Harsh Thakkar (00:01.102)

All right, we're live. And my guest today is Sharmeen Roy. She is the Street Strategy and Science Officer at DoseMe. And we're going to be talking a lot about DoseMe, which is a platform that uses Bayesian dosing or statistics to figure out dosing for patients. We're also going to go into precision dosing and a bunch of other topics. So please welcome to the show, Sharmeen. Thank you for joining.

Sharmeen Roy (00:29.552)

Thank you, Harsh. It's really nice to be here. My pleasure.

Harsh Thakkar (00:32.366)

Yeah, so I want to start off by asking you, can you expand a little bit about what, you know, I know just from my research that DoseMe is a software platform that uses the Bayesian dosing statistics, but what are like the type of users that use that platform or when in the life cycle do they use that platform?

Sharmeen Roy (00:53.776)

Sure. Yeah, so I would say think of DoseMe as our company and DoseMeRx as our platform. And it's a precision dosing platform. So essentially, the method we utilize for precision dosing is Bayesian dosing. So there's Bayesian methods, which are basically any drug that is available to any medication that patients use, it's been studied on various populations. And what

Harsh Thakkar (01:08.334)

Hmm.

Sharmeen Roy (01:20.496)

What happens is that you learn from those populations how the medication works, how it's cleared, how it gets processed through your body essentially. I'm kind of simplifying it, but that's the information that we take. So there's models that are created based on certain patient populations to show, okay, this medication, this dose will take 10 hours, 12 hours, what have you to...

clear someone's body. So it creates these models. So we take those Bayesian population -based models and integrate that with patient information. So if you are a patient in a hospital, we take your lab information, your demographics, your age, weight, height, all that information about the patient, and sort of combine it with that information that we have from the models.

Harsh Thakkar (01:52.462)

Hmm.

Sharmeen Roy (02:16.4)

And that creates a precise individualized dosing for your patients that you're taking care of. Who is the user? I would say it's clinicians. So it's a platform that's intended to be at the bedside. So it simplifies these complicated methods and complicated dosing methods at the bedside so you can have it within minutes or seconds.

Harsh Thakkar (02:27.278)

Hmm.

Sharmeen Roy (02:42.576)

as far as the patient goes. So the users are most users in the US, I would say, are clinical pharmacists. As we, you know, as clinical pharmacists in an acute care setting, so in a hospital setting, are really integrated with the possible teams, the multidisciplinary teams, and they're focused on the dosing for those patients.

Harsh Thakkar (02:49.998)

Okay.

Sharmeen Roy (03:05.008)

And so they're the users. However, I would say, and I'm looking largely since we are a company that a software that's available internationally, our users vary in the international markets. They're clinical pharmacologists, they're physicians, providers. So basic, you know, it's really intended for all types of clinicians that are taking care of patients.

Harsh Thakkar (03:25.422)

Okay. And you mentioned that the Doze .me RX platform also takes into account some of the other patient data. So does it like integrate with softwares like Epic or something like that?

Sharmeen Roy (03:38.224)

yes, yes. So we have our platform is because we want to make sure that it's available to all sizes of hospitals, whether they're large academic centers as well as smaller critical access hospitals. So we want to make different methods of making it available. So we have it available on sort of like the web. You just go on the cloud and can use it for your patients.

Harsh Thakkar (03:39.438)

Okay.

Harsh Thakkar (03:59.79)

Mm -hmm.

Sharmeen Roy (04:04.304)

but it also integrates with all electronic health records. So EPIC, Cerner, as well as some surveillance software is just because, depending on the institution, they may utilize our platform via different workflows. So we want to make it available. So it is integrated. And the nice thing about integration is that it brings all that patient information, sort of populates it into the software automatically to just, again, streamlining the workflow for the clinicians.

Harsh Thakkar (04:17.39)

Mm.

Harsh Thakkar (04:29.006)

Right.

So why is precision dosing so critical compared to other past ways of dosing? Can you shed some light on that?

Sharmeen Roy (04:44.688)

Yeah, so that's a, you know, I can probably talk for hours about that. But really, you know, we've been doing things in healthcare and medicine a certain way that one size fits all kind of mentality. And our patients that we're treating are no longer the same ones that were used to maybe approve, get the approval for the medications, right? Those populations have changed. You know, many medications that were approved, you know,

Harsh Thakkar (04:48.334)

haha

Harsh Thakkar (04:57.358)

Mm.

Sharmeen Roy (05:11.12)

in the fifties are still being utilized now. And the patient population was very different, right? Our population was very different. And as time has gone on, you know, we've learned that as an industry, that one size fits all is not very, it's not the best approach for our patients because there's a lot of variability. So even, you know, as we're talking, you may metabolize a medication very differently than I would.

Harsh Thakkar (05:12.654)

Mm -hmm.

Sharmeen Roy (05:37.584)

And so that is what's not taken into account in the methods that were used previously. And there's more of a, I would say a movement towards making more individualized dosing to take into account those specific parameters that, you know, I would have, or you would have, or any other patient would have to, you know, exposure of the drug to make it safer for patients, as well as make sure that we're, you know, being effective and managing the toxicities for those patients. So precision dosing, you know, it's,

It's not a new concept, I would say. It's not something new that we've been doing, right? Because as a pharmacist, I can say I'm trained in pediatrics. So I always tell everyone that pediatric pharmacists have been doing precision dosing before it was the thing that everybody was focused on, right? Because if you think about the patient population, it's from birth to adolescence. And those...

Harsh Thakkar (06:17.038)

Mm -hmm.

Sharmeen Roy (06:32.176)

The 18 year old is very different than a newborn baby and how they metabolize medication. So it's really better for our patients. And we really need to make sure that we're dosing our patients that we see in real life versus the ones that were studied or in clinical trials that were maybe done many years ago.

Harsh Thakkar (06:35.342)

Right.

Harsh Thakkar (06:54.478)

So like when you were explaining, so not going the one size fits all approach and getting more individualized, finding different characteristics or variability of that individual. So what are some of the challenges that come in? Because I'm assuming like there are certain things like age, geography, sex, all the different things that you...

probably precision dosing considers to determine that variability. So what are some of the challenges when you're going this route of making it more individualized versus one size fits all?

Sharmeen Roy (07:36.368)

So the biggest challenge is that clinical trials are done on populations and groups and cohorts of patients. We're not taking one patient per dose and trying to figure out how each individual metabolizes it, which is where our platform takes that model information. So you may hear the term model -informed precision dosing. So that's taking an existing model that has been established.

for medication that's been approved. So it doesn't necessarily have to happen in the only the drug development process, but even once the medications are available. So these models get developed for certain cohorts of patients, whether it's a pediatric cohort or patients who are critically ill or have some sort of renal dysfunction, for example. So these models get developed and what...

What our platform does is takes that peer -reviewed validated model and it integrates, again, the patient information to help inform the dosing and individualize the dosing. So that over time it learns and it allows you to then individualize the dose for those specific parameters for the patient. So it's...

it's a step towards the right direction because in reality there's not going to be every single cohort that's studied and a randomized clinical trial method. And that's where we have to take really advantage of the technology that's available to take that real world data and real world patient data to see how we can improve on the dosing that may be available or applicable to the patients.

Harsh Thakkar (09:22.158)

Hmm. So I'm just curious. I don't know if you know this about this topic, but like, how would it be different? Let's say, again, I'm not sure I don't work in clinical trials. So I don't know how, you know, the stuff works in that area. But if, if it's, I'm thinking more along the lines of like a cell therapy, a person, a patient who's in a clinical trial for a cell therapy where, you know, the, the number of people are

very small, right? So maybe there's more chance to have that individual data from that patient versus a big population. So are there specific strategies for precision dosing based on what the intent of that trial is or what therapeutic area that trial is targeting?

Sharmeen Roy (10:13.072)

Yeah, so there's, I guess it's a little bit different depending on sort of what type of patients or what type of therapeutic area. So for example, you have infectious diseases or oncology. Oncology is probably where you're referring to cell therapies and that are very individualized. I would say that the support for that exists in clinical trials because it is for that specific patient, right? You're creating that.

Harsh Thakkar (10:22.382)

Mm -hmm.

Sharmeen Roy (10:42.512)

treatment for that individual taking their cells and potentially coming up with treatment. I think there's another aspect of it is so in the drug development process, you go through multiple phases. So you start with phase one, then two, and three. In phase one, I would say it's the healthy patient population. You're studying the drug how.

how it behaves in a healthy population. Then you move on to a more well -defined cohort of patient where it becomes a little bit smaller, but that information still does not in real life translate for the patients we're treating. There's a whole set of underrepresented population that we need to do better for. And those are as our...

Harsh Thakkar (11:21.806)

Mm -hmm.

Sharmeen Roy (11:33.584)

population is aging, you know, we don't have a lot of information on the elderly and how they metabolize medication or how much exposure they're getting from a given medication that may be appropriate for someone who's younger, but may not be appropriate for the elderly. So, you know, I think there's different approaches for depending on the medications that we're studying or the area. There's a lot of data, I would say, in the drug development process that

is used to get the approval of the drug and then it kind of sits on a shelf somewhere. We don't really get to utilize it and apply it to those underrepresented population and how we can learn from that to better dose that population. So I hope I answered your question.

Harsh Thakkar (12:10.094)

Right, right.

Harsh Thakkar (12:21.006)

Hmm.

Yep, yep. And I want to switch gears and go into, you know, Bayesian dosing. But before I go there, I tried to find out what's the difference between precision dosing and Bayesian dosing. I read some blog articles and I'm even more confused than I wanted to be. So from what I understand in layman's term, precision dosing like takes into account some of, you know, the patient characteristics or whatever data is available.

Sharmeen Roy (12:32.944)

Mm -hmm.

Sharmeen Roy (12:42.672)

Hahaha.

Harsh Thakkar (12:53.55)

but Bayesian is more like a statistical method that takes in real time data. I'll let you explain for our listeners, because I'm sure they all have this question.

Sharmeen Roy (12:59.888)

Yeah. Yeah.

Yeah, so precision dosing, it's a term that's used often, right? It's used in so many different kind of, by so many different types of companies and software. There's personalized, there's different, in the umbrella of precision medicine, there's personalized medicine, there's individualized medicine. So it gets a little bit confusing, I think. Bayesian dosing, Bayesian is a method, right? So it's looking at, so think about it as,

Harsh Thakkar (13:09.55)

Mm -hmm.

Harsh Thakkar (13:30.286)

Hmm.

Sharmeen Roy (13:34.256)

So before you start a patient on a medication, you have this prior information. So you have a population pharmacokinetic model that is telling you some information about how the dosing could be done for a patient to start them off. Now, as soon as you get some information from the patient, so whether the patient demographic, lab values,

Harsh Thakkar (13:51.182)

Mm -hmm.

Sharmeen Roy (14:00.432)

some individual components of their physiology and you combine that with the Bayesian model, then you get the posterior. So it's pretty much after you dose the patient with the first dose and you combine it with their patient data, that's when you get more precise dose. So it's a method used to get to precision dosing. And like I said, it...

Harsh Thakkar (14:27.374)

Hmm.

Sharmeen Roy (14:28.688)

There is also a term called model informed precision dosing, and that's kind of used interchangeably with the Bayesian statistical methods, taking into consideration all the models that are utilized to then inform you in better dosing the patient. So I guess an analogy I could make is we use our Google Maps or GPS system, whatever, on your smartphone.

And if you know that you're going to a given address and say it's your new work, you started a new job and you're trying to figure out what's the best way to get to drive to it, if you're driving to work that is. And so you start out with an estimate, right? And try to start out with an estimate of, okay, it's gonna take me 30 minutes to get to my work. Now, as time goes on, you start getting more information in that path, right? You start getting your...

Harsh Thakkar (15:05.038)

Mm -hmm.

Harsh Thakkar (15:10.67)

Thankfully not. Yep.

Harsh Thakkar (15:18.446)

Hmm.

Sharmeen Roy (15:25.52)

Maybe it's the time of the day, maybe it's the traffic information, or maybe other components that are informing that route. So as time goes on, you may get better at getting to your work faster because you figured out a different route, or you figured out more of your individualized patterns to get to it. So I guess that's an analogy I could make that you start out with an estimate, and then as you actually take that route, you learn from that route.

Harsh Thakkar (15:45.006)

Hmm.

Sharmeen Roy (15:55.183)

you know, what, how to get to your destination faster, if that makes sense.

Harsh Thakkar (15:59.534)

Yeah, that's a great example because I've seen many times, especially on long trips where I'm going somewhere and it takes four hours, but halfway through the trip there's an accident or something and it shows the color red on the route and then it's like, hey, if you take this detour, you're going to save 30 minutes or you're going to add 25 minutes, but there's no traffic. So that's a great example of how that can be used.

Sharmeen Roy (16:26.736)

Yeah, yeah, because it's taking all that data to give you that starting point, right, of where, how long it will take. But then as you kind of start on that path, it informs you into what the actual time would be. So I guess that's the easiest, I think, most applicable method. It's using similar kind of analogy.

Harsh Thakkar (16:30.638)

Hmm. Right.

Harsh Thakkar (16:42.446)

So, yeah.

Harsh Thakkar (16:48.814)

Yep. I'm actually gonna look it up after this interview to see if Google Maps and they use the same statistical method to figure out. They have to use something to give us that data, right? I'm not sure what they're using.

Sharmeen Roy (17:03.536)

Yeah, I'm sure they're using many different algorithms to inform it. And I think that's the thing, right? There's so much data. And it's not just in the Google has data, but even in health care, there's just so much data. And we just have to be better about how we utilize it, because it's as good as the data. It's.

Harsh Thakkar (17:14.19)

Mm -hmm.

Sharmeen Roy (17:31.248)

It's as good as you want to make it. So if you think about, and I know we're kind of going off tangent here, but if you think about all the data that's out there, so you have your EHRs that have all the lab data, your medication administrations, your diagnostic information. And there's patients who now have wearables. So you may have an Apple Watch that's

constantly tracking a lot of your health information or your smartphone. So I think there's all these different kind of methods of how we're getting data. It's just now how do we best utilize it, right? Because it can be very informative in personalized medicine is not just about dosing a drug, right? There's so many other components. And I think as pharmacists, we're sort of like,

we're gathering, if you're kind of taking care of a patient with a multidisciplinary team, you're gathering all that information. You're looking at culture results, you're looking at lab data, you're looking at any patient changes in weight or fluid status, and you're taking all that into consideration and dosing a patient. And what...

our platform does is kind of adds a little bit more information to that. So, you know, giving you a little bit more precise prediction for your dosing.

Harsh Thakkar (18:56.214)

Hmm. You're okay. And another thing that I was, this is for anybody who's listening and you want to like nerd out on this topic. It's pretty interesting because there are applications of Bayesian. So Bayesian statistics or dosing comes from like implementing Bayesian Bayes theorem, which is what you are kind of like explaining, you know, how to get

how to find the posterior with like the likelihood and bunch of other stuff. I don't know the theorems. I'm not gonna say it on air, but for anybody that's interested in figuring this stuff or learning more, look up Bayes theorem. I think it was a guy named Thomas Bayes who created that theorem. And that's kind of what Sharmin is explaining. And I'm sure if you look it up, you'll find some articles of how it's used. And...

I also saw that it's used in other industries like finance, right? I don't know for what reason, maybe for gauging like stock prices or market trends or I'm not sure, but I saw the article that says, yeah.

Sharmeen Roy (20:04.56)

Yeah, I'm not sure how it's utilized, but if you think about it, it's the probability, right? So you're kind of predicting the probability based on previous posterior information. So it really can be utilized anywhere, and it has been used in drug development for some time now. So it's just...

Harsh Thakkar (20:12.494)

Mm -hmm.

Sharmeen Roy (20:29.872)

taking those, our goal is to make it easier for the clinicians to have as much information as possible to make the best decision. So we know that in healthcare, the resource allocation is always a concern. There's just not enough people. So we wanna make sure that clinicians aren't spending time.

doing calculations and rather than spending time on doing calculations, they're spending more time with the patient and making sure that they're being adequately managed and optimizing their therapy. So however we can assist with that. And the software does help you do that complicated math at really, really fast at the bedside. So you can make those decisions faster and also make it safer for the patient. And it helps you identify any.

Harsh Thakkar (21:11.374)

Hmm.

Sharmeen Roy (21:22.8)

adverse event they may be having or any risk of kidney toxicity, for example, for some of our antimicrobials. So it's just to help make it safer using the patient information as well as the model that is available. So that's where the term for model -informed precision dosing comes because it's taking into account various factors and making it very...

Harsh Thakkar (21:44.878)

Hmm.

Sharmeen Roy (21:49.04)

available at the bedside. I think historically what these models have existed and the dosing has happened, but it's just not very easy to implement at the bedside. So that's what the software does is makes it easier and integrates it with the clinicians workflow.

Harsh Thakkar (22:05.806)

Okay. And not to get too technical, but do you know if the software has like any AI or machine learning algorithms that are running in the backend or like how does it intake the data and output whatever, you know, the clinicians need?

Sharmeen Roy (22:22.672)

Yeah, so it does have algorithms that are built in so it's you know, like I said, we what we do is implement the existing model that has you know for any given drug I'll use vancomycin as an example just because it's commonly used So that's an antimicrobial and it requires has a narrow therapeutic window and so there's you know, therapeutic drug monitoring that's done meaning

Harsh Thakkar (22:31.342)

Mm -hmm.

Sharmeen Roy (22:46.704)

when you give it to a patient, traditionally you start them on an established approved dose and you do monitoring of levels and then you adjust the dose. We've gotten much better at dosing that. It surprises me that it's a drug that was approved in the late 50s and we're still just as recent as 2020. There was guidelines that came out on what's the best method to dose it. And so,

What the platform does is allows for that monitoring to happen in real time and produce a dose based on not only the model that's implemented, but patient information and taking, you know, there's an algorithm that's built into the platform that allows for that quick output.

Harsh Thakkar (23:36.75)

When I was looking up the DozeMe LinkedIn page, I saw that it said somewhere that the platform also helps simplify AUC compliance and workflows and stuff like that. Before we go into that, do you want to maybe explain to our listeners who don't know much about AUC, like what it is, and then if you can explain how it does simplify the compliance?

Sharmeen Roy (24:03.056)

Yeah, so AUC is area under the curve. And so if you think about sort of a graph and how a medication, you know, when you give a medication, the curve it follows. So you give a medication, it gets to its highest concentration and then your, you know, your kidneys eliminate the drug. So it has a clearance. So area under the curve is literally the area that is under the curve. So that is what's...

Harsh Thakkar (24:15.566)

Mm.

Sharmeen Roy (24:32.752)

measuring the exposure, right? So that's the exposure of the medication for a patient. So for vancomycin as an example that I used, there was, in 2020, there are guidelines published that traditionally, the way we were monitoring vancomycin was trough base. And that is a concentration, which is at the end of an interval. So if something is given every 24 hours, for example,

at that 24 hour point, what's the lowest concentration? So that's trough and it's the minimum concentration. And so that's how, you know, so we would, we had kind of a range that was recommended that you would follow and you would measure the level and see, okay, is the patient, you know, between that 10 to 15 range. With the new guidelines in 2020, the guidance, the recommendation was to use AUC guided dosing. So the exposure,

Harsh Thakkar (25:06.318)

Hmm.

Sharmeen Roy (25:30.64)

of the drug rather than the trough because it's not the most appropriate method to dose. And it's also not the safest because there's lots of literature to show that there's an increase in nephrotoxicity if you're using trough -based dosing versus area under the curve. So it's not only more effective, but also safer for the patients to use AUC -guided dosing. So AUC calculations, you can do them. You can do them manually.

It's a long list of calculations that you have to do, which take a lot of time. And it's not easy to just kind of do that at the bedside when you're taking care of 30 patients at a time. So what the software does is, again, allows you to do that quickly. It calculates the AUC. You can indicate this is my target AUC I need to get to for this patient. And it will produce a

Harsh Thakkar (26:14.19)

Right, right.

Harsh Thakkar (26:27.214)

Hmm.

Sharmeen Roy (26:29.36)

dose recommendation according to that target. So if you give this dose, you'll achieve the target AUC of say, anywhere from 400 to 600.

Harsh Thakkar (26:33.23)

Interesting.

Harsh Thakkar (26:40.494)

Okay, okay, that's super helpful. So I want to ask you in your role at DoseMe, can you give us like a glimpse of a day in a life of what you are up to or what your team is up to? Because I'm curious to know, you're doing so many interesting things, but I want to understand what do you do in your work day or what does your team do?

Sharmeen Roy (27:03.12)

I wish it was the same every day, but every day is different. I oversee our product and our science team. So there's a big component of research and collaboration and working with our users or clinicians, as well as I oversee our customer experience with implementation. So any new hospital that is signing up to get our software, we...

Harsh Thakkar (27:04.566)

Yep.

Sharmeen Roy (27:28.88)

help, you know, we have a whole implementation process that we work with their site champions or clinical champions that they identify to get it, you know, the software rolled out, all the training and implementation that goes along with it. So, you know, that's part of what I oversee as well as our product and science team. So I work closely with our bioinformaticist and our develop, you know, our technical team just to kind of

They work on feature updates, new model implementations, because as I guess more awareness is around precision dosing and model -informed precision dosing, there's a need to increase the number of models we have. So it's not limited to just a handful of medications. We're expanding.

to specifically there's a big growth in oncology space because of the toxicity associated with those medications. There just needs to be a better way and more individualized dosing for those patients and for those drugs. So we're sort of constantly evaluating the need, the pain points, and also where we need to kind of focus as far as drug development of the models.

So I work closely with our science team and identifying those and implement them as well as with our customers. The nice thing is that with all this data that we have, you have essentially real world data, right? So this can be used for a lot of research and maybe either streamlining processes or also updating protocols in their patient population. So...

I work closely with our customers to see where we can support them in the research initiatives they have and how we can support them with our platform to get the data they need to be able to conduct some of that research.

Harsh Thakkar (29:25.87)

Mm -hmm.

Harsh Thakkar (29:33.966)

Okay. And for any of our listeners for whom this topic was something new, maybe they learned a little bit about precision dosing and Bayesian dosing, can you recommend any reading materials or any blogs or any books where if they want to dive deeper or learn more, what do you consume in terms of information to stay up to date with all the latest trends in this area?

Sharmeen Roy (30:01.104)

Yeah, sure. So if they want to learn more about just kind of sort of general information, you know, our website has a lot of resources with about Bayesian dosing and other specific drug models and evidence around that. As far as what I do to keep up, I'm involved with many organizations. You know, we it's it's sort of a continuous learning process for all of us, right? There's so much new information.

that is continuously being published. And I think more, the nice thing is that, so recently, because there's a focus in precision medicine as well as model -informed precision dosing, recently I was at a conference, ECMID, that's one of the largest European infectious disease conferences. And I was really, really pleased to see that there's...

there's a lot of focus on model -informed precision dosing and how to bring this to the bedside. So there's a lot of great work being done in those groups as well as a lot of clinical pharmacology groups. We're constantly looking at where we can apply this, what other medication we're learning more about to see, whether we're seeing toxicities and we need to improve on dosing. So yeah, it's, I think,

just being involved with those organizations as well as keeping up with the research that's coming out of there.

Harsh Thakkar (31:30.926)

Yeah, yeah, thanks. Thank you so much for coming on. You know, this is what I love about this podcast is because if I didn't have this podcast, I would not have had the opportunity, you know, to have this conversation with you and learn about some topics that I don't know much about. So thank you so much for coming on to the show. And before we sign off, do you want to, you know, let audience know where they can connect with you or?

Maybe if they're interested in DoseMe or looking at a demo, do you want to just give the social media information where they can reach out to you?

Sharmeen Roy (32:06.64)

Sure, they can find me on LinkedIn. That's usually the only social media platform I would say I'm active on. You can also always go on our website to get any contact information on dosmierx .com. So yeah, happy to answer any questions. If they're interested in seeing the platform, we can certainly arrange for that as well.

Harsh Thakkar (32:12.27)

Yep.

Harsh Thakkar (32:28.078)

All right. Thank you, Sharmin. I really appreciate your time and wish you a good rest of the week. I had.

Sharmeen Roy (32:35.344)

All right. Thank you, Harsh. It was nice to be here and it was a pleasure meeting you.

Harsh Thakkar (32:39.63)

Thank you.