MJM MedTalks
A Conversation with Dr. Ahmad Haidar

S01E03

Dylan Langburt, Khiran Arumugam, Ahmad Haidar, for the McGill Journal of Medicine1
Published online: May 1, 2023

1McGill University
mjm.med@mcgill.ca

Abstract

McGill Journal of Medicine (MJM) MedTalks is a podcast series where members of the McGill Faculty of Medicine and Health Sciences are interviewed on topics related to career, research, advocacy and more. The aim of MedTalks is to open a space where faculty members can share information and advice for trainees in healthcare and medical sciences. In this episode, MJM Podcast Team members, McGill medical student Dylan Langburt and MSc candidate Khiran Arumugam interviews Dr. Ahmad Haidar about his work in the Artificial Pancreas Lab. Dr. Haidar is an assistant professor from the department of Biomedical Engineering, Faculty of Medicine and Health Sciences here at McGill. He leads an interdisciplinary research program that applies feedback control theory and mathematical modeling to diabetes, psychological, and clinical problems. Since 2011, Dr. Haidar’s research aim has been to develop and clinically test novel artificial pancreas systems with the use of Bayesian modeling and isotope tracers to study the pharmacokinetics and pharmacodynamics of dual-hormones (insulin and pramlintide). He is the first to develop the dosing algorithm of the artificial pancreas system. This interview will potentially cover concepts such as the artificial pancreas, diagnosis and treatment systems, automated delivery systems, diabetes (Type 1), biomedical devices, glucose-isotope tracers, glucose physiology metabolism, etc. Today’s conversation is divided into three parts: 1) Questions regarding the history from the discovery to the evolution of insulin; 2) Questions focusing on specific objectives in Dr. Haidar’s lab; 3) General advice for medical/research students. The show notes include a transcript of the podcast, a more detailed content overview, glossary of important terms and resources and references. This podcast is produced and edited by MJM’s social media team members Dylan and Khiran with input from the entire MJM Podcast Team. Please see our website www.mjmmed.com for more information, including a link to show notes.

     

Content overview

0:00 Introduction;

0:33 Episode Overview;

1:26 Part 1: History of insulin;

1:26 Glucose sensor technology;

2:52 Glucose monitoring implications;

4:57 Part 2: Dr. Haidar Artificial Pancreas Lab Research;

4:57 What is an artificial pancreas?;

7:28 Why a randomized controlled trial in diabetes research?;

10:35 The aspect of artificial intelligence behind the artificial pancreas;

14:49 Differentiating the benefits of combined hormone effect in the artificial pancreas;

20:23 The artificial pancreas' influence on carbohydrate counting;

24:31 Where do you see the field of type one diabetes treatment moving?;

28:15 The importance of interdisciplinary and interprofessional in healthcare;

29:22 Part 3: General Advice for Trainees;

29:22 How did you find your way in academia?;

32:07 The implication of medical devices and other digital health technology in healthcare;

Outro;

Glossary

Links and papers

Transcript

0:00 DL: McGill Journal of Medicine (MJM) MedTalks is a Podcast series where members of the McGill Faculty of Medicine and Health Sciences are interviewed on topics related to career, research, advocacy and more. The aim of MedTalks is to open a space where faculty members can share information and advice for trainees in healthcare and medical sciences. In this episode, MJM Podcast Team members, McGill medical student Dylan Langburt and Khiran, an MSc candidate, interview Doctor (Dr) Ahmad Haidar about his work in the Artificial Pancreas Lab. Dr. Haidar is an assistant professor from the department of biomedical engineering, faculty of medicine here at McGill. He leads an interdisciplinary research program that applies feedback control theory and mathematical modeling to diabetes, psychological, and clinical problems. Since 2011, Dr. Haidar’s research aim has been to develop and clinically test novel artificial pancreas systems with the use of Bayesian modeling and isotope tracers to study the pharmacokinetics and pharmacodynamics of dual-hormones (insulin and pramlintide). He is the first to develop the dosing algorithm of the artificial pancreas system.

1:03 DL: Today's conversation is divided into 3 parts. Firstly, questions regarding the history from the discovery to the evolution of insulin. Second, questions focusing on specific objectives and doctor Haidar's lab, and finally, general advice for medical and research students.

1:19 DL: Doctor Haidar, thank you and thank you for agreeing to do this interview with us.

AH: My pleasure. Thanks for having me.

1:26 DL: Firstly, insulin is talked about a lot in terms of the discovery and the treatment of diabetes, but I think less talked about are the glucose sensors. The name diabetes mellitus refers to the fact that patients produce large amounts of sweet tasting urine. Can you describe the technology behind the early glucose sensors and some of the challenges in progressing to real time sensors used today?

AH: Yeah, absolutely. So, in the past actually we didn't use to call them glucose sensors, we used to call them glucose meters. So, in the early days we used to measure sugar level in the urine. And I think around the 70’s we were able to measure them using capillary blood. So, these are the glucose meter that you commonly see. And, what happened in around 2000, we started to develop technologies to measure sugar level in a continuous invasive manner. And this was really revolutionary because it's the first time you don't only see sugar level at four or five times during the day and you miss what happening in between, you actually can see the sugar level changing at every one minute or five minutes interval and that gives you an idea about the average sugar level as well as the variation of the sugar level. This is actually what draws up the whole research area of artificial pancreas systems, because before that we just couldn't measure sugar level in a continuous manner.

2:52 DL: And so have you seen improvements in treatment in people who are monitoring their glucose, like, continually throughout the day?

AH: Yeah, absolutely, because in the past, let's say someone using an insulin pump. And every time they need to know their sugar level, they need to prick their finger and measure their capillary blood. They do it three, four, five, or maybe seven times a day. These are the most adherent, like people with type one diabetes. But, then let's say at night before you go to bed, you might get your sugar level. You are at 15. We call that hyperglycemia. You think you need 2 units of insulin. Now, the problem is that from day-to-day there is large variability in your insulin need. So, if you do exercise during the day, you need less insulin at night. If you eat a high fat meal, for example, pizza for dinner, you need more insulin at night. If you're stressed, you need more insulin at night. So even though your sugar level is at 15, you don't know exactly how much insulin you need to bring your sugar level down to target. So let's say you have two units, but actually you need 2.5 or you need 1.5 and that makes it challenging. So, people often over or under correct based on their sugar level. However, if you have your continuous glucose monitoring system or the glucose sensor, currently you are at 15, it tells you now you are 15 and in 10 minutes you are at 14.5 and so on. You can automatically adjust the insulin delivery and the pump based on the sugar level as they change throughout the night. So, you give a bit of insulin, your sugar level drops. You give a bit more and then every 10 minutes you adjust insulin delivery and that allows you to give the right amount of insulin for that particular night. So, really the glucose sensor didn't only improve our ability to know sugar levels, but also improved our ability to give the right amount of insulin.

4:57 KA: I am going to be talking to you about the work in your lab, in the artificial pancreas lab. So, in light of the discoveries and the advancements made in the care for type one diabetic patients, how did you end up in this field of work relating to developing the artificial pancreas. And at the same time, what exactly is the artificial pancreas for our audience?

AH: Yeah, perfect. So, I don't have an exciting story. I get that question quite a lot. My story is a little bit boring, but I will say it anyways. My background is in control engineering. So, I did electrical engineering for my Bachelor’s degree. I did a mechanical engineering Master’s. And control engineering basically, for example, an autopilot for an airplane. Or, if you want to control the temperature in a chemical reactor, this is control engineering. So it's more, it's a theoretical or applied mathematics field that you can apply to mechanical system, electrical system, or a chemical system. Now when I was starting my PhD, this is around the same time where glucose sensors started to come to the market. So again, the ability to make your sugar level in a continuous manner and that allowed us for the first time to be able at least in principle to build what we call an automated insulin delivery system. Or, the artificial pancreas where you can automate insulin pump delivery based on the sugar level reading and mathematically this problem is a control engineering problem. So, you need to decide how you're going to give your insulin to bring sugar level to target at the same mathematical problem as trying to build an autopilot for airplane or trying to control temperature in a reactor and so on. So, this is actually how I got into this field. I was looking for something in the medical field and these sensors came around the same time, so it was just perfect. Perfect timing for me. Now, I know I talked about it briefly, but the artificial pancreas is basically nothing but a smart insulin pump and the name “Artificial Pancreas’ is actually misleading. So, I'm not sure I like using that term even though a lot of people use it in the field including myself. But it's like calling an airplane an artificial bird. I mean an airplane is not an artificial bird, it is a completely different mechanism, uses physics and mathematics to mimic the functionality of the bird which is to fly and the same in an artificial pancreas which is just a smart insulin pump. A mechanical system that automates insulin delivery based on the sugar reading and it just mimics the functionality of the pancreas.

7:28 KA: So, you conducted a randomized control trial and what I want to know is more about why you selected a randomized control trial. Is that something very common in terms of a study design for diabetic diabetes research?

AH: Yeah, absolutely. So, we run actually several randomized control trials. So, medicine is a bit different than engineering. In engineering, if you want to build a device, usually you are driven by what we call a specification of the device. If someone tells you: “I want you to build me a motor that gives me this or that efficiency”. You design it. You test four or five motors. If they give you that efficiency, you're done. Now, in medicine it is a bit different because the first thing you have, you have large variability between individuals. So, you need to test your system and multiple people, but also you need to know how do you perform compared to what we call a control arm or a comparator. In the drug side, people often talk about placebo. If I give you, for example, a drug to improve or let’s say to reduce body weight or to improve, for example, depression, you can just give people the drug without having a comparable group with the same characteristics who have placebo. So, you can see how much improvement you are getting compared to the other group because maybe when you bring people into clinic to give them the drug, you know, the fact that they interact with the physicians, with the nurses, they become a bit more conscious about what they eat and then they change some behavior stuff and then they reduce their body weight irrespective of your pill. For you to know exactly how much your pill is reducing body weight, you have to have a different group who will go exactly the same in study design and in study procedures, but they take placebo. And, in the same concept, we have for the device that we compare the device to existing therapies to see how much improvement we have over existing therapies and both these therapies will have the same study procedure. So, randomized control trial is really the highest level of scientific evidence. Maybe another quick example, you know during COVID, I remember we had Trump going around and saying that the hydrochloric that he was suggesting actually would work. If you see actually that paper, they had to have the paper in France and they had I think 12 or 15 patients with COVID. They gave them the drug and they indeed, all of them recovered. So, really if you look at that study and you're not really an expert and for example study design you will say well all of them recovered. We know from the media people often die from COVID, then that drug is actually working and that's it. This is what we should be giving to people. But, actually, no. We don't know because there is no control arm and these people if you didn't give them the drug probably would have recovered, and maybe would have recovered even faster. This is why it's important to have a control arm, because that actually gives you exactly how much benefits you're getting compared to standard treatments.

10:35 KA: I guess what was very interesting about your work is the fact that you use artificial intelligence as part of an automated insulin delivery system and you mentioned it briefly when you were talking about the artificial pancreas itself. But could you elaborate just behind the integration between biological, mechanical, electrical and mathematical approach and the artificial intelligence aspect?

AH: Yeah, absolutely. Maybe I should clarify first that in our lab, we actually don't build the glucose sensor, and we don't build insulin pumps. So, we use commercial glucose sensors and commercial insulin pumps. What we build is the mathematical dosing algorithms that actually decide how much insulin to give based on the sugar levels. Now, the main challenge here is that you have large variability between people and you have large variability even within the same individual. So, what I tell people building an artificial pancreas or an automated insulin delivery system is actually much harder than building a controller for an autopilot for an airplane, because imagine you have one controller that should work for different or twenty-eight different models of airplanes and also why the airplane is flying. The dynamic changes then it doesn't know you lose the wing or something like that and the algorithm has to be adaptive so that it actually still continues to fly safely. The same for us like we have this algorithm that should work for different individuals and within the same individual should actually address all kinds of potential variability that you get. Now, we don't use artificial intelligence. Again, this is a term that some people use sometimes just to you know sound a bit more fancy. We use more what we call control engineering algorithms or something that we call model predictive control. So, we have a mathematical algorithm inside each pump or inside the phone that is running the algorithm. That algorithm represents what we think the patient sugar level will look like if we give a certain amount of insulin delivery. Based on that mathematical algorithm, we always choose the best insulin profile and the algorithm updates automatically the model based on the sugar level using what we call a Kalman filter and this is a technique commonly used in aerospace or GPS, for example. So if we use that, yeah.

12:52 KA: Interesting. It's good to know that there is a lot of differences between building a medical device versus a mechanical device, for sure.

AH: Yeah but another device like an airplane. You test it, you fly it four or five times and you do the same on your algorithm and that's it call today. We just can't do that with devices unfortunately, yeah.

13:12 KA: Do you somehow find ways to integrate the use of computer simulations?

AH: Yeah, absolutely. Yeah. Because, I mean, for us, when I run our clinical trial. Now, depending on the studies, some studies cost us $300,000, some studies cost us $1,000,000. Now, before we run this study, we need to make sure the algorithm is actually good and is going to perform. I mean, the last thing you want to do, you spend $1,000,000, you run the study, it takes you a year and a half and then you realize, oops, we should have designed the algorithm slightly differently. So, one way to reduce that risk, we use what we call computer simulations. A little bit similar to like flight simulators where you put a pilot before you put them in the real airplane. We actually have a mathematical model for different individuals living with type one diabetes that we run on a computer and we test our algorithm using computer simulations. So, that allows us to optimize the algorithm before we run a clinical study, but also allow us to test scenarios that it will be either impractical or unethical to test a clinical trial. But, you know, that might happen in a clinical trial. Let's say in one of the scenarios you have infusion failure which means that the insulin pump just delivers insulin, but the insulin doesn't go through the body because the infusion set of the insulin pump has failed. Now testing that in clinic is possible but practically difficult, potentially unethical. So, we test that simulation and we know that if it happened in real life at least we have a feeling about how the algorithm would work.

14:49 KA: In your RCT, you look also at insulin alone, artificial pancreas versus insulin and pramlintide artificial pancreas. Now, my question is, what is the purpose of having one sample include a mixture of insulin and pramlintide?

AH: Yes, pramlintide is actually is an analog for the hormone called amylin and that hormone is secreted with fixed ratio of insulin and people without diabetes. But, then when you have diabetes or type one diabetes, you lose the ability to secrete insulin, as well as amylin so that drug actually slows gastric emptying, increases the satiety, and produces the glucagon level. So, it has some effects that can prevent what we call hyperglycemia or high sugar level. So that drug is called Pramlintide, has been in the market for a while, but people don't use it because an injectable medication that you have to give three times a day. And the last thing you want if you are living with type one diabetes is to have a pump and on top of that you have to give yourself extra injections. So, what we're proposing in our lab is that if someone can build what we call a “co-formulation” so we can bring these two insulins together well. Well insulin and pramlintide, together in one drug and you put them in the insulin pump can you improve the glucose control? So this was the hypothesis, but we didn't have actually a co-formulation. So what we did, we had two pumps and we simulated the co-formulation by delivering both hormones with fixed ratio. And we run a randomized control trial to compare that new dual hormone system versus an insulin alone system to see how much improvement we are getting. And you run multiple trials and we have short benefits and based on our data, there's actually a company in France called the Adocia that are currently developing a co-formulation and we recently completed for them a study which was first in human trials. Actually, we did here at McGill. The first time ever someone tested this drug and human that we did here based on the earlier studies that we did. That company actually felt very strongly about developing a co-formulation and they gave us to do testing with that so.

16:57 DL: A quick follow up on that in terms of the co-formulation, because insulin is delivered in terms of units, right, a patient might give themselves 5 units in the morning. Is it a fixed ratio of insulin to to pramlintide that is physiological?

AH: Yeah. So this is a good question. So, we actually deliver one unit of insulin for every 6 micrograms of pramlintide. Now, we like to think it's physiological, but the truth is that it's not fully physiological because the pancreas when they secrete, when it secretes insulin and pramlintide with fixed ratio you expect to see in the plasma both hormones are with fixed ratio. But, when you need to deliver actually insulin and pramlintide subcutaneously, the absorption pattern of the two are different. The pramlintide absorption is very fast while insulin absorption is actually slower. So, if you give both are boluses, you peak of pramlintide after 15 minutes and you peak of insulin after one hour. So, in the plasma it is not fully physiological. I mean you replace amylin that is lost to a little bit more physiological, but we're again, again we're not building an artificial pancreas. We just like try to replicate some functionality using other techniques.

18:15 KA: I think another question I want to know is, do you have any experiments that have tried to combine the use of insulin with the glucagon? Mixtures of those two together? I know it's expensive in the market, but has there been any advantages of even hypothesizing that?

AH: Yeah, absolutely. It's actually, what we have done. So before we did our pramlintide study, we have done several insulin and glucagon studies. I think we run maybe 8 or 9 randomized controlled trials. Again, we use 2 pumps and what we showed is that if you add the Glucagon you actually get better glucose control, so you have slightly less hypoglycemia and also allow you to have more aggressive insulin delivery which allows you to improve hyperglycemia. And you can think about it like driving a car, if you have only got and every time you need to break, you have to take your foot off the gas and wait for, I don't know, 20 seconds before you can stop. You will drive very slowly, but if you know you have a break, you will drive fast and a break whenever you need to break. But as you mentioned, glucagon is expensive and glucagon also, at least the commercially available one now, are not compatible with pump use. So if you put them in the pump, they're not stable for three days. However, there are few companies who looked at developing stable glucagon. And now we have stable glucagon that was developed by a few companies which is suitable for pump use. And that the company called Beta Bionics that actually started by two professors in Boston, so one of them at Harvard, the other one at Boston University. And now they have that company and they're trying to commercialize a pump which has dual chamber, one insulin and one Glucagon. But my understanding now there are at the stage where they need to raise money to run this big, big trial because they need to show also the delivery of a glucagon is actually safe chronically. We know glucagon is safe if someone has to be in hypo we give them glucagon and they wake up. We know that particular case is safe, but we don't know if someone gives them glucagon every 10 minutes or every two hours over the long term, is that actually safe. So they need to test that in their study.

20:23 KA: Knowing that carbohydrate counting focuses on carbohydrate as the primary nutrient affecting postprandial glycemic response in patients with diabetes, how does the electromechanical artificial pancreas influence this meal planning approach? So what outcomes do you have you looked into or do you expect to assess in a work like this?

AH: Yeah, that's really good question. So, people with type one diabetes, currently they're asked to do what we call carbohydrate counting. So, that means for every meal they eat, for every snack they eat, they need to look at it and estimate the amount of carbohydrate. And based on that amount, they have individual insulin to carbohydrate ratio and based on that they decide how much insulin to give. So, let's say someone’s ratio is one to ten, which means that when they have 50 gram, they take 5 units and they when take 30 gram, they take 3 units and so on. Now, I don't need to convince you how difficult that is. I mean, just imagine you go to a restaurant and you have no idea what they put in the food or imagine you are in a social event with friends and you have to look at the food and you have to think about how many carbs are there and you have to give yourself insulin. So, there are some literature that shows that it's quite not only challenging for people, but also can put some pressure in social scenarios where they feel potentially embarrassed and so on. So, what we're looking at and other people are looking at that. Can we make these systems smarter so people do not need to do exact carbohydrate counting. So, the next potential step is that while instead of doing carbohydrate counting maybe they will choose if the meal is small, medium, large or very large in terms of carbohydrate. So, instead of having thinking this is 45 gram or 55 or 50, well this is just medium. So at least this simplifies the process of carbohydrate counting and because the insulin automated system looks at the sugar level and adjust automatically, then if you are not exactly precise with your carbohydrate entry to the system, the system will still function well. Now, another step would be simply announcement. We mean that at meal time you press a button and that's it. The system will give a small insulin and then based on the sugar level, we'll give you the remaining insulin. Now the most extreme scenario, we have a fully closed loop where you tell people to wear the pump, wear the glucose sensor, eat what you want and don't worry about it. Now that one have been tried by multiple people and we couldn't achieve good glucose control because insulin is very slow. Because imagine you eat 100 gram meal of carbohydrate, the sugar level observed very quickly you start to give insulin, but insulin peaks after one hour and by that time your sugar level is already high. So, with current insulin analogs we just cannot get fully closed door. Potentially we can do that maybe if you add pramlintide, if we add SGLT-2 inhibitors but not fallible insulin alone. But it's an active area of research that us and other people are trying to work on just to improve quality of life of people living with type one diabetes.

23:30 KA: OK. So, yeah, it makes sense. If there's a lot of limitations with the insulin levels peaking right away in an hour or so, that doesn't make it easy. But of course, carbohydrate counting is a big focus in this field of research too.

AH: Well, actually there is, maybe I could comment on that. There's large variability and insulin absorption. So some people, they get a peak after 30 minutes when they give a bolus. Some people they get a peek after two hours. And it's just hard for the algorithm to handle all these cases. You know, like you have to have an algorithm that knows exactly what to do, time to peak and also changes over time within the same individual. So if someone is running an insulin pump. Currently, the recommendation is you change your infusion set every three days, so the catheter. And there's something we discovered recently that is called “Tamburlaine effect” means that the insulin absorption becomes faster by around 10 minutes every day. So on day one of the infusion set, you have a different pharmacokinetics profile than day three of the infusion set. So that even adds complexity to the whole, to the whole thing.

24:31 DL: So as a medical student, we're starting to see the use of some sort of off-label medications that are used for type 2 diabetes like SGLT-2 inhibitors in type 1 diabetics. And there's also some promise in terms of like autologous beta cell transplants from induced pluripotent stem cells. Where do you see the field of type one diabetes treatment moving? Do you see it moving more towards medications to transplants or the more widespread use of medical devices?

AH: Yeah, good question. I mean I think devices definitely in the last 10-15 years took over. I mean if you look at the use of the glucose sensors, I think within three or four years it actually increased by tenfold in young children. So, definitely the devices are taking over, but at some point they will saturate. And what we know now like initially I remember 2010-2012 people are talking about artificial pancreas system. Everyone was expecting once we get this to the market, everyone would have perfect control and that's it. What we know now is that even when you are on an artificial pancreas system or an insulin pump, half the people with type one diabetes still do not achieve its maximum targets. So these systems are not perfect yet. Now they are getting better and better. So you know if they have faster insulin and potentially better algorithm you get to improve. But I don't think they will ever be perfect. So there are definitely more room for drugs to come in. And we do our studies, actually, we do studies with SGLT-2 inhibitors. We just published 2 trials. But the complexity with this drug is that if you look at SGLT-2 inhibitors, you get big glycemic benefits in type one diabetes, you potentially have some renal benefits, but you have an increase in DKA risk, which is life threatening. And This is why actually FDA rejected the approval for FDA and type 1 diabetes. In Europe, they had approval for certain doses, but also for subpopulation of type one diabetes. But the, now you have GLP-1s studies in type one diabetes, data has been conflicting. Now you have long acting GLP-1, semaglutide, for example that lasts for one week. We don't have much data in type one diabetes. Our group actually just started the study with that. Now we have tirzepatide which is for weight reduction which is more for obesity and Type 2. Now will someone test it in type I, so I don't know. I mean again, these are developed mostly for Type 2. Also, the challenge there is that companies don't have interest in getting labeled for Type 1. And then there's also the issue, I mean, if you look at SGLT-2 inhibitor, AstraZeneca withdrew. Like they had the license to sell for type one, but they withdrew it in Europe because they were asked by the regulatory agencies there to have a black box warning that this drug can induce decay in type one diabetes and they were worried that this little symbol will impact their sales in people with chronic diseases or heart failure or type 2 diabetes. So they withdrew the license. Of course, you know like JDRF for example, they were unhappy, so juvenile diabetes research foundation. They were unhappy with that and they made a strong statement about this. But from companies perspective, they just don't have much incentives actually to push for these drugs in type one diabetes and I think we'll see more and more off-label use.

28:15 KA: I want to also talk about if there is any interdisciplinary and interprofessional relationships within your lab with other professionals, healthcare professionals, with the work you do. Yeah, absolutely. So our love is. It's truly a multidisciplinary. So we have currently we have around 10 people, half of them they have what they call engineering backgrounds or some electrical engineering, some biomedical engineering and they do the system development and some of them they do algorithms and the other half they're mostly some like we had in the past were nurses, we have one clinician and technologist and we have people who come from basic science background, but they run the chemical components of the lab. And for me, a true interdisciplinary lab is the lab that publishes high quality papers and more than one field. And we're lucky that we can publish good quality papers and engineering journals and we have good quality papers and clinical journals. And yeah, so.

29:22 KA: Interesting. I think that interrelationship within that lab that connects to different aspects of research definitely reinforces the objectives you guys have said. So that definitely does contribute to that. I guess another question regarding just for the advice for students is how did you find your way in academia?

AH: Yeah, I mean. I mean, I've been always enjoying academia and school. And honestly, I mean you need two things to succeed. So you need passion. So someone has to love research and spend hours doing research, but also you need luck. You really need it, I mean again in my case before I worked in this area of research, I was going to work on thermoforming machines and I'm sure if I did my PhD in that topic I wouldn't have the same career that I had now and the reason I moved to work on insulin pump systems is that someone developed a glucose sensor 2 years before I started to do my PhD. So this is really pure, pure luck. And also sometimes when you do your PhD you have good luck, good supervisor stuff that are not really under control that actually work well for you. But also, it's not like alone will make people successful. You need to be , you need to have passion, you need to keep trying and sometimes an opportunity comes and you just like take it and move on with it. So I'm not sure if that answer your question. I mean I think if you're asking about clinician. When we lack clinical scientist. So definitely if you look at 30 years ago, ago, 20 years ago, we used to have a lot of people who are 50% clinician they spend their time in connect and 50% in the lab doing research and a lot of them, they do wet lab research. Now we don't have that. It's really rare. We have people doing 100% time clinic. We have people they do 100% of their time research. But then that's what it prevents. Sometimes the clinician will be driven by problem they see in clinic they actually solve in their lab and sometimes you have this disconnect between the two. So if there are any med school students maybe listening to me, I would say that we really need people like you to have interest in research and the earlier you start either volunteering or maybe spend, you know, if your in you’re first years spend 3 months doing research maybe at the end of your fellowship try to do one year to do research. We really need people like you because this is the kind of people that generating new evidence and improve clinical trial guidelines and basically bring in new discoveries.

32:07 KA: Very well said and you definitely answered all the questions I had next. But I guess one last question I would want to ask is in terms of developing a medical device that improves the delivery of healthcare treatments and services for chronic conditions. Whether it's involving the different, various different digital health technology. What advice would you give to students or faculty when with the limitations and the challenges that you've noticed throughout your lab work?

AH: Yeah. So you definitely need to understand the regulatory process. So medical devices, so they usually they are classified. So here in Canada we have four classes. In the US they have three classes and the highest if the device has higher risk, they would have high class, The Class 3 medical devices are for like closed loop system would be Class 4 medical device. Now class one medical device, for example a wheelchair, actually it's considered a medical device. A toothbrush is a medical device and depending on the classification of the medical device, the regulatory requirement either to do testing or to try to commercialize it are very different. So, really the most important skill you have or maybe the most important knowledge you can try to acquire is related to the regulatory process to test and commercialize medical devices. I mean the other thing is that some people are, it's a bit unrealistic to expect to develop a medical device in academic lab. Like I people tell me you're developing an artificial pancreas, but the truth is that I'm actually not developing an artificial pancreas. I'm building a prototype system and then doing clinical trials to show that it works and then we collaborate with industry to try to bring these devices to the market and if you're talking about insulin pump it will cost around 50 million to 150 million to bring an insulin pump to the market. The glucose sensor around 200 or 250 million. So still much cheaper than bringing a drug to the market, but still a large amount of money and we just don't have that amount of money in academia. And really what we can do as academics is build prototype, do proper strong rigorous clinical trials to show that it works and then try to collaborate with industry who will take this invention and bring them to the market and who are willing actually to do, to put that investment.

34:29 KA: So I guess in a sense that that's sort of how you bring up commercializing the work you do. I guess that's how you commercialize your results and to a certain extent, right?

AH: Yeah, exactly. I mean I remember 2017 or 16, I thought I could build an artificial system myself and I went around to try to talk to investor. But then within three months I realized that I was dreaming, I had to quit McGill and then I had to try to raise 50 million and I just realized that this is an extremely difficult job, I just don't know how to do it. So I want to try to do the easy approach by talking to companies, see who has interest in commercializing a system and then we can contribute with some components and in that system and this is what our approach has been and we have done this with basically multiple companies and different technologies because we do also work in the open loop side. So, people who are doing multiple daily injections, they're wearing glucose sensor, you can make an adjustment to their insulin dose on a weekly basis. So, it's not like an insulin pump where you do that every 10 minutes. It's more on a weekly basis, you can adjust the dose. So, we also have algorithms there and in order to commercialize we try to actually collaborate with industry, yeah. The other important thing that I learned is that you really need to patent any invention you have. I mean, unfortunately, now you have our friends at the MNI, at McGill, they keep pushing the idea of open science. I think that is a big mistake. Yes, there are some devices and drugs where if you publish, you can commercialize them. But, really when you talk to companies, the first question they ask you, do you have a patent? Is your invention protected? Because again from their perspective why would they spend 100 million to bring your invention to the market if they're not going to get the return and for them to be able to get the return, financial return for that it needs to be protected by patent and then they can spend that money and they only have a window of 20 or 25 years to actually make the profit and after that you get either by similar or you get generics. And I think I just don't understand this honestly obsession with open science because that in my view you actually harm your chances of commercializing and really the best thing you can do if you want to commercialize is to patent protect it and then go around and try to find someone who actually willing to put that money. Even if you want to have your own startup, you need to get an investor. Why would an investor give you 3,000,000 if you are going to give them a lecture on open science? They will give you 3,000,000 if they know they can get back 6,000,000 in five years. And the only way you can do that if you're actually protected. Now there are some cases where publishing is actually good. So, I'm not saying everything has to be patented, but really in most fields patent is the way to go.

37:24 DL: Very interesting, interesting point. I think open science is interesting, but maybe a little bit too idealistic for the world that we're that we're living in today.

AH: Yeah. And you can still publish. I mean the idea of patent is basically the concept behind it is a contract between you and the state where you disclose your invention and they give you exclusive right to use it for a short period of time. But by forcing you to publish it, you actually improve innovation because other people will see your invention and they come up with their own idea and they can improve on that. But if you're going, let's say tomorrow we abolish the whole patent law and we don't have any more patent. What will happen is that everyone will be hiding their invention, they will not be publishing it and if anything it will have a negative effect because what people don't get is that most inventions happen in industry. I mean we would like to think like, you know, in academia, universities, is where everything happens, but the truth is that the big invention, the exciting stuff, they really happen in industry and these people, you know, the system is that if they don't make money, they will not make these inventions and you really have to protect their work by patenting it. And again, in my field we do algorithms and if there are really no patents, why would any company publish their algorithm? I mean they published you know, cause enough for them. Like the algorithm is not just the equation that they have, it's actually the amount of clinical testing they did on the algorithm. So, if they have an algorithm and they spend 20 million testing it in clinical trials and they just publish it. Well, another company would come, would take that algorithm and they will not be having to spend 20 million to make sure it works and then they can go to the market with the same product with half the price and you actually end up being bankrupt and then you want it just doesn't work. Now again some examples maybe you know I hear from time to time examples where open science has promoted innovation.

39:15 DL: Well, I just wanted to say thank you so much to you Dr. Haidar for taking time of your week to to chat with us. It's been, it's been really interesting to to hear your opinions and and your experience and we look forward to seeing how the artificial pancreas and how your work evolves and changes the treatment of diabetes.

AH: Thank you very much and my pleasure.

39:38 KA: Thank you to our audience for joining us on another episode of the McGill Journal of Medicine, the MedTalk series. This podcast was edited and produced by MJM's podcast team members. Feel free to reach out to us on Twitter or Instagram and McGillJMed or by e-mail. We would love to have your feedback and don't forget to join us again for our next episode.



Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.