Welcome And Meet Shannon Lantzy

Speaker 0:03

David Klonoff 0:15

Hello. Welcome to Diabetes Technology Report. I'm David Klonoff, an endocrinologist at Sutter Health in San Mateo, California. I'm with my co-host David Kerr, who will introduce himself.

David Kerr 0:29

Thanks, David, and hello to everyone. As usual, I'm speaking to you from Sutter Health, which is in Santa Barbara, California, and today we have a very special guest. Hello, Shannon Lantzy, who's speaking to us from Maryland. It's lovely to hear you today. We'd like to begin these podcasts with finding a little bit about the person. So how come you've ended up being interested in diabetes?

Shannon Lantzy 0:56

Thank you so much for having me on today, and thanks for the question.

From NASA To FDA Decisions

Shannon Lantzy 1:00

My background is sort of non-traditional to diabetes technology, I would say. I started in math undergrad and then was a technology consultant and strategy for NASA, and then went in to get a PhD in information systems. And it turned out to be a human decision-making, both the macro-scale economics and micro-scale consumer behavior and behavioral economic experiments in my PhD. And when I realized I wanted to apply that work, I went to FDA, applied for a job, and then instead of getting a job at FDA, I got that job, but I moved into a consulting role and created a regulatory innovation group. And one of my favorite projects was with the Center for Drug Evaluation and Research, CEDAR. We that project was to apply decision science methods to some of FDA's most difficult or close call decisions. And I had the opportunity to support the review team that was doing the first review of soda-glyflosin, which is an SGLT2. And the the the one of the critical questions was could the evidence supplied for patient preferences flip a decision from not tolerating a small but strong signal of DKA risk versus all of the other benefits like weight loss and glycemic control? And ultimately it was a it was a split decision and a negative decision of the advisory council. But through this process, I learned about how difficult it is to make population-level decisions for people who have very, very disparate preferences among a, I will say a patient population, but a person with diabetes population.

Cybersecurity For Body Worn Devices

Shannon Lantzy 2:55

And then later I moved on to CDRH, Center for Devices, and focused on cybersecurity and met with some of the folks who were in charge of writing both the policy and evaluating the cybersecurity of AID systems and the intricacies of cyber requirements for devices that live on the body and administer life-saving and life-preserving treatment were complex, a regulatory challenge. It was very intellectually interesting to me. So both of these are incredibly meaningful work. But I sort of took on it as a mission and has attached to myself, myself to some of the sort of initiatives like interoperability and software throughput, specifically in AID systems, and later uh my my spoon shot initiative, when I spoke to parents who told me how little sleep they get after their kids are diagnosed in the first, in the in the first phase. Um and how underserved. I mean, how much space I guess there is for for better technology in this area.

David Kerr 4:01

So you you're gonna have to explain at least to me and probably some of the audience about spoon shots. What what's this spoon theory that uh you have?

Spoon Theory And The Spoon Economy

Shannon Lantzy 4:12

Yeah. So the spoon sh theory is not mine, but spoon theory is is a theory that uh it's hard for someone who does not suffer from a chronic disease to to truly grok, understand what it's like to live with an energy, energy draining um chronic, chronic disease or condition. Um and so Christine Rendino, Ms. Rendino, uh described spoon theory in that she had believe if I'm correctly, she had lupus, and she said to her friend, look, I wake up in the day uh every day with a limited number of spoons, and everything I do has a cost and number of spoons. And she showed, you know, brushing teeth, one spoon, making dinner, three spoons, and she might have to trade seeing a friend out in a in the world for um the next day being in bed to recuperate. And so that spoon economy is her limited amount of energy. And that that is often invisible to folks who don't um experience the same level of medical burden. Um, and AID systems are great, and some of the other systems are great for baseline. Not everybody has available, they're not available to everybody, but that they can do a lot for the engaged um user who has AID systems available to them in their health system. Uh, but there's still a lot of burden. There's a lot of uh management, there's a lot of cost, uh, there's a lot of physical burden. Uh, and I'm I'm not the best person to expand on every kind of burden, but there's a there's a lot more we can do than just A1C in time and range. And so spoons, more energy on the total life is the is the goal. Uh that's how I define the goal.

Spoonshot And Faster Safe Progress

Shannon Lantzy 6:00

And then spoon shot, it's a play on moonshot, but the the thought here was um there are many areas of technology development. They're not completely applied to AID, not least because it's not the biggest, it's not a big tech money maker in comparison to, you know, um Amazon's ring camera or my microphone, even. Um, there so it what if we could bring um the best in cybersecurity and the best in software throughput or in validation engines or the frontier models to the um process of med tech throughput from an unmet medical need to feature development, technology development, all the way to the met need? What if we could accelerate that by bringing in expertise from the outside and calling it a spoon shot? And so that was the concept. And the concept is underway in several in several

Consulting On Regulatory Ready Tools

Shannon Lantzy 6:51

ways.

David Klonoff 6:51

Um What sort of uh role do you have uh in your interactions with FTA? Uh, you're a consultant. Uh what sort of consulting do you do?

Shannon Lantzy 7:03

Um I would say it's like most things, maybe a little not traditional. I have advisory programs where they look a little bit more like retainers, and I advise um technology vendors who are selling into medtech. So, for example, there might be there's a software behavior and understanding tool that is very successful in other industries, um, but medtech would only be able to use it effectively if it's characterized and validated with regulatory alignment. And so I help bridge that gap. Um, my role with FDA is informal. I don't have any contracts with them now. I used to. Um, so I used to be a federal explicitly only an FDA consultant, and my team was explicitly FDA. Um, now my any any interactions with FDA are with my clients who may be seeking a medical device development tool qualification. Um, and the tool qualification can be clinical or non-clinical assessment tools designed to assess the performance of a medical device.

David Klonoff 8:03

Uh you're familiar with AID systems. What do you see are the barriers to having better systems that will be better for patients and not have to use up so many spoons?

Integration Testing And Shipping Updates

Shannon Lantzy 8:15

Uh system integration and trust and predictability of the uh regulatory process. Um there's also scalability. So scale is sort of a business challenge in a regulated environment. You know, if you've got 33 regions to get into and you've got, you know, multiple different versions of either of software and hardware in flight to get into the regions all at once. And in any given region, there's a different type of reimbursement and therefore cost recovery or price, that's a that's a big economic problem. And so if you've got, if you're omnipod, uh you're leading the market, but you need to have that growth engine, um, that's a big barrier. I I like focusing rather though on the inside of the med tech, um, which is the throughput. So for example, the sort of the easy example is we we want to get to one day patches. I want to be able to take a patch fix and ship it all the way to the factory with new firmware in a day. The what the barriers for that is are not making the patch. Sometimes it's deciding whether or not a patch is necessary. It's not really shipping the patch, it's valid testing the patch. So system testing is extremely difficult. Um, system integration within the med tech. So we've got cloud, mobile, controller, um, pump, or CGM. And then of course there's interoperability testing. Um even for established interoperability partners, testing can be pretty complex, difficult, and costly. Does that answer your question?

David Kerr 9:49

Yeah, Sean, and this is really interesting. So you're solid describing healthcare has now become a consumer product.

Shannon Lantzy 9:57

In fact, that's what I pitched last year. Yes. So if if we can if we can ship Angry Birds updates uh quickly, what's the difference in MedTech? Well, the difference is we need to make sure that, you know, if we fail with Angry Birds and we brick it, nobody's gonna get hurt. Some people might be disappointed. Um, but we ha we have to only go as fast as we could fail tolerably. So systems fail. Nothing's quote, safe. But the question is how fast we can go? Can to we can we go to foreseeably fail only tolerably, which means if a system goes down, it tells you it's down and you can get it back up really quickly. Um, or if uh we ship a bad patch, we can reship the better one and never, you know, overdo, you know, over or underdose such that we get to um a high, high or a low low. Um so those are in intolerable failures. I don't think we could get two consumer technology uh speeds as if there was no high consequences, but I do think we could get a lot faster. Um and the incentive for medtech to get that much faster is probably not high enough alone. So Spoonshot is to increase the incentive for um other parties in in including the federal government, um, like ARPA H, to invest in accelerating software throughput for MedTech. I sort of cut you off there. Did I answer your question?

David Kerr 11:28

It it does. But of course, what it's the two letters of the alphabet which are staring at me are AI.

Where AI Fits In MedTech

David Kerr 11:35

And what you're describing here is an open door for AI to come in. Are you a true believer? Are you a somewhat skeptic? And what what's your position on this?

Shannon Lantzy 11:46

I'll I'll say, look, I think it's about the outcomes. So I think if um we can measure, if we can apply AI and improve outcomes for people with diabetes or people using the devices, um, if we can improve outcomes, including speed, cost, quality for medical device companies who are shipping the products. And if we can improve outcomes for clinicians, um, we're doing well. And so I think there are four ways that we can apply AI in this context: AI as a medical device, AI in a medical device, AI to develop a medical device, and AI to evaluate a medical device. And frankly, mostly the AI there is going to be to develop those tools, which then become deterministic and better calibrated to being able to use repeatedly in a validated environment.

Patient Needs And Feature Gaps

David Klonoff 12:43

Uh what do you see as uh major unmet needs for patients given the current level of technology?

Shannon Lantzy 12:52

Well, I believe you would probably be better than me in that, but I think personalized preferences with tolerable acceptance criteria so that people can make individual choices. I have heard that it's relatively rare to have an algorithm tuned for pregnancy use case, or a two-year-old, or someone who doesn't want any alarms at all can't tolerate them. Um, uh someone who wants to be extremely highly engaged but doesn't want to be on, you know, open APS and wants hardware available on the open market, um, there's a gap between feature parity in the open source APS community and Night Scout community and the regulated commercial medical devices. I think when that feature gap closes, um my guess is we'll be a lot more of the way there. And then I'm also assuming that we need devices that are extremely cheap, extremely easy to use, uh fully closed loop with, you know, no interoperability. So, you know, what we're seeing from the uh, I guess pharmacens coming up with uh a single device uh with glucose sensing and insulin dosing, you know, on the on the order of of an omnipod size. Um I think uh I heard put portal, the implantable, so nothing on the body at all. Um I I think everybody has different preferences. Those preferences are extremely valid. And until preferences are met to the point where um someone that has type one has just as much freedom and flexibility as I do, I'm hoping I did a good enough uh job guessing to the answer, but you'd you'd tell me, tell me what I missed.

AI Limits Validation And Risk Tolerance

David Kerr 14:41

No, no, no, that's that's really good. Um just going back to the AI for one second. Do you think AI has reached physician level performance? Or do you think that there's still a long, long way to go, particularly in diabetes care?

Shannon Lantzy 14:56

I I think that it is going to force us to do much more objective characterization of the decisions and tolerances that we have for technology. So um uh FDA people can't just always describe how they make their decisions because every decision is somewhat unique and it's within the context of that 510K or that de novo or that AINN letter. And uh so people don't always have to meta describe, but in order to automate anything, you have to be able to describe and then characterize what good looks like. So, no, I don't think AI can uh do regulatory work or um develop a new feature that's fully validated, but it can do a lot more than people think it can do. Have you played with Fable Five yet? It's unbelievable, absolutely unbelievable what it can do. I I have a fully functioning web app that I've wanted to do for years and years and years that I I almost basically built in one shot. It's a complex card game that is only plays in a very small region that I've never been able to play outside of that region because nobody can learn it quickly enough. And now I have an app to do that. And it was just a kind of a spare time thing. Um, I don't think most people have access to the models I have access to. And I don't think most people are trying because they don't know what they can ask. What we're really gonna see, and I credit Sam Altman with this this concept, but what we're gonna see is that how big, how much they can do is gonna come out of the five to eight year olds right now who can start asking questions. Um famously, NASA had a test for creativity for its for new hires, and it ended up being a generalizable test for creativity. Most adults um couldn't pass it, like some less than 2% scored well on it. Um, but almost all five to eight-year-olds, if I'm remembering correctly, like 98% of five to eight year olds flying colors with creativity. And so we're self-limiting in many ways. Uh, and that self-limiting will start to come away with what AI can do. And uh yeah, so I'm definitely a uh I ride the line between Doomer and Boomer on AI. Um I think organizations are gonna be slower when the risks are high. Uh, and I think we're probably gonna see all the documentation get done by AI pretty quickly because it's verifiable, code works verifiable pretty quickly. And then the next frontier is really getting explicit about validation, um, system integration, which is hard, and tolerance, realistic tolerances. Because we tolerate death. Nobody likes to talk about it. Like even the regulator tolerates death. There's it happens, right? So putting into a model and explicitly saying, we'll accept two deaths in 20,000 is really hard for people to do, but that is the decisions that are like people are being made or that are being made. Once we can tolerate writing down what we really will move forward on, then we can get give AI like more um leash.

David Klonoff 17:56

Shannon, uh, this has been very instructive. You're working in a high-tech world with software, hardware, regulatory companies. Uh, I'll bet your day is quite interesting every day. And uh I'd like to thank you for taking part in this podcast today.

Thanks And Where To Listen

David Klonoff 18:13

The uh Diabetes Technology Report is available at the Diabetes Technology Society website as well as the Apple and Spotify websites. Uh thank you. Thank you, Dr. Kerr. And uh we will see everyone pretty soon at our next podcast. So, for David and myself, goodbye and have a nice day.

Shannon Lantzy 18:34

Thank you, Shannon. Thanks so much for having me.