus accumbens, and the amygdala. These patterns hint that people forget the names of different foods or that they compulsively seek certain foods that trigger defunct reward pathways. The correlation with atrophy of the amygdala suggests a person may have trouble controlling emotions related to eating. In both groups, preference for sweet tastes correlated with atrophy in a brain network previously reported to control sucrose preference in healthy people; it comprises the nucleus accumbens, temporal occipital cortex, and cerebellum.

The involvement of the occipital lobe and cerebellum in this study is surprising, since these are not typically associated with sporadic bvFTD and semantic dementia, wrote Jennifer Whitwell, Mayo Clinic, Rochester, Minnesota, in an accompanying editorial. “The authors may have uncovered a previously unrecognized role for these structures in FTD, suggesting autonomic and visual contributions to problems with eating,” she wrote. However, she urged caution when using voxel-based morphometry with relatively few participants and a lenient threshold for determining atrophy, as the authors did here. Whitwell suggested validation in another study, and functional MRI to see if these regions are functionally connected.

Warren noted that the study raises questions, including whether other eating signatures might correlate with specific genetic mutations, such as in the C9ORF72gene. He wondered how early these changes occur in the course of disease, and how shrinking brain structures shift network dynamics. “The work should provide a platform for a more systematic and physiologically informed study of these complex and highly clinically relevant behavioral changes,” he wrote.

Ahmed and colleagues are investigating how these dietary changes affect cholesterol levels and body mass index. Whether these food proclivities could be treated to slow the disease remains unclear. For example, elevated cholesterol and insulin resistance improve prognosis in ALS, which shares risk factors and pathology with FTD (for a review, see Ahmed et al., 2016). “Our natural inclination is to stop people from eating all these foods,” said Ahmed. “But maybe the body is adapting to changes in metabolic rate, and patients who eat more are trying to improve their prognosis.”—Gwyneth Dickey Zakaib


  1. Although it has been recognized for some time that changes in eating behavior and food preferences are salient features of frontotemporal dementia and semantic dementia (indeed, this is included among the defining consensus criteria for behavioral variant frontotemporal dementia), the biology of these changes and the brain basis for them has not been well defined. The work of Ahmed and colleagues takes significant steps forward in studying the eating behaviors of FTD quantitatively under carefully controlled conditions with test meals (rather than simply by historical reports), in a well-characterized patient cohort, with detailed structural neuroanatomical (MRI) correlation. Their study design has allowed the authors to present one of the most comprehensive and neurobiologically relevant delineations to date of the brain networks that mediate abnormal eating behavior. These networks show some specificity for particular syndromes (behavioral variant frontotemporal dementia and semantic dementia), which may hold a clue to the clinical profiles these patients present. However, the networks implicated are very extensive, including areas involved in the sensory processing of food, associating it with biological and emotional meaning, monitoring internal homeostatic signals (such as hunger and satiety), and programming complex behavioral responses such as food-seeking. The work should provide a platform for a more systematic and physiologically informed study of these complex and highly clinically relevant behavioral changes.

    At the same time the work raises a number of further questions—such as, can even more biologically specific eating signatures be identified (for example, in association with particular genetic mutations), how early in the course do eating changes signal FTD, how do the structural brain changes relate to dynamic functional changes in brain networks, and can the latter be measured in a way that would allow us to assess the impact of interventions on underlying physiology and metabolism, while damage is still recoverable? Addressing these questions will require longitudinal study of larger (including genetic) FTD cohorts, ideally with task-based as well as task-free functional MRI so that we capture the brain at work. We will also need to remain alert to clinical observations made in individual patients.

    Some of the work from our group on this topic complements the work of Ahmed and colleagues. For example, we have shown that some patients with FTD lose the meaning of flavors and food more generally (Piwnica-Worms et al., 2010; Omar et al., 2013). Patients with Alzheimer's disease often seem to show reduced appetite and lose weight for reasons that are not clear, and we have recently seen two patients with semantic dementia who have food aversion rather than overeating. Observations like these will need to be reconciled with the data from Ahmed and colleagues if we are to arrive at a really comprehensive picture of eating behavior in different dementias.


    . Flavour processing in semantic dementia. Cortex. 2010 Jun;46(6):761-8. Epub 2009 Jul 14 PubMed.

    . Flavour identification in frontotemporal lobar degeneration. J Neurol Neurosurg Psychiatry. 2013 Jan;84(1):88-93. PubMed.

Make a Comment

To make a comment you must login or register.


Paper Citations

  1. . Quantifying the eating abnormalities in frontotemporal dementia. JAMA Neurol. 2014 Dec;71(12):1540-6. PubMed.
  2. . Sensitivity of current criteria for the diagnosis of behavioral variant frontotemporal dementia. Neurology. 2009 Feb 24;72(8):732-7. PubMed.
  3. . Amyotrophic lateral sclerosis and frontotemporal dementia: distinct and overlapping changes in eating behaviour and metabolism. Lancet Neurol. 2016 Mar;15(3):332-42. Epub 2016 Jan 26 PubMed.

Further Reading

Primary Papers

  1. . Assessment of Eating Behavior Disturbance and Associated Neural Networks in Frontotemporal Dementia. JAMA Neurol. 2016 Mar;73(3):282-90. PubMed.
  2. . Uncovering Neuroanatomical Networks Responsible for Abnormal Eating Behavior in Frontotemporal Dementia. JAMA Neurol. 2016 Mar 1;73(3):267-8. PubMed.