Gerry Lee, MD (Host): Hello, everyone, and welcome to another episode of Allergy Talk, a roundup of the latest in the field of allergy and immunology from the American College of Allergy, Asthma and Immunology. In this episode, we'll be reviewing more articles from Allergy Watch, a bimonthly publication, which provides research summaries for the latest journals of allergy and immunology. And you can earn CME Credit by just listening to this podcast or our new thing, a video podcast. To get CME credit, go to education.acaai.org. And again, if you're listening to this on audio, please visit our YouTube channel, youtube.com/@allergists. And of course, we also have the ACAAI Committee on DocMatter, where we can continue the conversations about the articles reviewed today.

Hello, everyone. My name is Gerry Lee. I'm an allergist immunologist at Emory University, and I'm an Assistant Editor of Allergy Watch, and I'm here joined by the Editor-In-Chief of Allergy Watch, Dr. Stan Fineman.

Stanley M. Fineman, MD: Hello again, and thanks for having me. For the next one more month, I'm the Editor-in-Chief of Allergy Watch. And I'm going to be turning it over-- and we'll talk about that a little bit later-- to Dr. Joshi. And I'm also an adjunct faculty at Emory, and I've been in allergy practice for over 45 years.

Host: And for the third chair, we're excited to be joined once again by Dr. Shayam Joshi, an Associate Professor and Section Chief at Oregon Health and Science University and Associate Editor, soon to be Editor-in-Chief of Allergy Watch. Shayam, welcome back to Allergy Talk.

Shayam Joshi, MD: Thank you so much for having me. I'm really excited to talk about some articles today.

Host: Okay. Well, I could not think of a better way to kick off our inaugural video podcast by these three so interesting articles. Stan, we have to start with you. We're on this technology theme, so let's talk about AI. What do you got for us?

Stanley M. Fineman, MD: So, this was a fascinating article that was published in the Annals of Allergy earlier this year. And it's entitled Can ChatGPT Provide Parent Education for Oral Immunotherapy? And it's from Canada, from Montreal. And what they did is they did a questionnaire. They took 14 common questions that parents have regarding oral immunotherapy. This is food now, for food allergy. And they entered them into a ChatGPT specific version, and they then categorized the responses that the ChatGPT made for either basic, advanced, or medical. And then, they had allergy immunotherapy healthcare practitioners, professionals from the North America, and even some from the United Kingdom, who were reviewing their responses.

And they measured them basically on a 10-point Likert scale, and they checked for response readability, for understandability, for reproducibility. And these were all assessed using something called a Flesch Reading Ease scale, and also a Flesch Flesch–Kincaid grade level score. And these are basically tools that can estimate at what grade level and how complex the answer is in terms of trying to understand it.

We all know that we're all doing oral immunotherapy now for food allergy. We know parents have a lot of questions. We know everybody's busy and unable to spend the time that we'd like a lot of times to try to help the patients answer questions. And of course, a lot of times parents are just going to Google and ChatGPT to ask them the questions to find out themselves. So, this is a real happening here. And I thought this was a fascinating attempt to try to quantify the responses.

And so, what they did-- okay, let me talk about the results here. So basically, the results to the 14 questions and the questions included things like, "What is oral immunotherapy? How effective is immunotherapy? How does it progress? Can you do multi-allergen immunotherapy?" And then, those were basic questions. Then, the advanced questions were, "What happens if my child misses a dose of immunotherapy? or "Can we do multi-allergen immunotherapy?" Then, medical questions were categorized as things like, "Will oral immunotherapy worsen my child's eczema or asthma?" or "Can my child eat a normal diet during oral immunotherapy?" You know, those are medical type questions. So, those are the three categories that the authors separated the questions into.

And then, they ran it through their ChatGPT and had the experts read the responses and evaluate on a Likert scale what it was like. And the bottom line was that the basic, advanced, and medical were really within a similar mean ranking of an 8.6, 8.4 and 7.8. So, they're all pretty similar. The length of the responses varied between 450 words to 650 words, which is obviously a lot. And the bottom line-- and I don't want to go through all the numbers here because it's too complex, we're just talking about this, but they basically had multiple comments suggesting that the ChatGPT response was repetitive, was long-winded and had information overload. So, it wasn't that it was that incorrect. In fact, it was correct about 80% of the time, roughly. But it was just repetitive, long-winded and had information overload.

They also checked the readability. We mentioned a specific score. They felt that there was low readability scores because it corresponded with a very high grade level, the responses. So, the responses grade level were between 13 to 17 years of education, which is pretty high when we consider a lot of times when we write instructions for our patients, we try to keep it on a sixth grade level, which is like the normal newspaper.

Then, the understandability was another one I thought was interesting. The understandability scores were between 73% and 84%. And then, in terms of similarity, those were between 80-90%. So, the bottom line was that the medical-specific questions, including the ones including diet and adverse results, they received the lowest scores for ChatGPT. And the other problem was the fact that it was not that readable. It was too complicated. So, what the authors concluded was that there are significant barriers to implementing AI as a formal tool in medical education right now. And that ChatGPT may have some potential in the future as an assistant to assist patient education. But right now, it's not ready for prime time. And the results, as I said, 80% of the healthcare. But even though 80% recommended the use for it in patient education, it still has a long way to go. So, I still think the human touch is important. And I think that's what we lose when we do too much with AI. And I'd be interested in your comments.

Host: I think my first comment is it sounds like they talked to the physicians who knew what they were reviewing and not the patients. Is that right, Stan?

Stanley M. Fineman, MD: That's correct. They did not talk to patients. They only were talking with physicians. And so, well, I shouldn't say-- they said healthcare providers, allergy specialists. So, I don't know exactly their education. That wasn't listed.

Host: I get where this paper's coming from, but I don't know, I think I'd be interested what the patients think, like when they get this information, maybe they like that or they have the opportunity to ask followup questions to clarify.

I think the third issue is that, even this paper, I think it was using 3.5. ChatGPT 3.5 is so much different from the latest now. At the time we're recording, we're in like, I think, ChatGPT 5 now. So, I personally notice a large difference over time, and I would assume that most caregivers are just not going to ask one question. They're going to have a dialogue.

So, ultimately, Stan, I'm not saying that I don't appreciate the human touch either. But absolutely asking someone immediately versus like sending on myChart and getting a message back. I don't know, patient's might prefer that. Shayam, are people using AI in your clinic? What have you seen?

Shayam Joshi, MD: Yeah. I think what makes this really challenging is OIT is done so differently at different clinics, at different hospital systems. And what the ChatGPT is learning from are what's available online, right.? So if certain clinics are publishing some stuff online or the very few papers that are not behind a paywall that are being accessed, those are all what ChatGPT is learning off of. And there's going to be difficulty because all the protocols are slightly different, all the kind of, "What are you going to do if you miss a dose? What do you do if you have a reaction or slightly different?" which I do think complicates how ChatGPT is synthesizing the information and pushing it back out.

I'm wondering, in the future, if we're able to retrain a specific AI model with a certain set of documents that are, let's say, most frequently asked questions or everything you can know about one OIT protocol, and you ask that system. That seems like a much more usable method, whether the college does that or whoever to create a packet of information that you can learn off of, that seems like a much more usable tool for AI. Because right now, I do think it's hard when you have so much alternate information out there that ChatGPT is learning from.

Stanley M. Fineman, MD: Yeah, I think you're right about that. And in fact, I look back, Gerry, this was a free accessible version from April 2024. So, it wasn't that long ago, but...

Host: That was a long time ago in AI terms.

Stanley M. Fineman, MD: Okay. Well, I guess it was. Anyway, and I do agree with Shayam, I think that, unless you're using some sort of a standard guideline or standard type of protocol, then that would be a challenge.

Host: Right. Well, we're definitely going to hear more about this. And, again, I think I agree with everything we're saying. We just want our patients to get accurate information that's evidence-based and aligns with the treatment plan. So, we continue the partnership between physicians and patients. And I would love AI to help us with that and not make that more challenging.

Well, anyways, let's just go to the next article. And Shayam, we continue to try to educate our patients as well as our colleagues about anaphylaxis. And it looks like there was this attempt to try to harmonize all these different guidelines. Tell us about it.

Shayam Joshi, MD: Yeah, I think this is a really important paper for everybody to be aware of. As stewards of patients with anaphylaxis and understanding anaphylaxis, it is our job to talk with other specialties, especially our ER, urgent care brethren that understanding really how to define anaphylaxis in a way that's easy for clinicians as well as patients is key.

So, the title of the paper was Anaphylaxis Definition Overview and Clinical Support Tool, the 2024 Consensus Report. And so, this was an international consensus guideline from about 46 members, international expert panels. These are from around the world, really the experts in anaphylaxis. But what was unique about this was that they didn't just include experts. They included 31 medical stakeholder organizations. And probably most importantly, they included 15 patient advocacy organizations. Because when we're putting this information out there, if patient's can't understand it, and it's not important to the patients, there's a lot of limitations around it.

And so, why they decided to do this was that the kind of classic or traditional way that we look at anaphylaxis was first by the NIAID, and the Food Allergy and Asthma Network definition, this consensus definition came out in 2006, it was great. It was very important for harmonizing how we think about anaphylaxis, how we use anaphylaxis in terms of research. But it had some holes in it. And so, in 2020, the WAO, the World Allergy Organization, they created an updated criteria. And they changed how we thought about it. It was always in three kind of categories of unknown ingestion, likely ingestion, and ingestion of a known allergen.

What the WAO did was that they actually narrowed it down to just two criteria. So, they had skin symptoms plus some other secondary organ system, or they had known or likely allergen exposure plus respiratory, cardiovascular, GI symptoms, but didn't really have to have the cutaneous manifestation. While in theory, that's great, but now we have two sets of anaphylaxis criteria. And that's a problem. So, patients are getting a little bit more confused. When you're looking at research papers now, some are using the original NIAID definition, and some are using the WAO definition. And so, it's even making it harder for us to compare research studies now.

So, this group got together, and they're like, "All right. Let's use both of these definitions. Let's include lay people in this definition. And let's define it. Let's create an overview." This is not meant to be a practice parameter, right? We just had our joint task force practice parameters come out in anaphylaxis. This is not meant to replace that. This is more of a, "Hey, let's create a clinical support tool that patients can use, that allergists can use, that ER physicians can use, primary care can use," a consensus that everybody can look at and be like, "Okay, this is how we're going to look at anaphylaxis. This is going to help support the patient the best way we can." And really, just to promote ease in clinical practice.

And so, what they did, I think the most important part about this entire article is they created this figure. It's figure two in the actual paper, if you're pulling it up at some point. But it breaks it down into what defines anaphylaxis. And so, they still stuck with the three-definition approach. So if you have no known allergen exposure and then you have skin plus some secondary organ system, or if you had likely or known allergen exposure and you had two or more symptom involvement, that's what we generally use. And most of the time when we're thinking about anaphylaxis, there's two-system involvement. And then, there's also a third option if you had known allergen exposure and isolated respiratory or isolated cardiovascular involvement. And I like this. I think the NIAID consensus guidelines is how most of us have kind of been trained and thought about anaphylaxis, and I like how they still stuck with this three-tier approach. I think it's visually put in a way that's really nice for ER physicians and urgent care to really be able to look at it and implement it.

But beyond just the definition, they also have a portion in this figure that shows what's the treatment. And I love how they just put epinephrine. They didn't put diphenhydramine, they didn't put steroids. They just put epinephrine here. And so, again, this is something we see all the time. I don't know how it is in Atlanta. But oftentimes my patients, especially kids with food allergies, they end up in the emergency room. They end up getting antihistamines, systemic steroids, maybe albuterol. They get watched for a few hours. Things progress. They eventually get epinephrine like an hour and a half after they're in the ER, instead of recognizing it is anaphylaxis and treat them for anaphylaxis, treat them with epinephrine. And so, they just have this kind of red box in here. The treatment is epinephrine. Here's the dosing for manual and autoinjector and how frequently you can give it.

And then, the last bottom part is just looking at the organ systems and what symptoms specifically constitute skin manifestations or respiratory manifestations or cardiovascular manifestations. And so, as allergists, I think this is something that we are very well aware of. But I do think this is a really important tool that we can use for patients, that we can use for our colleagues, especially colleagues that are going to be seeing anaphylaxis more often.

So, that's really the main purpose of this. I think it's just something allergists need to be aware of and need to see this figure a few times just so it could be disseminated appropriately across the medical field.

Host: One thing I wanted to add, Shayam, is that they actually help clarify some of the infant symptoms as well. I think sometimes it's underappreciated where if you go by strict guidelines, the subtle signs of especially someone who's not a trained pediatrician or has never seen infant anaphylaxis before may miss. So, I think that was also very good that they put that front and center in each specific organ system. And again, for those who want to be familiar with that, I think in the show notes, we'll link that article. And I think that graphical abstract or figure is just freely available.

Shayam Joshi, MD: Yeah, I guess we can call that out really quickly on the podcast in case you can't see it. But the infant symptoms that they call out here are mottling They call it repetitive lip licking as a mucocutaneous symptom, they call out a hoarse cry in the respiratory, and persistent and unexplained tachycardia in the cardiovascular side. So, these are all things that infants may display that, again, won't be caught with urticaria or kind of what do we classically think of anaphylaxis. But they do call it out in bold as some of the symptoms. So, good call out, Gerry.

Stanley M. Fineman, MD: Yeah, I think it's great that they did this. I mean, we've had trouble for years, as you mentioned, so that we could try to compare studies and also use of epinephrine. I mean, I recall when we we're still trying to get people away from using Benadryl as a first-line drug. And now, fortunately, this just clearly states use epi. We're going to talk on the next podcast about different ways that people treated anaphylaxis. And we'll talk about some of the fact, and a lot of times, especially in emergency rooms, they're not always using epinephrine as a first-line. So, we can talk about that a little bit more. But clearly, there are evidence-based guidelines that we can hang our hat on. So, that's good. Thanks for bringing it up.

Host: Yeah. Making it simple, easy to read, clear. I think it hits all the boxes. So again, definitely, share that and disseminate with your colleagues. We highly recommend it.

Okay. Well, let me just wrap up this inaugural video podcast with just a really interesting article that hit the press a few months ago. This is a study that is entitled Cysteinyl Leukotriene Stimulate Gut Absorption of Food Allergens to Promote Anaphylaxis in Mice. So, this was published in Science in August. And again, this is relying on mouse models to explore questions that we think about in the field of food allergy. And one is that some people even with IgE to food react and some don't. And why is that? And there's probably multiple reasons, but they actually have seen this in mouse models of food allergy as well.

So, there's different genetic backgrounds in mice that you can purchase and study. But they all respond differently to traditional allergen models. And the one that is most resistant to this classical oral food allergy model where you would give an adjuvant, in this case, cholera toxin and some sort of allergen, like peanut or egg, is a background called C57 black. And so, they could do the typical ejection anaphylaxis action models where you do an injection in the peritoneum, sensitize them, they make IgE, And then, they would have a reaction if you injected them again. But if you give them the food, they don't undergo anaphylaxis as opposed to other backgrounds. So, these investigators kind of want to understand this, like, what's the difference? And maybe does that give us insight on the variability of allergen responses amongst food-allergic individuals. So, they show their model. Again, they give multiple doses of peanut plus an adjuvant like cholera toxin. And we'll have little blood pressure cuffs on mice or anything. So, they look at temperature drops. So, anaphylaxis is usually measured by a drop in temperature when it occurs in mice. And so, they did show that the susceptible strains did develop anaphylaxis. And then, the C57 black mouse, even though injection, it causes anaphylaxis, it's resistant to the oral reaction.

And so, they did this for genetic screen to try to determine what genetically is different about that C57 black. What is their responsible gene or group of genes or genetic region? And they were able to localize, using this for genetic screen, a specific difference that explains a lot of this resistance to anaphylaxis if they take it orally. And it was dipeptidase 1. So, what the heck is dipeptidase 1?

Well, back, back, back, back, back, and we all know that we're either studying for ABAI or we took ABAI. And essentially, you may remember, in the leukotriene pathway, we have the cysteinyl leukotrienes, LTs C, D, and E. And the conversion of LTD4 to LTE4 is catalyzed by dipeptidase. And so, these mice seem to have increased dipeptidase 1 activity. Therefore, much lower levels of the earlier  cysteinyl leukotrienes, LTC and LTD. And interestingly, if you actually gave LTD4 to these resistant mice, they have similar oral anaphylaxis like their colleagues. And then, what they did was very interesting is then they administered anti-leukotriene therapy.

Now, Montelukast unfortunately only blocks one of the receptors,  cysteinyl leukotriene receptor 1. That didn't seem to do much. But when you go upstream and block 5-lipoxygenase with Zileuton, and you take out all the  cysteinyl leukotrienes, they actually had an impairment in this oral anaphylaxis, even in the susceptible mice. And the mechanism of that is really this oral absorption that the food allergen is not translocating through the gut to cause a reaction.

Now, obviously, all this was boiled down into a media article that says, "Is  Zileuton the next treatment for food allergy?" which is obviously the title of this podcast as well. And certainly, again, we're not saying that mice are human, that we give cholera toxin to sensitize people. That's not what we're doing here. But interestingly, that just talks about then is there a potential role of cysteinyl leukotrienes in the intestinal permeability, and is that one of the co-factors responsible for these variable responses that we're seeing with food allergy? As you can imagine, they want to test this hypothesis in humans. There are some description of maybe human clinical trials to see does giving Zileuton, when you eat a food allergen, modify your allergic response? I think that that's one of the questions they want to see if they can bring this translational project to the bedside. But again, I thought it was a very intriguing article. It makes us think about all the different factors of why two food allergy patients with IgE to food react differently. And I think I'm looking forward to what we learned from this.

Shayam Joshi, MD: I love the idea how we're trying to repurpose medications that we've used before, right? I have a pretty large AERD population, and we still use Zileuton not infrequently because it does help in that specific population, because we know they have increases in the leukotrienes as well. And so, I think this will be a very interesting option, especially if it can give selective ability to-- like, if patient's are traveling, they need some level of protection for a short period of time versus our current options do take some time to kind of build up and provide that protection. If this is an option that they can take starting a few days beforehand or whatnot, that they'll be able to have some level of protection, that'll be a great option for a lot of our patients.

Stanley M. Fineman, MD: I agree. And I mean, I remember using the Zileuton, especially in our aspirin-sensitive patients. And this could be a repurpose, the fact that somebody's thinking about the mechanism of the whole cascade of mediators.

And I guess I'm going to make another plug for Allergy Watch, because this is in the journal Science. I mean, allergists don't routinely look at that. And so, I think that the fact that we're bringing this forward for the allergists to know and the allergy community, kudos to Allergy Watch for doing that. And Gerry, you were the editor who brought that up. So, kudos to you and thank you.

Host: Well, social media brought it to me, so I bring it forward to y'all. So, anyways. Well, again, I appreciate you listening and potentially watching our first video and audio podcast. If you enjoy what you're listening to or watching, please rate us on YouTube, please rate us on your favorite podcast app.

And maybe you have thoughts. Please give us your emails, your comments. What do you think about AI and allergy and immunology? What do you think about the potential of Zileuton for food allergy or even the new anaphylaxis guidelines? The email to contact us is allergytalk-- one word-- @acaai.org. And just remember we have all these other episodes available for you to get CME on. The website is education.acaai.org/allergytalk. And of course, the published issues of Allergy Watch are archived on college.acaai.org/publications/allergywatch. Well, again, I'm excited. I enjoyed this very much. And I hope to see you for the next one. Everyone have a wonderful day.

Stanley M. Fineman, MD: Bye.

Shayam Joshi, MD: Take care.

Disclaimer: The ACAAI is presenting this podcast for educational purposes only. It is not medical advice or intended to replace the judgment of a licensed physician. The college is not responsible for any claims related to the procedures, professionals, products, or methods discussed in the podcast, and it does not approve or endorse any products, professional services, or methods that might be referenced.