Episode Transcript
[00:00:00]
Jill Brook: Hello fellow POTS patients and beautiful people who care about POTS patients. I'm Jill Brook, your horizontal host, and today we are discussing a research project and publication that uncovered some problems with a tool that many patients might be using to help select their medications, the DailyMed website, that many of us use to look up ingredients in our drugs.
Now, this has to do with accuracy of the information there and what we can do about it. So it's a really important topic. Now our guest today is one of the four authors on the paper, Mike Brook, data scientist. And if his name sounds familiar, yes, he is my husband and occasional host here on the POTScast.
He's the editor of the POTScast.. He does so much for this community. He makes the DoodleThru comics with me. He created WhatTheBleepCanIEat.com for our community. He created PatientsCount.org with me, our [00:01:00] non profit that provides data collection and statistical analysis for physician researchers who don't have academic backing.
But, when he's not doing all of those things for the POTS community, his actual true career is being head data scientist for the Alaska Native Tribal Health Consortium.
And your claim to fame is that you have mastered walking three dogs at the same time while getting AI to help analyze data for you.
Yes, my, my morning routine is to walk 3 dogs at the same time. Now, granted, they are small, but still there's 3 of them at the same time. And I have my headphones in and I talk to ChatGPT and ask it to do things for me. And then by the time I get home, it's done.
So, today we're talking about a paper that came out last March 2024 called Inconsistent Excipient Listings in DailyMed, Implications for Drug Safety. [00:02:00] And the authors were Laurence Kinsella, the autonomic neurologist, jill Brook, me, michelle Briest, who is a compounding pharmacist, and she has been on the podcast before, and Mike Brook, that's you.
Jill Brook: And I think I speak for everyone on the team, you did some cool and difficult things on this project that nobody else could do. So, let's talk about it.
Mike Brook: Well, let's, before we get to the nuts and bolts of what we did on this thing, let's, why don't you level set everybody on excipients? What are excipients otherwise known as inactive ingredients? Can you give us the patient's eye view on that?
Jill Brook: Yeah, sure. So excipients refers to all the ingredients in a medication product other than the active ingredient itself. So it can include things like dyes, preservatives, binders, fillers, [00:03:00] and, you know, I'm a big fan of Dr. Lawrence Afrin, who talks about excipients a lot, and he would say that even some of the unlisted, like contaminating ingredients, such as maybe, like, microparticles, things that could shed from the plastic vials, or the bag, or the tubing, that those would count as excipients also.
So, anything in the drug, besides the main ingredient.
Mike Brook: Okay. I know drugs are all different, but a typical drug that has probably one or maybe a tiny handful of active ingredients, but this may have a lot of inactive ingredients, can you give us an example?
Jill Brook: Sure, yeah, and I'm gonna everything I'm about to say comes from a paper by Reker et al. 2019, and these are some really awesome researchers at MIT and Harvard who wrote a paper called "Inactive" Ingredients in Oral Medications, and they put the word inactive in quotes, as if to say, are those inactive ingredients really [00:04:00] inactive?
And spoiler alert, they concluded no. But what they said is that on average the inactive ingredients provide about 70 percent of the mass of an oral medication and that an average tablet or capsule contains about 280 milligrams of excipients and 164 milligrams of active ingredient. And so they said that the excipients can add up to several grams per day
if you're on several different drugs at the same time. And some of them are associated with allergic reactions, even in minute quantities. And they made a point of saying that even in the teensy tiny doses that they would be in a capsule or a tablet, it can be enough to trigger symptoms. And just as an example, lactose
can be in there in a big enough dose to cause problems for people with dairy, or gluten can be [00:05:00] in there enough to cause people issues if they have celiac disease. FODMAPS could be enough to cause some malabsorption or issues in the GI tract. And they say this is a problem because it may contribute to poor compliance or an inability to take a drug when it's not the active ingredient that's the problem,
it's just all the little extra things thrown in there. And then one last thing, they, they make a point of saying 38 different excipients have been shown to act as allergens. And so, you know, could cause a real allergic reaction, even anaphylaxis.
Mike Brook: You gave examples of gluten and dairy and FODMAPs and things like that, but those are just the ones that somebody might know. I mean, it's, there's a lot of things where this never gets out of the realm of being a mystery, yet certainly a lot of patients and physicians have [00:06:00] noticed patients having reactions to some things or just not tolerating certain versions of drugs.
Jill Brook: Yeah, absolutely. And on that list of ingredients that can act as allergens for example, the most common one is lactose and it's in 45 percent of medications. The second one is cornstarch. That's in 37%. Polyethylene glycol is number three. That's in 36%. So, it's not like anyone seems to have tried to take the allergenic ingredients out of very many medications.
Now, now there is good news though. They also write about how most drugs do have a number of different formulations available. So if you don't tolerate one, you could try another one. And in their paper, they even have an impressive graph where they talk about you know, the most common drugs and how many how many different formulations [00:07:00] there are for those.
So, for example, if you had to take levothyroxine for your thyroid, then there is almost 150 different formulations that you could take.
And on average, each of those formulations has about 14 excipients each. So you could go looking for one that does not contain the excipient that's causing you problems.
Mike Brook: So certainly a major detective effort, but in order to get anywhere on that, you need good data. You need to know what's actually in these things and that's where DailyMed comes in, yeah?
Jill Brook: Absolutely. And that's why we needed you. So do you want to first just tell people what is DailyMed and how would a patient use it?
Mike Brook: Yeah. So DailyMed's a website, it's a government website, and it's from, uh, the National Library of Medicine. So you can just Google for DailyMed. And what it is is basically [00:08:00] a search engine for drugs. And, um, it has, something like 150, 000 records in it and you can search by name and there's, there's a, there's something called an NDC code, which is a sort of a special number that identifies drugs.
And you can search by a number of different things. And what you end up getting is essentially a rundown on this drug. It's really what it really is, is everything, it's the official documentation of what's in that drug, what it's used for, what it's not used for, what the warnings are about it,
what the ingredients are, obviously, what the molecule looks like, you know, all of this kind of stuff is on there.
And that's a lot of information for every drug. And so what DailyMed does is it makes it searchable and arranges it in a friendly way.
Jill Brook: Okay, cool. So if I, if I had an experience where I took one form of Allegra and it seemed okay, but then another day I took a [00:09:00] different formulation of Allegra and I thought I had a bad reaction to it. What would I specifically do with this website to help figure out what might have been the issue?
Mike Brook: So what you'd probably do is you'd do two searches. You'd try to find the two versions of Allegra that you had taken. And that's not always easy. I think if you search for Allegra out in DailyMed, you're going to get dozens or maybe even hundreds of versions because it's going to have things that are generic, things that are white labeled, things that are repackaged, it's going to have the store brand, and that kind of stuff.
So it's going to have all those different versions of Allegra. And so finding the ones that you took can be a little bit tricky, but that's just kind of a matter of, you know, okay, here's the, here's the packager on this one. And sometimes you can look at, you can look for something that looks like a, some sort of a code, that's the NDC code and you can search on that.
So, if you were to do that, and you find the two, one that's working for you and one that's not, you could pull up the [00:10:00] inactive ingredient lists for each of those. Now, the active ingredients are going to be the same by definition. That's what that's what it means to be a different version of the same drugs, the same active ingredients. Now, where they're going to differ is probably in the inactive ingredients. And that can include things like
fillers and colorings and flavorings and you know, that, that kind of thing too. So if you were to pull those two up and you see, oh, this one has titanium dioxide and that one doesn't, well, that might be your culprit right there.
Jill Brook: Right, right. So it really helps with the detective work. Now, the reason that we pulled you into this project is that there was a number of physicians who work with patients who have dysautonomia and POTS, and Mast Cell Activation Syndrome, EDS, all the things that make patients sensitive to medications.
And so they're always telling their patients to use DailyMed to make sure that they're not taking in excipients that are or triggers for them. And they had a few patients say when [00:11:00] I look in DailyMed, I'm getting some conflicting information.
Mike Brook: So I think what they probably noticed is that the drug inserts, so that, that piece of paper that comes with the drug, that, that has all that tiny, tiny fine print that if you were to look close enough on that, you'd probably find a list of inactive ingredients on that. And, somebody probably looked at that and then went out to DailyMed and said, no, this is the exact same drug.
How come these lists are different?
Jill Brook: And the same NDC code, so not only the same drug, but the same drug product.
Mike Brook: Yes. Yes. Those really need to be the same. And so what that, that, what that led to was a little bit of confusion, skepticism and concern, I would say, about what's out there. You know, what is, is this DailyMed thing broken? You know, what's going on here?
Why are we not seeing what we expect to see in this thing? So that's what really kicked off this project. And the reason I got involved is because there's a data source underlying this whole thing [00:12:00] that, that that DailyMed is really displaying, but DailyMed is not itself the source of truth.
They're not the agency that essentially owns this information. That would be the FDA. DailyMed is from a different agency that's just making it more friendly. So, one of our hypotheses is, well, maybe the DailyMed website is just hopelessly broken.
So, so one of the things we did is we went directly to the FDA and got the, the data files, the source files. And then we went to DailyMed and we confirmed that, no, these are exactly the same. They are exactly the same. So that's not the problem. The problem is not that DailyMed is somehow losing data or messing data up or something like that.
So what we eventually found is that the issue was internal inconsistency within each of these data files. Now, what, what drug packagers are obligated to do is to document what's in these drugs in this special file format. This is a file format called SPL, [00:13:00] Structured Product Labeling. It's, you can, you can search for this and you can read the spec on it if you really want to. Suffice it to say, it's this giant file. It's a huge thing. It's this big hierarchical file. And every different drug has one, basically. And so the, the drug packagers are obligated to submit one of these.
This is the official documentation of what's in this thing, who should use it, who shouldn't use it, all that stuff.
Jill Brook: And I think we figured out it was the drug companies who submit the information.
Mike Brook: Yeah, they call them packagers and so I think what that means that I think that includes those that manufacture drugs, so that could be 1st party, it could be generics for drugs that are available as generics. It could also be, I think, white label. So they call them packagers, but I think it's all of the above, but you're obligated to still get some level of approval from the FDA for these things. And the way that you document all that is in this file.[00:14:00]
Jill Brook: Okay, so we figured out that DailyMed itself was just only reflecting data that it was getting from the drug packagers through the FDA file.
So what do we mean by inconsistencies in the label?
And then how did we go about testing it?
Mike Brook: So here's, here's what we found. This was the problem. It's that inactive ingredients or excipients are typically reported in at least two places in this file. There's what's called the description section, which is a narrative. You're probably used to reading these if you look at a drug insert. It's the part that has, often has a little molecule diagram and it says, such and such a drug is useful for treating this, blah, blah, blah.
And then oftentimes the last sentence in that paragraph will be this drug also contains the following inactive ingredients and it lists them, but it's very much intended to be read by a person. There's another section in that file that is more of a, a table of ingredients. It's kind of like a, it [00:15:00] almost looks like an ingredient manifest for the, for the for the drug.
And that also contains inactive ingredients. Now, so we've got inactive ingredients listed in two places. I think everybody would agree those lists should be the same. This is all in one file referring to a single formulation of a single drug from a single packager. They
should really be the same. And the problem is, they're not in a lot of cases.
And I'll tell you how many in a second.
Jill Brook: Let everyone just take a moment to guess. If you were to look at a hundred random drugs in DailyMed, how often would you expect to see internal inconsistencies between these two lists? Just kind of, just kind of guess.
Mike Brook: Think about what number might be problematic if, if you, if you had this. And I think, you know, people that have MCAS and MCAS doctors and things are, are gonna definitely resonate with this idea that this would be a problem [00:16:00] and, you know, almost any number is too high.
Jill Brook: And actually, when we started this project, I asked the physicians that brought this to our attention, what number would be problematic? Because that determines our N size. We need to look at a lot more drugs if you say 1 percent is problematic then if you say, you know, we're fine with it up until you get to 20%, you know, 20 percent is good enough.
And what I heard from them was around about, you know, they thought 1 2 percent
would be a problem.
Mike Brook: So that was kind of the consensus is like, oh boy, one or 2 percent is, is a, that's a, that's a bombshell. You know, if it's that high, well, the number that we found was about 40%, 40%. And so, and when I say about, there's a couple of reasons for that. One is we took a sample. So we took a random sample and what we did is we selected we, we, we got a list of the 50 most common drugs because we wanted to [00:17:00] do drugs that people actually use.
So the 50 most common drugs, and then we randomly selected 2 versions of each one. And we just manually, we went out to DailyMed and we inspected these two sections, and we just said, okay, well, this one says this, this is this, this one says this, are they the same? And in most of were the same, 60 percent were the same. 40 percent were not.
And the other reason that I say that that that number is kind of fuzzy is because it's kind of a matter of opinion whether something is a problem or not. So you can think to yourself, well, if if one side of one part of the file says water and another side of the file says distilled water, I'm going to give that a pass.
They're technically not exactly the same, but close enough. Okay, whatever. If one of them says starch and the other side says corn starch, [00:18:00] that's a problem. I think that's a problem. I think, you know, a lot of people in the MCAS world would say that that's a problem.
People with corn allergies for sure.
Mike Brook: That was incredibly common. What is definitely a big deal
is when one part of the label contains a list that has inactive ingredients and some of those are just plain missing from the other part and we saw that actually quite a bit. You're going to give some examples of that. But one thing last thing I would say is one hypothesis we had is that there's one section of the label that's quote correct and one that's just derelict. Right, that one has all the stuff and another and the other of them is missing some stuff.
Mike Brook: That wasn't true. It was it varied. Sometimes it was in the description section where you saw a bigger list and sometimes it was in the other section where you saw a bigger list. So imagine being a patient and you know depending on which [00:19:00] section you happen to look in, you're going to get a very different picture for what was in that drug.
I wonder if you could we, we gave a couple of concrete examples in the paper. I wonder if you could walk through a couple of them.
Jill Brook: Absolutely. And so, one example was acebutolol hydrochloride. And just, you know, as a reminder, the label has two different places where it lists the inactive ingredients. In one place, the only inactive ingredient named was F, D, & C, blue number one. But then on the other place, it listed D&C red 28, D&C yellow 10, FD&C blue 1, FD&C red 40, gelatin, maize starch, povidone, stearic acid, and titanium dioxide.
And by the way, yes, a couple of those are on that list of 38 known ingredients that can be allergens. So, a person who only looked [00:20:00] at one section you know, we don't know which one of these is correct, but you would get a very different answer depending on what you were looking at. And so, believe it or not, we found 41 different labeling inconsistencies in 39 out of 100 products.
So that's where that 40, 40 ish percent came from that you had. So different kinds of inconsistencies. So like we said, some of them were more minor. So for example, the starch omitting the source, right? If you say, well, okay, you know, unless you have an issue with corn or rice, maybe it's not that big of a deal.
But that came up 10 times where it would be listed in one place or not the other. There was another four times where it was a very similar thing where a detail about the source of the ingredient was only included in one section or another, for example [00:21:00] peanut oil versus vegetable oil or corn oil versus vegetable oil.
And that would certainly matter to somebody with peanut allergy. There was one episode where it was just a completely conflicting ingredient, and that might have been a just a typo by somebody, but the example is the drug fluoxetine, where on one list, it had dimethicone, and on another list it had simethicone, and those are not the same thing.
Now there were a number of times when an inconsistent name was used for ingredients, like a synonym. And an example was in budesonide. On one list they used the term polysorbate 60, and on another list they used the term emulsifying wax. And so if somebody had an allergy to polysorbates, then you wouldn't necessarily know it when [00:22:00] you're reading it as emulsifying wax.
There was another 15 times when the when ingredients were missing from from only the tabular list, there was another nine times when the ingredients were only missing from the description section. And so all of that added up to 41 times that these two lists were not consistent.
Mike Brook: This was far more prevalent than any of us ever expected. I mean that the, these are very real concerns at a, at a rate that I think far outstrips what you can just ignore.
Jill Brook: And we should add that in our results, when there were inconsistent names used on the two listings, we did a lot of work to figure out if they were the same ingredient, and we only held it against them if that inactive ingredient with a different name was on that list of [00:23:00] allergens. And so that's something that came up all the time and it's just so frustrating as a patient trying to figure out what's in your, in your drug.
If you have to check both lists, you also pretty often have to do a ton of extra Googling to figure out are these two things the same name or different names for the same ingredient. But we didn't even hold that against it here. Okay. So talk about what you did next because we wrote this up, we published it in April of 2024.
It was in the Archives of Pharmacology journal. We, I don't know, I think we hoped that maybe, like, somebody from the FDA would call us and say, hey, we're gonna try to address this and fix it. Nobody ever did. And we had always, I think, maybe regretted that we only did a hundred drugs because we were finding so many errors and it was quite laborious.
And so at the time we had, we'd kind of daydreamed, Oh, someday we can use AI to do this and [00:24:00] continue the research.
Mike Brook: Yeah. So it turns out it's actually not that difficult to get AI to do this. So, just a couple of weeks ago, I decided to give it a shot and try to do not a hundred, but some thousands. I didn't want to do all 150, 000 because that would have cost a little bit of money. But 1, 000 was 10 bucks to, to do it.
So
Mike Brook: that's what we did. And what I did is I selected
MCAS relevant drugs. So I was able to sort of, query the DailyMed system. For example, I got all antihistamines. There's almost, there's hundreds and hundreds of antihistamines. I got Cromolyn and I got some other things that a lot of MCAS patients would maybe take.
Mike Brook: And so that ended up being about 1, 500 drugs. And I put them through an evaluation that I had an AI model do. So this wasn't anything where I had to do anything fancy like invent a new model or anything [00:25:00] like that. This is just essentially doing a little bit of text processing and then clever prompting of one of the GPT OpenAI models.
So this happened to be a model that had just come out called o3 mini. And I wanted to, you know, part of the reason I did this is I wanted to test that model. And it had just come out and I thought, okay, well, great, I'll do this. And so I basically through a series of little bit of experimentation arrived at a prompt that essentially got me what I wanted and by got me what I wanted I mean, I thought it had a about the right level of sensitivity.
I coached it to ignore things like water versus distilled water. I also told it to prioritize or put sort of like an importance rating on the issues it was finding, and I said that synonyms confusing synonyms, while they're important, they're lower priority. I, so, so it would tag those as sort of [00:26:00] low priority.
The starch corn starch situation that we mentioned, I had it put that low priority too. And I think a lot of people would claim that should be high priority. I put it low priority because it was so common. It was just absolutely overwhelming the data.
Jill Brook: Plus, you were finding much bigger issues.
Mike Brook: Yeah. So the, so the, the biggest issues were, like I said, missingness. So something appears over here and it just doesn't appear over here at all.
Jill Brook: And that happened with flavorings so many times that we started to wonder if they have some sort of pass, but the physicians that we've consulted say that there should not be a pass for flavorings, because, for example, something that's flavored mint, that's that's a known sensitivity that a lot of people have.
Mike Brook: But I will say that in my analysis, I had to downgrade flavorings because it was so common. So basically, it was a matter of me kind of going through and getting it, sort of coaching the AI to where it got, it flagged things in a way that I thought made sense.
Now, that was just my opinion. [00:27:00] Other people could disagree on what was important versus not. But nonetheless, it was, it was quite successful at doing it. So I had it just churned through about 1500 of these things, and then just sort of like tabulate the data as it went. And, once again, to my surprise, about 40%.
So our our sample of a hundred was
not, you know, just some weird lucky thing that we, we happened to hit. It was pretty much right. And that was with me really being pretty sympathetic to the drugs, I think, and like downgrading things like the flavoring and that kind of stuff. So that was, that was me being giving a pass to a lot of things that maybe didn't quite deserve it.
But nonetheless, 40 percent was approximately what we found among MCAS drugs, antihistamines, Cromolyn and so on.
Jill Brook: Drugs for sensitive people.
Mike Brook: Yeah. And, you know, I should be clear, the I was very careful to audit the results of the [00:28:00] AI. So, you know, I was pretty skeptical that it was finding what I thought it was finding.
So I think I, I reviewed hundreds and hundreds of these, of the things that were sort of the hits. Reviewed hundreds of them and went through and said, Oh, sure enough. And I didn't, it didn't make a single mistake in what I was doing, which was really remarkable to me. So 1500 tries, basically churning through and, and I would emphasize that this, like comparing one list of jargon terms, like drug ingredients and things that you don't see every day to another list where you've got synonyms, you've got, they're in a different order,
it's presented a little bit differently and then you have the missingness problem and all that. For a human, it's cognitively difficult to do that. Like, it's actually tiring to be looking at this big list of words and this big list of words and saying, where do they differ? It's actually pretty hard to do and this thing was great at it.
It's really a perfect use for AI, I think actually, and it did a very good job. So [00:29:00] that's what we found is out of the 1500, approximately 40 percent and these are legitimate problems.
Jill Brook: And in some of your noodling around, you played with some other drugs. So this wouldn't have happened in the Mast Cell Activation Syndrome drugs, but for example, in some of the more like skin related products. It found a type of error that we did not even think to look for, which was fragrances in products that claimed to be unscented.
Mike Brook: Right, so some of the things that are in this database are things like sunscreens. And like things, other sort of cosmetic things, topical creams, that kind of stuff. And that isn't what we looked at originally and that they're, they're not MCAS things, but I did sort of randomly sample and look at random things.
And, and that was one thing that did pop out is for things that would tend to have fragrances even for things that said right in their name, unscented, there's a fragrance. And I [00:30:00] know that's probably, you know, certainly an issue for, it would be an issue for you. And I think it would be an issue for a lot of people.
Jill Brook: Well, and there's many, there's many skin creams that I have not been able to use because they irritate my skin and you know, I've never exactly figured out why, but it could be things like that.
Mike Brook: Yeah, and what one of the points that, that one of the physicians made that I think is a good one is that the reason we did this is not to scare people off of taking drugs. The scenario that we have in mind is somebody goes on an antihistamine or something that's supposed to help and it's the 1st one they've tried, and they react to it.
Now, you really think that person's going to go shopping for one that they don't react to at that point? No, they, they're going to cross that off the list. And that might have been a therapy that would have helped. And so this idea that we're playing fast and loose with inactive ingredients is I think maybe depriving people is something that might help them, a therapy that could [00:31:00] potentially help.
Jill Brook: There's a lot of physicians who would say, who would say, if it's an important drug for you, try to find a version of it that you can tolerate instead of just necessarily giving up. And you know, that probably deserves a conversation with your doctor because maybe you had a real known side effect to the active ingredient,
but if it seems like it was something maybe more a reaction to an excipient, then maybe with just a little searching you can find a version that works for you. And so that kind of comes to our conclusions for patients, which is not to be discouraged and disgusted by all this. And like, geez, we can't even get accurate labels.
'Cause I guess we should say that all we could do was compare two versions of ingredients that appear on the same label. We don't know which one is correct. We don't know if maybe both of them are incorrect, who knows. But for patients who are really sensitive, what we wrote in the paper was that you should check [00:32:00] both lists of inactive ingredients,
and if they match, then maybe you can kind of feel comfortable that you know what you're getting. If they do not match and you find discrepancies, then oftentimes the drug company will have a 1 800 number like a helpline there on the label. And maybe you want to call that, and get a little bit more information from them, find out what, what, you know, really is in there and in, and just keep track of which formulations are working for you.
Mike Brook: And then do remember that as Michelle Briest, the compounding pharmacist said in her episode for us, it is possible to get many medications compounded where you control the excipients, you and your doctor decide what gets put in there and so you don't have to have any ingredient in there that you don't tolerate.
I guess this is evocative of a lot [00:33:00] of things that I've seen you go through as a patient. Which is that you kind of can't theorize your way into success. You have to try, unfortunately, you know, it's what works for one person won't necessarily work for another person.
Mike Brook: And even in the world of drugs, one version of such and such a drug might not work the same as the one you took last week, you know, and so it really is, it's frustrating, it's a little bit of a trial and error game, and this is yet another example.
Jill Brook: I know, what I thought you were going to say is, it's evocative of being a patient and even the things that are not supposed to be hard are oftentimes hard, like you're supposed to be able to just look up this nice database and get your information and even that turns out not to be straightforward.
Mike Brook: The world of chronic illness is full of metaphors, right? That's the one thing you can say about it. There's no shortage of metaphors.
Jill Brook: Well, we were very grateful that you helped make this project possible as you help make so many [00:34:00] projects possible. And we should start talking about more of them on the podcast.
Mike Brook: Well, if you keep talking, I'm going to, it's going to take me so long to edit this thing that I won't finish it in time. So let's wrap it up.
Jill Brook: Okay, listeners, that's all for today, but we'll be back again next week. Until then, thank you for listening. Remember you're not alone and please join us again soon.
