A Flooded Basement

Last summer, the basement of our first floor apartment flooded. I had (stupidly) stored some of our stuff in cardboard boxes down there, which predictably became covered in mold, as did some of the sections of wall that had been exposed. I took all of the moldy boxes out to the trash, which required bringing them up and through my apartment. The next day I started having a bad sore throat and fatigue, which stuck around and didn’t go away for months. Under the assumption that the issue was the mold in that particular apartment, I tried moving, but with no luck. My symptoms persisted. It turns out I was not allergic just to mold but to a panoply of allergens, both indoor and outdoor: molds, dust, cat, dog, cockroach, timothy grass, Norway maple, and more. Somehow that one exposure had triggered a whole range of allergies that I had never had an issue with before. My allergist recommended allergy immunotherapy, or AIT.

The idea of AIT is both simple and paradoxical: if you are allergic to something, regularly take a chemical extract of the allergen in order to attenuate the allergy. Upon hearing about this I was intrigued and made the mistake of asking why this would work, naively assuming that there would exist an answer to this question. This essay chronicles my attempts to find one.

Phenomenology and Effectiveness

First off, does AIT actually work? There is a strong consensus that there are many cases in which it works, though there’s debate around the edges about in what cases it works, how much it works, and why it works. Allergic rhinitis (nasal symptoms) and asthma are definitely treatable, but atopic dermatitis (like poison ivy) or food allergy are more debatable.

While much of the data on clinical outcomes is self-reported symptom and medication use (with all of the associated methodological issues there), there is also relatively objective data like reactivity determined by skin prick test that is generally in concordance with the self-reported data. There are also studies claiming that AIT prevents the development of new allergies to other allergens, though a systematic analysis from Di Bona et al claims that almost all of the studies finding this were at serious risk of bias due to inadequate controls and rated the overall quality of evidence in favor as low.

Astute readers may note that AIT follows one of the core principles of homeopathy (“simila similbus curentur”: like cures like). As a coincidence, Charles Harrison Blackley, who first discovered that hay fever is caused by pollen, was a homeopath. He did not, however, use a homeopathic approach to treat his hay fever, instead constructing various devices that would obstruct the contact of pollen with his nasal epithelium. He settled on a thin metal covering of the inside of the nose with a 4 millimeter hole in it for breathing. This worked, as he tested by walking through hay fields with visible plumes of pollen. Though it was not without tradeoffs: “it was necessary to train the sensitive mucous membrane to bear the daily contact of the instrument for eight or ten hours at a time during the whole of the hay-season… it was found that it was not possible to secure perfect freedom of breathing and complete interception of the pollen at one and the same time.” Anyway, it wouldn’t have mattered if Blackley had discovered AIT since nobody believed him about pollen, instead “attributing the symptoms to infection, odorous emanations of plants, or hysteria”.

There are many ways to administer AIT. The most popular is subcutaneous immunotherapy, or SCIT. If you or someone you know got “allergy shots”, say for a cat or dog allergy, this is what they got. SCIT is effective, but very inconvenient for patients. You have to make an appointment with an allergist once a week every week for three years. More recently, sublingual immunotherapy, or SLIT, has gained popularity. Instead of having to go to the allergist, you can self-administer drops under your tongue daily. At first this approach was resisted because it was considered less effective, but it has gained widespread use and the overall state of the literature appears to support it being equally effective to SCIT.

So it works. But how? There are three questions that are extremely important but left unasked in any of the accounts I read on the mechanism of AIT.

  • First: AIT is mechanistically identical to vaccination (and in fact, is called allergy vaccination in much of the literature). However, on a high level, it has the opposite effect. With a normal vaccine, you initially do not have an immune response and the vaccine gives you one. With AIT, you initially have an immune response, and the AIT takes it away. Where does this difference come from? In this frame, the question of mechanism really becomes two questions. How do you tell the difference between AIT and natural allergen, and what changes in order to attenuate the allergic response? Most discussions of mechanism focus only on the latter.

  • Second: AIT involves exposure to a chemical extract of the allergen you are allergic to. However, allergen sensitization also also involves exposure to those same chemicals. If it’s not a function of the presence or absence of a specific compound, what is the key difference between AIT and natural exposure that leads to the downstream opposite effects of sensitization versus tolerance?

  • Third: AIT will work even when it’s given in conjunction with continued environmental exposure. For instance, you can get allergy shots for cat dander while still owning a cat. Why is that?

Crash Course in Immunology

“Immunology is where intuition goes to die.” - Ed Yong

Before we can address those points, I’m going to summarize as briefly as possible just the parts of immunology you need to understand for that section to make sense. If you’re already familiar with immunology, you can skip over this.

The immune system is split broadly into two components: the innate system and the adaptive system. The innate immune system consists of cells and receptors that recognize general signals that something is wrong. The adaptive immune system, on the other hand, consists of cells that custom-manufacture immunity to particular pathogens. In the case of a COVID infection, the innate immune system might recognize viral RNA generally, while the adaptive immune system recognizes the particular form of some component of the spike protein on that particular virus, but not related viruses.

The adaptive immune system has two main classes of cells: B cells and T cells. B cells produce antibodies, which recognize and bind to the three-dimensional structure of proteins, while T cells are presented shredded up bits of protein and recognize their sequence (one-dimensional structure). B cells and T cells typically live in the lymph nodes. They are alerted to what is happening in the rest of the body by antigen presenting cells, who find bits of potentially interesting material and bring them to the lymph node for inspection.

Antibodies are shaped like a “Y”. The top two parts of the Y (the variable domains) are what binds to the specific target that your body cares about. There are a combinatorially massive number of possible configurations of the variable domain since you should be able to generate a response to theoretically any foreign protein1. The bottom of the Y (the constant domain) is the part that interfaces with the rest of the immune system. There are 5 main classes of the constant domain– IgM, IgD, IgG, IgE, and IgA– though for our purposes we will only need to worry about IgG and IgE.

These two types of antibodies are components of two broadly distinct systems of related cell types and signaling molecules: the type 1 and type 2 responses. The type 1 response is typically involved in responses to intracellular pathogens, like bacteria or viruses. Type 1 responses involve IgG. If you have developed immunity to COVID, either by exposure or immunization, you probably have IgG antibodies with a specificity for the spike protein on that virus. Type 2 responses, on the other hand, mediate immunity against parasitic worms and in allergies and involve IgE antibodies. If you have allergies you probably have IgE antibodies that recognize the protein you’re allergic to. (Unfortunately these are both oversimplifications that we will get back to complicating later.)

Each antibody class has a corresponding set of FcRs (fragment constant receptors2) that will recognize it. Annoyingly, the receptor names use Greek rather than Roman characters to indicate which antibody class they correspond to, so IgG binds to FcγR, and IgE binds to FcεR.

Mast cell degranulation, image by [John Kimball](https://www.biology-pages.info/M/MastCell.gif)

Mast cell degranulation, image by John Kimball

Most types of antibodies, including IgG, spend most of their time circulating through the blood and lymph. IgE, on the other hand, hangs out bound to FcεR receptors on the surface of mast cells. Those mast cells live in the skin, where they quickly come into contact with allergens upon exposure. When mast cells recognize an allergen using the IgE that is bound to their surface, they “degranulate”, spilling out compounds that both directly harm whatever is bearing the allergen as well as cause an inflammation response and recruit other immune cells to the site. This is what causes the immediate swelling and redness characteristic of an immediate hypersensitivity reaction, which in an extreme form can turn into anaphylaxis. One of these compounds is histamine. Most common allergy medications are antihistamines, which block histamine receptors, attenuating the downstream signaling effects of the degranulation.

Attempted Explanations of Mechanism

Type 1 vs Type 2 Immunity

Most of the literature on mechanism is about describing in detail at a biochemical level the shift from type 2 (allergic) immune response to a type 1 (intracellular) response. There used to be a bunch of literature demonstrating a “see-saw” tradeoff between type 1 and type 2 immunity, where each system repressed the other. It is now standard to point out that this view is no longer accepted (in much the same way that cognitive science papers must ritually flog behavioralism lest they be suspected of promoting it). Now the dominant narrative is that type 1 immunity is built up in addition to type 2 immunity.

One of the shifts involved in moving from type 2 to type 1 immunity is moving from an IgE-dominated antibody response to an IgG-dominated antibody response. There are a ton of papers all over the literature that report increases in IgG, and more specifically the subtype of IgG known as IgG4 (yes, IgG has 4 different subtypes, sorry), so this phenomenon seems robust. The most popular explanation is that IgG competes with IgE for binding the antigen- if IgG is bound, that means it can’t bind IgE, and therefore can’t stimulate mast cell degranulation, preventing allergic symptoms3.

The evidence gets weaker when you try to correlate the IgG response to actual clinical outcomes. One group found no correlation between specific IgG1, IgG2 or IgG4 and symptom scores despite significant rises in all of those over the course of treatment. They could only get a correlation when looking at the ratio of IgG4 to IgG1, but that sounds like p-hacking since there is no theoretical motivation behind that quantity. Another group found that allergen-specific IgG4 levels have no correlation to clinical outcomes, though they do find a correlation between clinical outcomes and functional IgE suppression for two different assays where IgG4 and IgE competed for allergen binding. Djurup and Malling 1987 found that high IgG4 was strongly associated with failure of AIT.

This account does answer our question about normal vaccination vs allergy vaccination. Both of them function by inducing an IgG response, it’s just that in the case of ordinary vaccination that leads to immunity to a pathogen, whereas in allergy vaccination it leads to cancelling out the allergy. The main problem with this explanation is that it’s missing the most important part. We know phenomenologically that allergic responses shift to non-allergic responses as time goes on, and this accounts for that, but it does not explain what features of the stimulus induce the downstream changes.


While there may be specific differences depending on the specific way in which a particular extract is prepared, the primary distinction between an environmental allergen and a chemical extract of the same allergen is its size. For instance, pollen particles are in the range of 10-100 microns wide, significantly larger than a solution of the constituent allergenic proteins. The innate immune cells that directly interact with allergen-bearing particles upon exposure (as opposed to downstream T cells and B cells) can detect how large something is and change their response accordingly. Specifically, the type 2 system, responsible for allergies, also functions in helminth immunity. Helminths are multicellular animals significantly larger than phagocytic cells. The immune system can’t rely on phagocytosis and instead need to use other strategies- check out this video of eosinophils attacking a parasite by swarming it and degranulating around it. This suggest a simple rule: small things induce a type 1 response, since they can be phagocytosed, and big things induce a type 2 response, since they cannot.

However, there’s a problem with this hypothesis: it is common practice in animal models to induce sensitization using an allergen extract. This is true for house dust mites in both mice and sheep, ragweed in mice, and ovalbumin (egg protein), the most commonly used allergen. Since both native allergen and extract can sensitize, that rules out any explanation based on systematic structural differences between the two.

Route of Administration

Perhaps the difference is in the route of administration. There is a lot of literature on the phenomenon of oral tolerance, for instance, which argues that oral ingestion provides cues that prevents the development of allergy. In this story, children will develop allergies to foods if they are exposed to the allergic compounds externally (by skin contact) prior to eating them. The generalized version is that if you were sensitized to an allergen by exposure in one route, you can become desensitized by exposure via a different route.

That doesn’t seem to necessarily be the case for AIT, though. AIT for pollen administered by nasal inhalation, the same as the natural route of exposure, is effective against rhinitis (that is, nasal symptoms), though not against asthma.

A bee sting is effectively an injection, which is the same as the standard route of administration for bee venom AIT. Or is it? A bee sting and an allergist’s needle may get to different parts of the skin. The skin has multiple layers: the epidermis, the dermis, and the subcutaneous space (or hypodermis). In humans, the epidermis is 0.05-1.5 mm thick, and the dermis is 0.3-3.0 mm thick. Wasp stingers average about 2.5 mm in length whereas needles for subcutaneous injection are more like 10-15 mm. Figure 7 of this paper indicates that bee and wasp stings penetrate about 0.4 to 0.7 mm into pig skin. This implies that stings will go into either the the dermis or possibly epidermis in very thick skin, but not into the subcutaneous tissue, where standard AIT treatment is administered. How much of a difference does this make? Theoretically a large one, as mast cells typically reside in the dermis. However, the dermis and the subcutaneous space are adjacent to each other. Depending on how much diffusion there is from the injection site, it may not make much of a difference where exactly the venom goes.

I couldn’t find any papers directly addressing the question of bee venom spread. We can estimate it by looking at the diffusion of other proteins throughout the dermis, but there is an added catch. A major component of bee venom is hyaluronidase, an enzyme that breaks down hyaluronic acid in the extracellular matrix. This makes it easier for the venom to diffuse through the skin, so we need to take that into account. Looking at charts of diffusion over time for hemoglobin (64 kDa compared to 18 kDa for hyaluronidase), it is possible for proteins to diffuse through the dermis, either permeabilized with hyaluronidase or not, on a ~10 mm scale in the range of tens of minutes. This is both quickly enough that there should be an interaction but not so quickly that there’s no obvious distinction. So the intradermal and subcutaneous injections are different routes but also not so different as to be unambiguously separate, as oral vs intramuscular would be for instance. On net though, I will count this in favor of the route of administration theory.

Spread of hemoglobin through tissue from [Hayes and Reed 1950](https://journals.sagepub.com/doi/abs/10.3181/00379727-75-18197)

Spread of hemoglobin through tissue from Hayes and Reed 1950

The intradermal route in particular is relevant to allergen research as that is where the standard sensitivity test is done- so the gold standard of sensitivity to an allergen is specifically dermal sensitivity. There are some papers exploring intradermal AIT for pollen sensitivity, but unfortunately none of them look compelling. A 1933 paper claims 92% improvement and has a large sample (N=322) but has no control arm. More recently, one paper had positive results on cutaneous late skin tests but had no measure of clinical impact, another found no difference on combined symptom-medication score and worse outcomes in some of the secondary outcomes, and another claimed positive results on medication score but not symptoms, which is suspicious.


As with all medicine, perhaps the issue is simply one of dose. When it comes to chemical concentration, quantitative differences lead to qualitatively different outcomes. There are a couple of options here.

First, the AIT dose could be lower than the environmental dose. In this case, the lower dose acts as a sort of progressive training, often called Mithridatism4. However, AIT seems to work regardless of whether you do prior to exposure or in conjunction with exposure. For instance, pollen AIT can be taken during pollen season and still work, and cat dander AIT can be taken while owning a cat. This is odd. If the environment exposes you to 1000 (arbitrary) units of X, and you take 1 unit of X, you now have been exposed to 1001 units of X. That shouldn’t make a difference, since it should fall well inside the range of variation in environmental exposure.

It is also not clear what exactly is changing in response to this increasing stress. In muscle development from exercise, it is at least partly due to overcompensation of muscle cell size and count in response to (though I have not looked deeply into this and suspect if I did that it would be much more complicated than that). For the immune system, it is less clear what that compensation would look like. Proliferation of regulatory cells? Deletion of memory cells that recognize the antigen? Something else?

Perhaps instead the dose is significantly higher than environmental exposure. Naively, it seems like this couldn’t work, since it would immediately induce a severe immune response. Not necessarily. Very high dose exposure to an antigen can induce tolerance to that antigen, a phenomenon relevant for vaccine design, antibody engineering, and clinical administration of biological drugs.

So, what is the dose difference? This is a surprisingly difficult question to answer. While it is relatively straightforward-ish to define the dose of allergen administered in immunotherapy (though this is stymied by a lack of standardization across extracts), it is basically impossible to compare that to the environmental exposure for many allergens such as pollen and house dust mites. Exposure varies drastically depending on the season and the specific activities and living conditions of the individual, and the sensitization may have happened in a different environment than where they are currently living.

Despite that, we can still make estimates of the environmental vs AIT doses of house dust mite exposure. To get numbers that directly compare, I will look at Der p 1 antigen, that is, the primary allergenic compound in the common house dust mite Dermatophagoides pteronyssinus.

First, how do we measure the amount of environmental exposure to dust mites? There are multiple methods available for sampling dust: vacuuming the area, sticking adhesive to a surface, running an air pump with an electrostatic filter, or having subjects wear an intranasal air sampling device. Intranasal sampling has the benefit of measuring more directly the air that actually gets inhaled, but does that at the cost of lower volume of material sampled and greater discomfort for subjects who have to wear the device.

Gore et al use intranasal sampling to estimate that the yearly exposure to Der p 1 during sleep is 2 μg/year, which they compare to another study using personal air samplers which would imply (under the same assumptions of generalization) 0.2 μg/year of exposure. There are some issues with this number though. First, another site reports that due to the large variation in allergen per particle, the assay they use is only reliable at counts greater than 30, which only 8 out of their approximately 300 measurements hit. So this measurement should be regarded as a very rough estimate. Second, makes the assumption that the primary source of exposure to dust mites over the course of the day is while sleeping in bed. A different study attempted to measure this assumption by tracking the entire day using a very clever impaction sampler mounted to a clock motor and worn on the body like a bandolier, and found that sleeping only accounted for 10% of daily average exposure. Naively multiplying Gore’s values by 10 to correct for this gives a range of 2-20 μg/year.

Intranasal pollen sampler from [Graham et al 2000](https://www.sciencedirect.com/science/article/abs/pii/S1081120610624106)

Intranasal pollen sampler from Graham et al 2000

Impaction sampler from [Tovey et al 2016](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0153414). The impaction plate is mounted to a clock, which rotates over the course of the day, meaning that each part of the plate can be analyzed separately and cross-referenced against the subject's activity at that time.

Impaction sampler from Tovey et al 2016. The impaction plate is mounted to a clock, which rotates over the course of the day, meaning that each part of the plate can be analyzed separately and cross-referenced against the subject’s activity at that time.

How does that range compare to what you would typically get in AIT treatment? This is tricky since standardization of HDM extracts in the US is done by AU (active units, a measure of skin response to the extract), not concentration of major allergen. However, Laurenas-Linnemann et al have a good paper on comparing maintenance doses from different extracts. Using the figures in their paper, I estimate that the recommended dose comes out to 10-100 μg / year5. This puts the environmental and AIT doses in the same ballpark, though with the AIT dose potentially higher.

Bee venom is a simpler case than aerosol exposure. Though it the natural and artificial routes of administration are not identical, as discussed above, they are close enough that it may be meaningful to compare. Typical bee stings deliver about 50-200 μg of venom. A standard procedure for venom AIT calls to start with 0.1 μg, increase to 1 μg 30 minutes later, then 10 μg 30 minutes after that, then doubling the dose once a day until you hit a maintenance level of 100 μg. This is perfectly consistent with the Mithridatism story- start below, and train up to the full environmental dose.

Poison ivy is potentially similar (though poison ivy is an exceptional case, which we will discuss later)- while it is difficult to quantify how much allergen is directly transferred from the leaf to the skin, we can measure how much extract it takes applied to the skin in order to induce a reaction. Epstein 1974 reports a maintenance oral dose of 1-2 mg per day, compared to observing reactions from 5 μg of extract applied to the skin- a 20-40x difference. Is this comparing apples to oranges? On the one hand, both the external skin and the GI tract are epithelia that can absorb urushiol, triggering an allergic response. On the other hand, the gut mucosa and skin are very different environments physically and immunologically. Notably, the gut has a lot more mucous than the skin. I was unable to find any papers explicitly addressing this difference in the context of poison ivy allergy, though.

In cases where we can’t precisely define the relationship between environmental dose and AIT dose, maybe we can gain some insight from the range of possible AIT doses. In a review of SLIT dosing in allergic rhinitis, Linda Cox states: “The literature from 1986 through 2007 shows approximately a 6000-fold range in doses found to be effective with SLIT. However, recent studies in large patient populations have demonstrated a clear dose response with an effective dose range that appears to be equivalent to one to two times the monthly subcutaneous immunotherapy dose administered daily or weekly (ie, 15 to 30 microg of major allergen).”

The 6000-fold range is interesting. Perhaps it falls entirely above or below the environmental dose, but it seems plausible that it falls on both sides, which would definitely be an issue for the dose theory.

Theoretically there could be a difference in the temporal distribution of doses. If you get a fixed amount at a particular time every day, that would be different from an unpredictable environmental exposure. It is hard to square this with the variation in treatment protocols, though. SLIT is taken daily, SCIT usually weekly, and intralymphatic AIT can be monthly.

The Curious Case of Poison Ivy

There is another snag in the AIT story: poison ivy. Earlier I said that allergic responses were driven by IgE antibodies. This is only true in the typical case of an immediate (type I) hypersensitivity reaction. Delayed (or type IV) hypersensitivity reactions, such as reactions to poison ivy, do not involve an antibodies and are instead driven by a T cell response. The allergen in poison ivy and poison oak is urushiol- an oily resin formed of a collection of related small molecules that can easily bind to proteins in the cell membrane. When these proteins are degraded and presented by APCs to T cells, they recognize the urushiol and trigger a Th2 response. What is recognized is the urushiol bound to some protein (known as haptenization), not the urushiol on its own. The reason the process is delayed is that the T cells are not resident in the tissue (unlike IgE-bearing mast cells), so they need to migrate from the lymph nodes to the affected skin.

Many reviews of the topic state that AIT is specifically for IgE-mediated allergies. This makes sense under the standard class switching mechanistic model: if AIT functions by skewing the Type 1/Type 2 response by changing the specific IgG/IgE ratio, a mechanism that involved only T cell responses would not be impacted. If urushiol allergies are not driven by an antibody response, and AIT does work on them, then that could be a problem.

But does AIT for urushiol work? The oldest tradition of urushiol therapy goes back to Californian Native Americans, who would chew poison oak or drink tea prepared from it in order to induce immunity, a practice which was sometimes adopted by Californian settlers. On that basis alone though, it can be difficult to tell whether the consumption had a causal therapeutic effect or just a selective effect. If people who have severe allergies to urushiol simply die on ingesting it, then the remaining individuals will overall have less severe allergies. Indeed, one 1942 report mentions occasional fatalities as a result of the practice.

French physician André-Ignace-Joseph Dufresnoy reported in the 1780s that he was able to prevent rashes in his patients by brewing teas using up to 12 leaves of poison ivy with only mild side effects6.

Several papers report New Jersey physician Dr R. Dakin as being the first to report the benefits of consuming poison ivy in 1829. However, this is what he had to say on the practice: “Some good meaning, mystical, marvellous physicians, or favoured ladies with knowledge inherent, say the bane will prove the best antidote, and hence advise the forbidden leaves to be eaten, both as a preventive and a cure to the external disease. I have known the experiment tried, which resulted in an eruption, swelling, redness, and intolerable itching, around the verge of the anus.” Sounds more like ridicule to me.

Urushiol consumption was also allegedly used for a decidedly different purpose in Japan. There are reports of Buddhist monks who wanted to become sokushinbutsu, or “a Buddha in this very body” and decided to intentionally go through a process of self-mummification in order to achieve enlightenment and prevent rebirth. One part of the extremely painful process is to drink tea prepared from the bark of the urushi tree. The tea would induce vomiting, drawing liquid out of the body, and cause buildup of urushiol in the tissues, theoretically helping prevent the decomposition of the body after death. Unfortunately, this effort is reported not to have been particularly effective. Most attempts at self-mummification did not adequately preserve the body, and after 1000 days of entombment, would be judged a failure, dooming the aspiring Buddha to rebirth in samsara.

So historical reports of the effects of oral consumption of urushiol are a mixed bag, sounding potentially useful but with common bad side effects. What about scientific studies? Also mixed. The literature is filled with small studies, most of which indicate positive results, but some of which disagree. For instance, Zisserman reports no effect on poison ivy in Boy Scouts given tablets containing 0.01 mg of urushiol. However, Epstein reports effects from oral AIT with maintenance doses of 1-10 mg per day, so the issue may be one of dosing. However, naively comparing dose information across studies is difficult due to the variability of quality in different extraction processes and lack of adequate standardization.

When searching for answers to this on Google, it helpfully provides a “featured snippet” answer card linking to a Backpacker article which claims that a 1987 study demonstrated that consuming poison ivy did not induce immunity. However, they misunderstood the study- it was trying to induce immunity using two compounds present in urushiol that were then chemically modified in order to make them less allergenic. The same group had previously reported success in hyposensitization with unmodified urushiol. (Always be wary of people “citing studies,” including me.)

At least one group in the US regularly does SLIT for poison ivy. They report subjective clinical improvement in 90% of patients as well as an average increase in concentration of antigen required to cause a reaction of 125x, though I was unable to find the conference proceedings at which they presented these findings.

It is also unclear how much publication bias is an issue here. The same 2019 paper mentions as an aside that “previous poison ivy vaccines, comprising urushiol in sterile vegetable oils injected SQ [subcutaneously], were withdrawn in 1994 for failure to demonstrate statistical efficacy (personal communication J. Slater).” I was unable to track down any record of what group or company may have been working on that. It is also unclear without further context what dose they were claiming to use, and whether that was perhaps too small.

If I were evaluating this literature on its own, I would strongly suspect that all the positive results were artifacts of the garden of forking paths, but given the related findings in other allergies as well as the historical precedent, I give it some benefit of the doubt. It may work, but the effects are not huge, do not last a long time, and require a large enough dose that you are essentially giving yourself poison ivy to prevent poison ivy. Epstein notes that “the price for protection… is high in terms of untoward reactions such as pruritus ani, general itching, urticaria, and other rashes during the procedure using purified extracts, so that patients quit with the comment that the “treatment is worse than the disease.””

That outcome is obviously bad from a medical perspective, but from a mechanistic perspective, it still provides a challenge. Why can you induce hyposensitization at dosages that also come with severe allergic side effects? And if poison ivy immunity is T cell-mediated, how does the IgG vs IgE story come into play?

IgG response may still come to the rescue. Stampf et al 1990 found that IgG from patients who had undergone poison ivy AIT was capable of suppressing allergic responses to urushiol in mice. Perhaps the same tolerance mechanism as in AIT for IgE-mediated allergies is at work here even if the allergic mechanism is different. However, this is weird given that urushiol is a hapten. Why can we generate anti-urushiol IgG but not IgE? They claim that the most likely reason is that the IgG actually binds the T cell receptors that recognize the urushiol, which they demonstrate by showing that IgG fractions that have been adsorbed to mouse lymphocytes work worse for suppression than fractions that have not done so. But why should human antibodies bind specifically to mouse T cell receptors? I am confused about why they didn’t do an ELISA assay to be able to definitively rule out the IgG directly binding urushiol. There hasn’t been any direct follow up to this work since, but Baldwin et al 1999 did manage to develop a monoclonal antibody against urushiol in mice, indicating that should be possible.

There are groups trying to work around this problem, fortunately. Hapten Sciences currently has a vaccine under development using PDC-APB, a compound designed to immunologically mimic urushiol while preventing side effects. It is in phase I efficacy trials whose results have not yet been published.

Following up on the homeopathy thread, there are many easily available homeopathic extracts of poison ivy that claim to treat it. Most cannot possibly work. For instance, one labels itself as “30C” dosing, which means a 10^-60 dilution of the original extract. This would leave no molecules of the original extract remaining in the product. However, one called Oral Ivy has “3X” dosing, which is a mere 10^-3 dilution, putting it in the realm of possibility. Unfortunately, the question remains: “10^-3 of what?” One Amazon review reports upset stomach after taking it, lending plausibility to the existence of meaningful quantities of poison ivy extract (or maybe being just as meaningless as the reviews on the 30C products). From the studies above you would expect many more negative reviews if it were in fact potent enough to work. Another reviewer was bold enough to actually rub poison ivy on themselves after taking it and found their reaction was as bad as it had always been.

My Experience With AIT

I started allergy immunotherapy about a month ago. Three days in I was extremely tired and wheezing. We discontinued for a week and then restarted with a lower dose, successfully ramping up to the maintenance level over 10 days. Since then I believe my symptoms have gone down somewhat, though it’s hard to distinguish those gains from long-term antihistamine use (claritin specifically) and more rigorous hygiene (aggressively eliminating dust by vacuuming, washing clothes, and running an air purifier). Spring is just around the corner, so that will be another test of whether this actually works.

How To Get SLIT

If you are interested in getting AIT, the most straightforward way is to get a referral for an allergist and have them test your allergies and mix up the treatment for you. Sublingual immunotherapy is less widely used than shots, however, so it’s possible you may not be able to find an allergist that will give you the more convenient drops. There is a hip new direct-to-consumer telemedicine company called Curex that will offer you drops at prices comparable to what I’m paying through my allergist. I haven’t tried them, but I see no reason why the drops would not be equally effective. That said, since they are remote, their allergy testing services use an IgE blood test rather than skin prick and intradermal injection tests. My allergist’s opinion was that IgE tests are largely useless, and this seems to be backed up by the literature. IgE, unlike IgG, is mostly localized to tissues. A blood test can only measure what’s in your blood. This is a problem unless your blood levels and your tissue levels are very tightly correlated. One study on IgE testing for food allergy reports an 80% false positive rate (though they don’t specify what was used as the gold standard for comparison). Furthermore, it is clear that IgE levels aren’t consistently improved by tolerance acquired by AIT. It feels wrong to use a test for determining whether you need a treatment that wouldn’t also be able to determine whether the treatment had worked. Serum antibody titers are used because they’re straightforward to measure, not because they are a reliable measure of the thing of interest.

In the US, SLIT is in a weird legal category where it is not approved by the FDA but allergists are allowed to administer it. This means that it is not covered by insurance. As a result, it is reasonably priced (since the price is something that you are expected to actually pay rather than a meaningless input into the trench warfare of insurance payments). Mine comes out to about $1.50 per day, which is more than worth it under the assumption that it will work completely, and probably worth it with the unclear improvements I’ve had so far.



Thanks to Ago Lajko and Eric Frank for feedback on drafts.

  1. The mechanism by which this occurs is called V(D)J recombination. It is extremely cool and I encourage you to look it up. 

  2. Technically, “fragment crystallizable” is the real etymology since it refers to the part of the protein that can be crystallized to determine a common structure, but “constant” is a better, if fake, etymology in my book. 

  3. The other two common explanations are that IgG could act primarily by binding Fcγ receptors on mast cells, signaling to the mast cell that it doesn’t need to degranulate, or that there could instead be idiotypic IgG antibodies- that is, IgG that binds directly to the IgE, blocking it. 

  4. Mithridatism is named after King Mithridates VI of Pontus, who allegedly built up immunity to snake venom by consuming it. It is certainly possible for exposure to venom to generate antibodies to venom: most known antivenoms are produced this way in animals. However, the Mithridates story is probably a myth: taking snake venom orally means that the venom is deactivated by stomach acid, and so could not induce immunity. Rock star and snake enthusiast Steve Ludwin fancies himself a modern Mithridates and injects himself with a variety of snake venoms once a week, touting the immunity and general health benefits it gives him. But injection seems to not work either. In 2015, Lautsen et al put Ludwin’s serum to the test and it failed to protect mice against venom from a species of snake he had exposed himself to. His serum had some IgG antibodies against some components of the venom, but none against the most potent neurotoxins. The authors point out that the difference may have something to do with the small size of the neurotoxins relative to other components of the venom, though it is possible to generate antivenoms against that class of neurotoxin in horses. It is not clear to me why there should be a difference between humans and horses as far as targeting antigen size is concerned. Probably we are just more willing to kill horses than people in the process of finding out the right way to do it. 

  5. They report that, in the US, house dust mite extract is standardized to 10000 AU / mL, has Der p 1 in concentrations 20-50 μg / mL, and is recommended at 500 - 2000 AU / month, which in combination gives 12 - 120 μg / year of Der p 1. It is worth noting that they point out those recommendations are derived from studies on European extracts, which they show have significantly less Der p 1 that US extracts (eyeballing the graph, 5-25 μg / mL). 

  6. I cannot help but add this great excerpt in full: “During the French Revolution, Dufresnoy’s fondness for poison ivy nearly sent him to the guillotine. After sending some young plants from his garden to a physician friend, he wrote in 1794 to ask, “How are our dear Rhus? How I long to see them!” The letter was intercepted and Dufresnoy was arrested on suspicion of conspiring with the Russians (Russes in French), who were then threatening to join the wartime coalition against France. Fortunately, the political upheaval that ended the Reign of Terror also deposed the harsh judge who was about to hear his case, and Dufresnoy was able to explain to the authorities that “dear Rhus” referred to medicinal plants, not menacing foreign troops. After this lucky escape the doctor’s story has a poignant ending. Following his death in 1801 his skeptical pharmacist brother dug up and destroyed all the poison ivy plants in his garden.”