Histamine and Anorexia Nervosa

Most of us have at some point in our lives taken antihistamines–drugs that block the action of histamine (e.g., Claritin, Allegra)–to relieve allergy symptoms. And while histamine is best known for its role in the immune response, it also has many other important roles in the central nervous system.

In the brain, histamine release is important for arousal (this is why antihistamines tend to make us drowsy). It has also been implicated in regulating appetite, taste perception, learning, memory, aggressive behavior, motivation, and emotion, among others (Yoshizawa et al., 2009; see this quick summary).

Alterations in histamine signalling in the brain have been implicated in a variety of disorders, including schizophrenia (Iwabuchi et al., 2005), depression (Kano et al., 2004), and multiple sclerosis (Wikipedia has a nice summary chart; or you can read this open paper for more details, too).

Of particular interest to us here is the role of histamine in food and appetite control (see this open access review paper for a more detailed exploration). As summarized by Yoshizawa et al. (2009), it is been reported that

  • increasing histamine decreases food intake (Brown, Stevens, & Hass, 2001; Masaki & Yoshimatsu, 2006)
  • blocking histamine increases feeding behaviour (e.g., Orthen-Gambill and Salomon, 1992)
  • histaminergic activity is increased by food intake after starvation (Oishi, Itoh, Nishibori & Saeki, 1987; Itoh, Oishi, & Saeki, 1991)

Essentially, there appears to be an inverse relationship between histamine activity and feeding: High histaminergic activity suppresses food intake whereas low histaminergic activity increased food intake. Histamine’s effect on feeding appears to be mediated through the histamine H1 receptor (more on receptors below).

There are also findings suggesting that the histamine system is more active in women than in men, particularly with regard to the histamine H1 receptor (e.g., Ghi, Orsetti, Gamalero, & Ferreti, 1999; Kasoako et al., 2005; Prell et al., 1991). In particular, previous studies have shown that women have higher densities of the H1 receptor.

Given these findings, the authors of the present study wanted to see whether the histaminergic system is perturbed in women with anorexia nervosa relative to women (and men) without AN using positron emission tomography (PET) scans, focusing specifically on the H1 receptor. The authors also hypothesized that perturbations in the histaminergic activity in AN patients would be related to the extent/severity of abnormal eating behaviours or negative emotions.


If you have some background in neurobiology, feel free to skip this section.

Neurons communicate with each other by releasing chemicals called neurotransmitters (e.g., histamine, serotonin, dopamine, etc.). These chemicals exert their action on other cells by binding to specific receptors on those cells (receptors are picky; they don’t just bind any molecule that comes their way, and they bind to some molecules more readily than others). This binding can have many effects on the cell. For example, it can make those cells more or less likely to participate in neuronal communication with other cells, or it can turn on/off genes in that cell (among other things). There are four histamine receptors, creatively called H1, H2, H3, and H4, which are involved in different pathways and activities (see this chart).

When scientists talk about changes in histaminergic activity in the brain (or serotonergic or dopaminergic, etc.), they could be talking about a variety of different things that affect the overall system. For example, changes in the density of histamine receptors on a cell membrane, changes in histamine metabolism, changes in how readily histamine will bind to its receptor, changes in how much or how often the histamine is released from a cell can all affect histaminergic activity. These changes can then affect our physiological and behavioural responses. In the case of histamine, for example, this could mean affecting the extent of allergy symptoms we experience.


To assess histaminergic activity in the brain, the authors used the [(11)C]doxepin radioligand. Radioligands are radioactively labeled drugs “that can associate with a receptor, transporter, enzyme, or any site of interest. Measuring the rate and extent of binding provides information on the number of binding sites, and their affinity and accessibility” (Source).

PET studies frequently measure something called the binding potential. The binding potential is a combined measure of receptor density (in this case, the density of H1 receptors that are able to bind to the radioligand) and the affinity of the radioligand for the receptor. In this study, the authors assessed the binding potential of [(11)C]doxepin to study the H1 receptor.

I don’t read PET studies often, so I don’t know if this is common practice, but I thought it was interesting that the authors also took into account the female subjects’ menstrual cycle: All PET scans were performed within a week after last menstruation.


The participants were 12 female patients with anorexia nervosa (restricting subtype), 12 healthy female volunteers, and 11 healthy male volunteers. The participants did not have a history of any other psychiatric illnesses. The average age of the participants was early 20s. AN patients had an average illness duration of 5.2 years (range: 3-9 years), and had an average BMI of 14.7 (versus 20.3-20.4 for the two control groups). Patients with AN had higher anxiety and depression scores compared to the controls.


Confirming previous findings, the authors found that the binding potential of [(11)C]doxepin was higher in many brain regions in females than in males.

More interestingly, the binding potential of [(11)C]doxepin was higher in AN patients than in female controls in two regions: the left lenticular nucleus and the right amygdala; there were no regions where binding potential was lower in AN patients than controls.

Contrary to the researchers’ hypothesis, there were no positive correlations between the binding potential of [(11)C]doxepin and extent/severity of eating disorder behaviours, state and/or trait anxiety, and depression scores. In fact, there were several negative correlations. This means that the more severe the eating disorder behaviours, anxiety, or depression scores, the lower the binding potential (i.e., the more similar the binding potential is to female and male healthy controls).


As mentioned earlier, females have higher H1 receptor density than males. Interestingly, in rats, ovariectomy (surgical removal of the ovaries) decreases H1 receptor density and estradiol replacement reverses this decrease. In animals, food restriction increases histaminergic activity and reduces food intake. The authors suggest that maybe these differences play a role in increasing the vulnerability of women to develop AN:

Until now, CNS disturbances seen in AN were mainly considered to be secondary changes due to chronic starvation. However, in the present study, higher BP of [11C]doxepin in several brain areas was observed in normal female subjects. The risk of developing AN may be increased by not only the social background that women want to be thin because people tend to admire a thin figure but also biological vulnerability associated with central histaminergic activity.

It is an interesting thought, but for the time being, it is important to keep in mind that we really have no idea if and to what extent the differences in the histaminergic system play a role in the development of AN.

The higher binding potential in the lentiform nucleus (a region of the brain important for fine motor movement, among other things) and the amygdala (the so-called “fear center” of the brain) are also interesting. However, besides the fact that they have to be replicated by a different group, we don’t know whether this increased binding potential is the cause or result of AN; we also don’t know what it means.

The authors predicted the opposite effect based on findings in depressed subjects (who had lower binding potential). The authors put forth several possible explanations for this finding, but without knowing whether it is the cause or result of AN, and assessing only one time point (i.e., not studying the patients after they’ve recovered or at least restored their weight), it is impossible to really know what this means.


The authors mention several limitations to this study, including:

  • a small sample size (I am sure this partly due to cost, PET studies are expensive to conduct),
  • comparisons between many different brain regions (increasing risk for certain statistical errors),
  • inability to rule out that the results are due to brain shrinkage in AN (I’m not sure why this couldn’t have been controlled for, though)
  • and inability to tell whether the differences in binding potential are due to changes in receptor density or receptor affinity (remember, binding potential is a composite measure of both)


As with a lot of things I blog about here with respect to neuroscience or genetics, the takeaway is usually: This is cool research, but we can’t really conclude much. The authors provide support for the idea that individuals with AN may have higher levels of the H1 receptor in several brain regions, but these findings need to be replicated. And if they are, we’d still need to do more research to know whether the changes we see between AN patients and healthy controls lead to or are the result of the eating disorder. The typical cause or effect problem.

The brain is complicated, eating disorders are complicated, and our ways of studying both are generally very subpar. This research is preliminary, as is our general understanding of the neurobiology of eating disorders. My goal is to illustrate how complicated the situation is, how hard to study, and how difficult to have a nicely wrapped “story.”

Also: I am not a PET researcher, so if you know more than I do and spot some mistakes in my post or would like to provide additional feedback or thoughts on the study, please do so! I would very much appreciate that.


Yoshizawa, M., Tashiro, M., Fukudo, S., Yanai, K., Utsumi, A., Kano, M., Karahasi, M., Endo, Y., Morisita, J., Sato, Y., Adachi, M., Itoh, M., & Hongo, M. (2009). Increased brain histamine H1 receptor binding in patients with anorexia nervosa. Biological Psychiatry, 65 (4), 329-335

Liked it? Take a second to support Tetyana on Patreon!


Tetyana is the creator and manager of the blog. She has an Honours BSc in Neuroscience and an MSc in Medical Science. She can be reached at tetyana[at]scienceofeds[dot]org.


  1. Hmmm, interesting. Like you, I come away from this review with more questions than answers.

    Perhaps the negative correlation between histamine binding and severity of symptoms is just a by-product of decreased leptin levels, and not related to behavior in this case?

    I also find it interesting that no changes in BP in the hypothalamus were observed, which is maybe a region hypothesized to show differences? Some of the rat studies suggest this could be an important area for histamine’s regulatory effects on feeding. However, the review didn’t indicate that any direct targeting of the hypothalamus with histamine agonists/antagonists specifically altered food intake, just that food intake was associated with changes in histamine levels within the hypothalamus. Even the animal stuff (as reviewed in the Ishizuka report) seems to have a lot of open questions.

    To more research!

    • Yeah, the point about leptin is a good one. On the other hand, what explains the higher levels in the LN and amygdala.

      I found the discussion part of this paper lacking. Did you read Yoshizawa paper? I was confused by this:

      “Furthermore, estradiol has been reported to facilitate histamine-induced excitation of ventromedial hypothalamus neurons (60). Thus, female animals may adapt better to starvation through the central histaminergic neuron system than male animals.”

      Are female animals better adapted to starvation? And if so, how much of that would be the result of increased fat storage relative to male counterparts?

      I also wonder if the apparent increase in BP is not really real but is just due to lack of proper controls for brain shrinkage in AN/as a result of starvation? I have no idea; I don’t know anything about PET statistics/proper methodology, unfortunately. “A:)” said “you would need to account for volume effects. There are ways to do that and if they’ve done it it should be described in the methods,” but I didn’t see anything in the methods.

      Ah, I just don’t know enough at all about feeding and/or weight regulation stuff. (I don’t know, it is never been an interest of mine and I find it hard to motivate myself to read papers on it.) Or PET imaging for that matter. I really, really should.

      I picked this paper because someone on Twitter asked about histamine and EDs, and I was actually surprised to find papers on the topic. Now I’m kind of kicking myself for writing though because I feel so completely unqualified!

      • OK, so my comments as promised. . .

        1. 12 is a small sample size for a clinical group (but their total sample size is OK I guess, given they have a total n of 35 . . .).

        2. They do actually mention correction for volume effects when they say this: “Region of interest based analysis was conducted to evaluate brain H1R-BP, minimizing the effects of possible brain atrophy using original BP brain images instead of using transformed BP brain images into standardized brain space.”

        Now I’m not entirely sure if they mean that they’ve derived their BP values from (i) the raw summated PET image (i.e. PET image representative of total signal picked up by the PET scanner – i.e. “counts” – over the entire scanning period ), (ii) a “BP image” which has a specific meaning in the PET field or (iii) if they’re applying a usual region of interest (i.e. ROI) based approach after co-registering the PET/MR image.

        My understanding of the latter analysis (and I may be wrong) is that this is pretty standard (i.e. the original PET data is messed with as little as possible) and wouldn’t fully account for volume differences across groups. During analysis, after co-registration (basically super-imposing the PET and MR image so that they are aligned), it is usually the MR image that is transformed into standard space for spatial normalization. This information is applied to the PET image and ROIs are then placed on the PET image with ROI dimensions refined accordingly, to fit. So there is no actual transformation of the PET image itself during this analysis.

        There are validated methods of correcting for volume effects which they don’t seem to have mentioned these here – this is concerning given that there is evidence of grey/white matter volume loss in AN patients – especially in the acute underweight stage and I would also wonder about sex differences/brain volume. (i.e. they are comparing an entirely male group to 2 entirely female groups and in most studies, sex ratio is usually equal across groups)

        3. The brain regions they’ve chosen to investigate are for the most part, SMALL (ex. left amygdala) – especially when you take into account the fact that they’ve not looked at their regions bilaterally, but separated each into left/right. Usually in PET, the smaller the region you are looking to image, the more noisy/variable/less reliable the results . . .

        4. The way their figures are set up (i.e. the bar graphs and tables) it’s difficult to know if there are outliers driving the data . . .

        At least in my experience, these data are usually graphed as scatter plots with brain region on the x-axis, PET measure (i.e. binding potential) on the y-axis and each individual represented as a point on the graph – this way you have a feel for scatter within the group. The mean for each brain region is usually represented as a bar across each group of points (see Rekkas et al., 2014 doi: 10.1001/jamapsychiatry.2014.250, for a good example)

        Even if they choose to use bar graphs (which might be a stylistic difference), the actual mean binding potential values and corresponding SD should be reported in the tables (and maybe also the F-statistic or t-statistic?). Instead, all I see that they’ve reported is the p-value and brain region – which is . . . odd.

        5. Small point – but they mention they required their participants to be free of alcohol, nicotine, caffeinated beverages, 1 week prior to scanning, but they don’t mention whether this was validated using urine drug screening or (in the case of nicotine), plasma cotinine (a nicotine metabolite)/nicotine levels.

        Given that a large portion of the psychiatric population smokes and AN patients, in particular, might smoke for reasons related to weight control, this would have been important to control for or at least report – especially if there is evidence that it might affect their PET measure.

        • Thanks A:)!

          So it seems I missed the part about the correction for volume effects. Do people ever report corrected and uncorrected data so that one could see the effects of the correction? Is that a thing? Also, I’m not sure I get how the sex ratio here would affect this correction? I’m a bit confused there.

          I agree with you on #4. Why not use the best graph to represent the data?

          Compared to all of the PET studies you read, how would you rate this as far as methodology is concerned?

          • In regards to volume effects, people do definitely report on corrected/uncorrected data – especially when this might be relevant to the outcome/group under study (i.e. see Kish et al., 2010, doi: 10.1093/brain/awq103, for a good example/explanation of this)

            In terms of how the sex ratio would affect correction – It is just my observation that if you are going to compare an entirely male group (i.e. perhaps larger brain volume?) to an entirely female group, you should at least explore how volume effects may or may not affect your outcome of interest (i.e. PET measure) – especially in regards to the comparison between the clinical population of females with AN and the male group . . .

            As a 2nd point, the authors mention that the tracer used for this study (i.e. doxepin) has 2 binding sites – a high affinity binding site (i.e. for the H1-R) and a low affinity binding site (??).

            I don’t know a lot about doxepin as a tracer, past what is stated in the article, but usually, you want a PET ligand (i.e. doxepin) to be very specific in terms of binding ONLY to your target of interest (i.e. receptor, protein, enzyme, etc.).

            In support of the fact that doxepin is selective for the H1-R in humans with only negligible binding at other sites, the authors cite a rodent study in which H1-R knock-out rodents displayed negligible binding (although they do not specify a definition for “negligible”).

            However, the authors don’t seem to reference any human study which might also support this fact (as an example, a study could be conducted to determine whether H1-R BP would decrease – relative to a baseline scan with doxepin – after administration of an anti-histamine with high-affinity for the H1-R).

            As to how I would rate this study in terms of methodology . . . The points I raised in the previous comments regarding how the data has been presented actually concern me more than any methodological issue (apart from volume effects) – because it makes it difficult to evaluate the authors’ claims independent of their own interpretation. . .

          • Wow, super-comprehensive, A :)! If I remember correctly, H1 ligands also have an affinity for 5-HT receptors. I have no idea if the dose of doxepin used in this study is selective in humans, but it seems like some sort of binding-competition assay should have been performed in the past to determine specificity? Or are those studies too expensive, time consuming, and often not conducted?

          • Hey Liz,

            You are correct that H1 ligands also have affinity for the 5-HT receptor (and many other receptors) – hence its original use as a tricyclic antidepressant.

            Tetyana has provided the link for a table below and if I remember correctly, the binding affinity for the 5-HT2C receptor is highest, next to that of the H1-R.

            There are really 2 methods of determining specificity of doxepin for the H1-R.

            1. As you had mentioned, a preclinical binding-competition assay in some kind of animal model.

            2. A blocking study using PET with the doxepin tracer. This basically mimics the binding-competition assay described above:

            Step 1: Scan#1 with doxepin only to establish a baseline measure of BP in a small sample of individuals

            Step 2: administration of some sort of drug (an anti-histamine comes to mind here) that has higher affinity for the H1-R than doxepin.

            *Theoretically, during this second scan, as the drug is already bound to the H1-R, there should be negligible doxepin binding.

            If there is still substantial doxepin binding, this might point to the fact that doxepin is binding to other sites (i.e. as the drug is already occupying the H1-R receptor).

            Ideally, both methods should be used to validate the tracer.

            I hope this clarifies things.


Leave a Reply

Your email address will not be published.

12 − 12 =