#871 Best of Juicebox: Standard Deviation and her Friends

First published on Jun 8, 2020. Dexcom's John Welsh M.D. does a deep dive on Standard Deviation, Coefficient of Variation, A1c, Time in Range and more. 

You can always listen to the Juicebox Podcast here but the cool kids use: Apple Podcasts/iOS - Spotify - Amazon Music - Google Play/Android - iHeart Radio -  Radio Public, Amazon Alexa or wherever they get audio.

+ Click for EPISODE TRANSCRIPT

---

DISCLAIMER: This text is the output of AI based transcribing from an audio recording. Although the transcription is largely accurate, in some cases it is incomplete or inaccurate due to inaudible passages or transcription errors and should not be treated as an authoritative record. Nothing that you read here constitutes advice medical or otherwise. Always consult with a healthcare professional before making changes to a healthcare plan.

Scott Benner 0:00
Hello friends and welcome to episode 871 of the Juicebox Podcast

Welcome back to the best of the Juicebox Podcast today we're revisiting episode 343. It originally aired on June 8 2020. And it's with John Welsh, a doctor who goes into a deep dive on standard deviation, coefficient of variation, a one C, and time and range. While you're listening today, please remember that nothing you hear on the Juicebox Podcast should be considered advice, medical or otherwise, always consult a physician before making any changes to your health care plan, or becoming bold with insulin. Are you a US resident who has type one or the caregiver of someone with type one, please go to T one D exchange.org. Forward slash juicebox. Join the registry complete the survey. When you complete that survey. You are helping type one diabetes research to move forward right from your sofa. You also might be helping out yourself and you're supporting the podcast T one D exchange.org. Forward slash juicebox.

This episode of the podcast is sponsored by cozy earth. Now you can get 35% off your entire order at cozy earth.com Just by using the offer code juicebox at checkout, I'm wearing cozy Earth joggers and a sweatshirt right now these joggers are like the best and our sheets are super duper super, super cool. And silky and soft. Also from cozy Earth. Cozy earth.com use the offer code juice box to save 35% The podcast is sponsored today by better help better help is the world's largest therapy service and is 100% online. With better help, you can tap into a network of over 25,000 licensed and experienced therapists who can help you with a wide range of issues. Better help.com forward slash juicebox. To get started, you just answer a few questions about your needs and preferences in therapy. That way BetterHelp can match you with the right therapist from their network. And when you use my link, you'll save 10% On your first month of therapy. You can message your therapist at any time and schedule live sessions when it's convenient for you. Talk to them however you feel comfortable text chat phone or video call. If your therapist isn't the right fit, for any reason at all, you can switch to a new therapist at no additional charge. And the best part for me is that with better help you get the same professionalism and quality you expect from in office therapy. But with a therapist who is custom picked for you, and you're gonna get more scheduling flexibility, and a more affordable price. I myself have just begun using better help. Better help.com forward slash juicebox that's better help h e l p.com. Forward slash juicebox. Save 10% On your first month of therapy. All right, let's talk about John Welsh for a second. John has type one diabetes. He's a physician. And he works at Dexcom. And he's on the show today because I reached out to Dexcom and said, I want to drill down deep. I want to understand granularly the way smart people understand what is standard deviation. And I know that might be like You're like Oh my God. That's what this episode is about. But no, no, listen to me, what we're going to talk about today, standard deviation, we're really going to understand what it is and how they come to those numbers. We're also going to talk about coefficient of variation. Now there's a lot of words you don't know. But by the end of this, you're going to understand. And you're going to understand why it's so important for you living with type one diabetes. After we get all this information into our heads, I started talking to John a little bit about how does he manage what does he call success at the end of the day. And it wasn't as much about the numbers, as you might think. But he really helped me to understand what these words that you know, maybe don't make sense to us right away. Just lay people what they mean, and how they're helping. You know, it used to be all about a one C right? You just tell you tell people like keep your eye one say here, this is what you have to do. But then all of a sudden you start hearing people talk about standard deviation and variability and this is going to help you to understand that even more. I had such a good time talking to John, that it got away from me. I was supposed to talk to him for an hour and like an hour and 20 minutes into it. I was like oh my god, I gotta let you go. He was like four We're minutes away from having to go to another meeting. And I just like, I'm sorry, go, go go. I found this incredibly interesting. I hope you do too. Because I really believe that the concepts that John and I spoke about today are at the core, they're the basis the bedrock of how you should be considering your health with type one diabetes, if you're looking for data to tell you how you're doing. These three things are a huge piece, you'll see. Please remember, while you're listening, that nothing you hear on the Juicebox Podcast should be considered advice, medical or otherwise, please Always consult a physician before making any changes to your health care plan. Becoming bold with insulin. I wanted to call this episode, sugar Adam. But anyway, you'll find out why. Here's my finding. And I've been at this for quite some time, being around the diabetes space, I guess. And when the powers that be whoever they be, decide that we should all be aiming for a lower agency, there's a way to disseminate that information they pull together, you know, industry people, and they give them the toxic here's why no one c should be here and not here. And here's what we've learned. And you know, you get that talk. And then those people find different stakeholders and influencers and they spread the word. And before you know it, when it's distilled out to the public, the message is simply, you know, the ADA decided that your agency should be this now. And that's what you're now going to hear your doctors, your doctors talking about. Like it's, you know, like, it's a rule handed down from my PI, though, suddenly, they have a different opinion. And if you don't pay attention, you don't realize that that's just how we get information out to people, right, there's no good way you can't call everybody in the world and say, Hey, by the way, your agency should be a little lower. Now, you do this. But often, while we're spreading that information, it lacks real context. And when this happened recently, I'm gonna guess in the last two years, when all of a sudden, you started hearing your endocrinologist tell you? Listen, it's really much more about variability, your standard deviation, and they started talking like that. There was no context with it again. And then suddenly, everyone's just, you know, they're walking around, like they learned something. And they say, you know, a one sees not as important a standard deviation, and then all the sudden the message becomes a one sees not important, and then it gets, it gets, you know what I mean? Like it gets ruined as people oversimplify things. And so I really want to leave this talk, just backwards and forwards understanding standard deviation. And when I reached out to Dexcom, I said, I need someone who can really do that, and no pressure, but they said it was you. So

John Welsh M.D. 7:50
I guess you know, if you looked around Dexcom, you would say, All right, we need somebody who can tell stories, who can talk in a straight line more or less. And my, just by way of introduction, I My job title is medical and scientific writer. So I love a good story. And I love especially those stories that have to do with numbers and stories that try to convince people that the truth is actually true. And numbers can really buttress a story, you say, hey, look, look what happens if you don't save for retirement. Here's, here's one way you could go if you spend your money in Las Vegas on that gambling table versus spending your money in an IRA or whatever. So the the idea that you can make convincing arguments with numerical data has always been attractive to me and, and that's why I did some residency training, I went to went to medical school, went to graduate school. And after medical school, I did residency training in laboratory medicine. And laboratory medicine is all about measuring things, and saying, Oh, you've got an abnormal value on one of your lab results. And here's why it matters. And here's what you should do to mitigate the risk of, for example, having a really high potassium level. So if you have good data, then you can make persuasive arguments and you can change people's behavior, hopefully, keep them out of trouble. In the case of a higher low potassium, you could save their life, if you get the doctors to intervene. In the case of some really abnormal lab value that might come up in the hospital context. The bigger question about about glucose values and standard deviation. We can get to that but you made the broader point about public health recommendations and man we are just right in the middle of public health recommendations with with the pandemic because there's there's a lot of uncertainty, which is gosh, you know, how can I go to the concert? Can I go to the restaurant? Can I go outside without wearing a mask and that the recommendations that we've been getting from public health authorities have been A little bit discombobulated maybe internally inconsistent and kind of frustrating at times. But I am with you though the idea that we can provide good evidence based recommendations with respect to goals in managing diabetes is, is a big interest of mine. I'm all about all about the numbers.

Scott Benner 10:20
Well, many, many years ago, I came to the conclusion for my daughter, that if I get what I expect is what I started thinking of it as I realized I had Arden's high line set at 200. And I always kept her under 200. So one day, I moved her to 180. And I was like, Oh, I always keep her under one ad. This is really interesting. So I kept pushing it down and pushing it down. And now my daughter's, you know, ranges 65 to 120. And mostly, we keep it in there. And when we don't, it doesn't go that far out. Right, I'm gonna go to 150. That's usually, you know, like, just now, I will use this morning as an example, two slices of toast, an avocado, butter, and an orange. And her blood sugar went to 148. And it's coming back now. And it's not over a longer yet. Beautiful. Right? And so, but her standard deviation will look bigger than someone else's. And I don't know if I'm making up things in my head, or, like, how is it possible that Arden can have a life like that, but her standard deviation could be higher than someone who's a one sees a point or two bigger than hers, and who have swings that are far higher and lasts longer. And so that's the one idea that keeps me focused on I don't understand standard deviation or not. And then when I start talking about it with the people that I that listen to the show, I come to realize that everyone's sort of got that, that confusion. So can we start very over simply. and standard deviation as an idea? Is a mathematical issue. Is that right?

John Welsh M.D. 12:01
Oh, it is it's it's a number that is used to describe a set of numbers. So for the case of folks who are using CGM, you might expect up to 288 numbers every day. And each number represents a glucose concentration. And you can use words to describe that set of numbers or you can use numbers to describe that set of numbers. The the average is a pretty simple number that it's easy to calculate, you would add up those 288 values and then divide by 288. And then you get the mean, in this case, it's the arithmetic mean. There's other flavors, there's the geometric and the harmonic mean. But we'll we'll leave those aside for now. But the arithmetic mean, tells you it's a measure of central tendency, where you might expect the average, if there is such a thing, an average value to fall. The standard deviation is is another number that's used to describe that set of numbers. And it describes the width of that distribution. So it gives you an idea of how surprised should you be when a number shows up, which is pretty far away from the main. So here's I've got a kind of wonderful document came out a couple years ago that looked at glucose concentrations in people without diabetes. And they they came out with normal values. And the normal value here for glucose was pretty close to where is it 99. And express this number 99 is the average and then they give you a plus and minus seven. That plus or minus seven refers to the standard deviation. And the standard deviation. If you imagine a bell curve that you might have seen in school, where the most popular value is right there in the middle, that's the mean value, in this case, 99. The plus or minus seven tells you how steep is the drop off on either side of that mean value. So in this case, the 99 plus or minus seven, if you were to go up to 106. In other words to the mean plus one standard deviation, you would expect to have about I'm sorry, let's go back and say 99 plus or minus 799 minus seven is 9299 plus seven is 106. So anywhere from 92 to 106. The expectation is that you would have two thirds of the values in that pretty narrow range. So if your goal is to have if your goal is to have quite a lot of stability, which in general is a good thing. You want that standard deviation to be low and normal people without diabetes, it is in fact quite low. 99 plus or minus seven is a very tight distribution. Two thirds of the values fall between 92 and 106. Okay, so Whether there's a calculation, we could walk through it if you want,

Scott Benner 15:02
please. Yeah, I was just going to tell you that when we're done. And I can say this because this won't go out until after I'm allowed to, but I'm wearing a Dexcom. Pro. I have been for a couple of days. Ah, so I can see, I'll be able to look while you're talking and figure out what mine is.

John Welsh M.D. 15:20
Oh, good. So are you able to see the real time data or not yet? No, I

Scott Benner 15:25
see it. It's not blinded. I'm looking at it on my phone.

John Welsh M.D. 15:28
Oh, okay. Well, I hope you're, I hope you're within seven points of 99. I hope you're well in the normal range.

Scott Benner 15:34
I certainly hope so too. But I am I, I was really, I have to be honest. As I put it on, I thought, I'm doing this so that I can see how a working pancreas attacks things brings them back what curves look like, I wanted to see all that because I thought it would make it easier for me to speak to people about about using insulin. But at the last second as I was about to do it, I thought am I about to find out I have like type two diabetes or pre diabetic or something like that as like maybe you know, and I just kind of was like, alright, well, if that's if that's the case, it's the case, I'm going to find out. But so far, so?

John Welsh M.D. 16:15
Well, I hope so. And when we do onboarding, we have people come work for Dexcom. And part of the onboarding process is, hey, look at, look at our product and look at what it does. And of course, it's voluntary, but we say all right, if you'd like to wear one of these, just to know what the experience is, like, we can get you set up with one of these. And our expectation is always your glucose values are going to be are going to be let me check boring. And you're going to have a really smooth ride throughout the day. You know, 99 plus or minus seven. But once once in a while we have we have people that come back and they say, you know, John, I learned something really interesting. And what's that? If I have if I have an entire pizza, I can get my sugar up to 180. And I say wow, that's, that's abnormal. And so people learn something, even if they don't have a known diabetes, they can learn something about diet and exercise that you know, I went for a long bike ride yesterday and I crashed I went pretty low. And then I had the the Coca Cola or the sugary drink. And then I saw my sugar zoom back up so you can learn a lot. And that's a general truism that you can learn a lot just by looking. But Scott, I'm pleased that you're wearing one of the CGM sensors and I hope you learned something I really

Scott Benner 17:39
am. I'll tell you already, I had two pieces, smaller pieces of homemade pizza on Sunday. And three and a half hours later, I got a push up from the protein and the fat probably holding the the crust of the pizza in my in my system longer. That was fascinating. And this morning, I had a breakfast that was just a piece of Turkey and toast. People are like oh my god so boring. But, but I smoked a turkey yesterday, it was so good. John, I want to have some sort of breakfast. So I took some turkey and I had a piece of toast this morning. And when I was done, I grabbed a navel orange. And when I ate the orange It tried really hard to push my blood sugar up. You know, not immediately but it was it was drastic, and my body attacked the drastic rise so much so that I was 74 straight down for a second before I leveled right back out at 80 It was amazing. I went from 74 straight down to 80 and stable in a fight in all my shin one five SEC five minute things. So I saw my body go oh, that's a lot of sugar from that orange. And you know, he's already put this bread in here, I guess you know, I don't obviously don't know exactly how my body's thinking but but the idea was I was I was starting to push up a little from the bread not greatly. But then I think when I added the the simple sugar, I just I got a really quick response. So I'm noticing that that every time I press with simple sugar, my body comes back more aggressively than it does with more complex carbs.

John Welsh M.D. 19:03
You know, boy, that's interesting and, and other people have described it to me where they'll, they might have some indiscretion, they'll say I'm gonna have a 24 ounce Mountain Dew and you slam the sugary beverage and you get this wonderful increase in sugar which you can feel in life is wonderful. And then what you described with the orange happens happens in a very dramatic way where they're the insulin kicks in and then the sugar plummets and then all of a sudden you have the the big crash after the sugar high comes the crash and that I think that's a manifestation of instability. And same thing. I'm going to make a quick little analogy to the cruise control on your on your car. What I hoped for when I engage the cruise control on my car is just a smooth ride. And and I don't want the car to be slamming on the throttle and slamming on the brake all the time. You I just want to be going at 65. All the way home. So I am very sympathetic to your experience with with high amplitude glycemic swings. It's it's a common thing, especially in the world of type one diabetes where we're all taking insulin.

Scott Benner 20:17
Yeah, it's it's very interesting. I'll tell you and I'll then I'm gonna let you get back to it. But the other thing that happened that I really didn't expect, but makes total sense, is that for about the first 36 hours, I wore it, every time I looked and saw my blood sugar stable, I had a horrible feeling of guilt. It was, it was really interesting, because my daughter has had type one since she was two, she's 15. Now I have interactions with 10s of 1000s of people who have diabetes, and they all would just, I don't, they would do anything to have that, you know. And it really, it really impacted me for in the beginning, I just was I felt very guilty for my pancreas working. It was a weird feeling. So, but I'm sorry, I shouldn't derail you, because we're talking about something that's, you know, you don't think it's complicated, but trust me, I do. So I shouldn't I shouldn't distract myself. But we were talking again, about about people, you know, who have a functioning pancreas. And you said, you know, let's pick 99 Is that is that that kind of center target? And you can go to 92 or up to 106? And then explain again, what I'm sorry, where were you headed with that?

John Welsh M.D. 21:23
Oh, sure. The value, I'm looking at a big article that came out a couple years ago, they looked at 153 People without without diabetes. And they put glucose monitors on him. And they they collected a bunch of data. And so the question, I guess the first question is, why would you care? Why would anybody bother? The answer is, well, we want to know what normal looks like. So we can decide if if a particular glucose profile is reassuringly normal, or if there's something going sideways on it. The 99 value from earlier is the mean, the standard deviation I gave you earlier is seven. And that tells you something about how wide the distribution is. So one standard deviation on either side of 99 would go from 92 on the low side up to one 106. on the high side, that mean plus or minus one standard deviation, the expectation is that two thirds of the values would fall in that relatively narrow range, two standard deviations 99 plus 14 is 114 113. on the high side, and then 99 minus 14, I guess is 85. Is that right? On the low side, so 85 to 113, the expectation is that you would cover an even higher percentage, I think 96% of the values would would fall in that range. And if you go out even further to plus or minus three standard deviations, the expectation is that almost all the values more than 99% of the values would fall within three standard deviations of that central value the mean. So that's, that's it in a nutshell, the calculation. It's not difficult, it's not trivial, but it's not difficult. I'm not sure if your audience would be interested in walking through it or just looking it up.

Scott Benner 23:19
Right now, John, this is very much meant to be for people who are interested in that. So I have a group of episodes, there's about 20 of them. They're called protests and they are deep dives into specific things about type one. And this is this is one so don't think of this as an interview as much as think of it is, we are really trying to pick this apart so that when someone listens through like, I'll be honest with you. In sixth grade, my guidance counselor told me I could take algebra halfway through algebra, I didn't understand algebra at all. And I thought, oh, my gosh, I'm terrible at math, I dropped out of it. A was a bad decision, because I followed a much simpler math track the rest of my time, which probably wasn't necessary. And just now, as you were talking, I, you know, you set up this scenario, and the standard deviation was plus or minus seven, and you started talking about out one, standard deviation two and three, and it just started to make sense to me. So you're doing a good job. Trust me if I understood what you just said, everyone listening has a chance to understand it as well.

John Welsh M.D. 24:20
Well, you're you're very kind and that's I'm very pleased to think that we're making progress toward the goal, then we can I can introduce the topic again and say the standard deviation is just a number that's used to describe a set of other numbers. The standard deviation, there's a calculation for it, it's a little bit involved, but involves, first of all calculating the mean for a population. The example that we used was the the mean value for people without diabetes, it's 99. You have quite a lot of values. You might have 1000s or 10s of 1000s of values. And this is where it gets a little bit tedious. For every one of those individual values in the set that you want to describe, you have to calculate the difference from the mean. And the difference from the mean is either going to be a negative number, or it's going to be a positive number, depending on whether the the individual value is higher or lower than the mean. You square that. So squaring a negative number, it gives you a positive number, squaring a positive number gives you a positive number. So you're going to get another set of numbers, which is the squared difference from the mean. And if you had 10,000 values in the set, you're going to have 10,000 squared differences from the mean, you have to add them all up, you get a sum of squared differences. And then you divide it by divided by the number of observations in the set minus one. So it's, it's a pretty complicated when you try to describe it verbally. But if you were to look at it on a sheet of paper, you would say, oh, it's, it's a series of steps. Add up all the squared differences from the mean, divided by a large number one less than the number of observations in your sample, and then take the square root. And then once you've taken the square root, bingo, there's your standard deviation. So it's, it's a few steps, but it's something that kids probably learned and then probably forget just as quickly as they learned it in, in middle school or high school algebra class.

Scott Benner 26:26
So how does clarity app like to simplify that all down? What is the clarity app looking at? When it tells me, you know, the, the standard deviation is 35? Can you like, distill it? What is it looking at to make that decision without the without the detail?

John Welsh M.D. 26:44
Oh, absolutely. So the statistics page, for the clarity app gives you some summary statistics. And just a quick little operational note, I wonder if you're able to see my page that I'm trying to share with you on the Zoom meeting? Yep. Oh, good. Okay. So maybe you should ask your question again. So we could rejoin the the post editing narrative?

Scott Benner 27:11
Oh, I just know, I was. What I'm worried. What I'm interested in is, is there's a clarity app, obviously. And it tells me, Oh, your standard deviation, or your daughter standard, if she is 35. Or some people are like, Oh, I'm struggling. And you know, my mind is 65. And I heard from a woman the other day that told me her doctor told her that anything under 100 was okay, which she very smartly was like, I don't think that sounds right. But I want to know, like, what does it look at? To tell me? My standard deviation is 34. Like, taking into account?

John Welsh M.D. 27:47
Oh, sure. Well, that's, I think I can get that one answered pretty quickly. We've got our statistics page. And if your audience wants to look at the Dexcom, clarity, web interface, there's a page all devoted to statistics. Looking right now, at my statistics for Monday, and this is every Monday for the past 30 days. So there's several Monday's in that sample, I've got a total of 1253 readings. And each one of those is estimated glucose value. And then the summary statistics, the minimum 40 Oh, that was scary, the maximum 244. So those are, those are not normal, the mean value 128. That's reassuring, and then the standard deviation 34. So to get that 34, the calculation that I just walked you through, which is look at every one of those 12 153 values, get the difference from the mean. So do the subtraction 128 Minus a particular value. You square each of those differences from the mean, add them all up, and then divide the total by 1252. And once you've done that, you take the square root of it, and it's it's 34. So there's, as I said, it's a little bit of algebra. But it's, again, the usefulness of it. 128 plus or minus 34, tells you that you would expect two thirds of those glucose readings to be within one standard deviation of the mean. So 128 minus 34 is just 90 something and then 128 plus 34 is 162. So you would you would expect most of my sugars to be in that in that range.

Scott Benner 29:41
Take for second example, I know we're going to oversimplify but describe what mean Yes.

John Welsh M.D. 29:51
Oh, sure. I mean, it's also known as the average value. So if you were to look at the NBA players As you say, Wow, NBA players are really tall. You might express that in numbers by saying the average or the mean, height of an NBA player is six feet six inches tall. So it's another word for average, it's a particular kind of average. But we don't need to talk about the other kinds of averages. Mean is usually just the arithmetic mean, you calculate it by adding up all the values, and then dividing that total by the number of values.

Scott Benner 30:31
So what I have here, what I'm looking at in front of me is 12 153 readings. There were 40 that were or is that under a certain number, those 40?

John Welsh M.D. 30:45
Oh, yeah, we're looking at these rows in the in the statistics, the number of readings, 1253 is a bottom, the minimum was 40. The maximum 244. And the mean value 128.

Scott Benner 30:59
Within within those 12 153 readings, there, the high was 244. The low was 40. But on average, this person's blood sugar was 128.

John Welsh M.D. 31:12
That's a that's a nice way to do it. And yeah, we're looking at, we're looking at my readings from the past month or so

Scott Benner 31:17
these are you Oh, my gosh, are you? Do you have type one?

John Welsh M.D. 31:21
I do. I've been living with type one for most of my life for past 45 years. And so far, so good.

Scott Benner 31:28
Show me like an example page. I didn't realize we were looking at your blood sugar. Well,

John Welsh M.D. 31:33
I yeah, you can spy on me. You can you can look at my summary statistics. Here we can we can continue with the summary statistics page. Yeah.

Scott Benner 31:43
And I'm gonna have some questions about it when you're done. But please keep keep going.

John Welsh M.D. 31:47
Oh, sure. And this is an incredibly number, it's a very useful way to get a numerical description of other numbers. And so far, so good. You know, here's, here's a guy, John Walsh, who is this clown anyway, and what is he doing talking about his glucose numbers. So John's, had a, at least one time where he went all the way down to 40. But the main value 128 is reassuring. And then we get down to some other statistics that talk about the median value, the median value is the value above, above which and below which half of the values occurred. So in my case, the median is 122. And that tells you that half of my readings were above 122, and half of my readings were below 122. So that's another measure of central tendency. The end, it's usually expressed alongside the interquartile range. And so you look at the, the value that is 75% of the way to the top, so 75% of the values are below at 25% or above it. And in my case, the the 75th percentile is 153. The 25th percentile is 103. So you can say with, with some confidence that half of my values were between 103 and 153. And those are the 25th and 75th percentiles, and the the interquartile range here has given us 50. And that's just the difference between 153 and 103.

Scott Benner 33:33
So the question here, if if Yeah, if if half of those range between 103 and 153. I'm assuming that the other half are how we arrive at the standard deviation of 34? Like, I'm assuming you need that information to to come back to the standard deviation?

John Welsh M.D. 33:49
Oh, no, no, the standard deviation, the standard deviation relies on all values. And it doesn't, it doesn't care so much about the distribution, it just cares about how far from the mean value the values are. So there's, there's there's another point that I want to make, which is the median value, in my case, 122. The mean value is 128. A lot of times those are very close together. But sometimes they're very far apart. And there's some special circumstances where the mean value is much, much different than the median value. And we can talk about those if you think it's interesting.

Scott Benner 34:32
I wonder what I do want to know is, is how much of sensor like so you know, I've my daughter has been wearing a Dexcom since seven, maybe Dexcom, seven or seven plus back then. And so, obviously, we see things at every generation, improve and improve and improve but I could still say that for Arden in the first number of hours. You know that you put on a new sensor it's not as I don't know, it's not as tight with its understanding of your blood sugars that maybe is on, you know, day two or like, you know, or there's a sweet spot through the middle where it's crazy. Arden uses a Contour Next One blood glucose meter, which is incredibly accurate. And for a large part of our sensor where the meter and the CGM are spot on with each other there within a couple of points. And when you're managing type one, there's a ton of like, good feeling about that, knowing that, you know, she wakes up in the morning, and it says her blood sugar is 96. Now whether or not her blood sugar is really 85, or it's really, you know, I don't know, 104 to me is of no real consequence. It's in that space. And I'm thrilled with that. Then I put it on, and I don't have diabetes. And I wake up and it says my blood sugar's 94. And I think, Oh, my God, I've been fasting all night. And I'm 94 and I do a finger stick. And I'm 85 It's amazing that those seven points to a person without diabetes is, it's a different impact than it is to a person. Right? And so it is seriously like, I wake up in the morning, 94 I'm like, Oh, I guess that's it, I'll just eat lettuce till I die. But you know, like, like, it's just, it feels like that immediately. And, but I take that same information coming from my daughter, I am completely comforted by it, not just comforted by it. But it leads me in my understanding of how to manage her insulin and her health and everything. My question is, is that knowing that the sensor is a little, you know, on the on the edges, it struggles a tiny bit more than it does in the middle? Is there something about my data that I can't look at to micro? Like, do I have like, how much time do I really need before? The inconsistencies in the data? And the consistencies in the data bounce out to where it doesn't matter that it's not all? Perfect? Does that make sense?

John Welsh M.D. 36:53
Oh, that's, yeah, that is a very common question. And I don't have I don't have a good answer, I can tell you how I deal with imprecise measurements in my own life. And, and I've got, I had a wonderful bike ride yesterday, here in San Diego, and I've got a fancy bike that has a built in speedometer, it's based on how many how many times the will completes a revolution. So there's a speed sensor built into the into the wheel. And based on that, you can calculate your speed. And I've got another fancy thing in my phone where you can get your speed based on satellite data from your global positioning satellite system. And and I looked at it and I found myself chugging along the road and and the the speeds, you want to guess if they were exactly the same. No, they weren't. I was going 20 miles an hour. If you look at the wheel sensor, I was going 21 miles an hour, if you look at the GPS coordinate, so measuring your blood sugar and seeing one number and then looking at your CGM and seeing another number. And and it's frustrating, because there's no good way to to know how excited or how concerned to be about discrepancies. There's always going to be discrepancies. It's a rare thing when when the blood sugar tells you you're 105. And then you get that 105 From the CGM. And I don't want to give medical advice over the phone like this. But there is the possibility that you could calibrate your your G six and based on the your confidence in a blood glucose meeting reading, you could say, oh, my GSX is reading a little bit low. I'm going to calibrate it, and then bring it back into better alignment with the with the blood glucose meter. So I know it's frustrating. I wish I had a better. I wish we had better devices for measuring glucose with even more precision.

Scott Benner 38:59
They're amazing. You've had diabetes forever. You know how amazing this stuff is. Just because you work there doesn't mean you can't say that. And it's actually been very interesting for me because of the pro doesn't allow you to calibrate or at least I just had to go with it. And it really sure it was it was it was interesting to live in the space because for my first maybe 18 hours, the glucose monitor was reading about 10 to 12 points higher than what the finger stick was was pretty consistent for those few hours. And I found myself thinking if this was my daughter, and I put a brand new CGM on her that thought she was 110 when she was 91. I'd be like, Oh my god, this is the most amazing thing ever. I love this thing. It's so amazing. Except you know, and I didn't have diabetes and I was like, Is my pancreas not working? You know, like it's very like it's a it was just such a very different thing. But beyond that initial feeling. It really did just cement my idea of how much I love this technology. And and because I can remember managing my daughter's blood sugar without a glucose monitor. And to think that she'd be stable at 110 or 91, ever for hours and hours at a time is insane, but it just never happened. But over these last few days, we've been eating the same meals. And her care is so dialed in, due to a large due in large part to the information that comes back from the Dexcom that her blood sugars and mine are largely matching before and after meals.

John Welsh M.D. 40:35
Congratulations. And that's just That's wonderful news. And, you know, it's, and I'm totally with you, we we can talk about the battle days when when you had to make a make a guest and a lot of times it was not a very good guess based on just a urine dipstick and you could say, oh, I'm spilling sugar into my urine and I need more insulin, and you would have to make a guess. And some of the highs and lows were pretty scary. And, and people you know, sad, sad to say that people are still dying from insulin overdoses, insulin, let me check, it's a poison, and it can kill you. And there's, there's a lot of downside risk to insulin, even though it's a huge blessing, we're coming up on the 100 year anniversary of the commercialization of insulin. So we're all going to celebrate and be thankful for the commercialization of insulin and the fact that we're not dead. But it's, it's a tough disease. And you wouldn't, you wouldn't wish it on anybody because it's really a lifetime burden. But I'm really pleased.

Scott Benner 41:43
I just had a conversation briefly online with a woman this morning, who even with all the technology gets incredibly low every day. So I was turning her on to the podcast as like, this doesn't need to be you're just you're not using your insulin correctly. And it's not that it's not that difficult to figure out how you know, so I turned around, I was like, Listen, I have an idea. Can I hit you with some questions and see if you have answers to them. These are questions that came from listeners. And sure, I'm not asking you now I understand you're a doctor. But I'm not asking you that way. I'm asking you based on this information, this data and how much you've seen it? Do you see? Do you see information in the data that would help people with the things that they're concerned about? So the first one simple? Do you know what a non type one standard deviation usually is? Is there a range where it usually falls?

John Welsh M.D. 42:34
For example, somebody with type two?

Scott Benner 42:36
No, no, no, just someone who doesn't have diabetes at all. Do you know where like, like, where? Oh, yeah.

John Welsh M.D. 42:42
Yeah, so we've got a we've got some data from a big study of 153 people without diabetes. Their standard deviation was was seven,

Scott Benner 42:55
seven. Okay. Okay, is there? Let's see how I want to say this here. So this is a type one question somebody is somebody's asking. If there's a lot of variability within the good range, say like, like 70 to 120, this person's kind of bouncing between 70 and 120. There what they want to know, for their health? And maybe you don't know, but would they be better off sitting at 120 than they would be from going up and down between 70 and 120?

John Welsh M.D. 43:27
Oh, I think so. And there's, this kind of leads into another number that you can get with the, the summary sheet, it's the ambulatory glucose profile is something that Dexcom has. It's, it's not exclusive to Dexcom, but it's called the AGP. The ambulatory glucose profile, what

Scott Benner 43:46
my things John, don't know, you really got to get creative in charge of in medical in general in charge of the stuff that that goes back and touches people. If you look at glucose for I'm sorry.

John Welsh M.D. 44:01
There's, there's a lot of syllables there. And there's a whole industry for you know, if you come up with a new drug, you have to hire a marketing firm to come up with a name for your for your new drug. But there's a digression for you. Anyways, is the numbers. The numbers that are on the top line of the ambulatory glucose profile, the average is there, the time and ranges there. There's another number here, which is the standard deviation, and then the coefficient of variation. And that's a number that I think has has a lot of usefulness because it tells you how big is your standard deviation compared to the mean value. And there's some clinical implications for that as high, high coefficient of variation is dangerous because it puts you at very much increased risk for dangerously low events for for hypoglycemic misadventures. So the the coefficient of very Question again looking at my own data for the past 30 days, my coefficient of variation 31.3. And is that good or bad or indifferent? It's, it's higher than I'd like it. But is it dangerous? And there was a fun article. Fun, I don't know, but useful anyways, the useful article came out a couple years ago, and some folks in France in the UK came out with an article in diabetes care. And they they said, CV coefficient of variation of 36% is the threshold to distinguish between stable and unstable sugars. Because beyond this limit, the frequency of hypoglycemia is significantly increased. And, and if this, my own CV here 31.3%, that's reassuring, it's low, which is good. And it's less than 36%, which tells you that I'm, I could still go low. But the fact that this CV is less than 36% is reassuring. I went to see my endocrinologist and he said, Hey, John, keep up the good work. You're probably not going to die of hypoglycemia before the next time I see you. And I was so alright. Yeah.

Scott Benner 46:15
John, you know, it's interesting that I see with my daughter who is, you know, a woman, a burgeoning woman, is that with our care, the same exact care we use on weeks and days where she's not impacted by hormones? Arden's standard deviation is 24 ish. But oh, my gosh, that's terrific, thank you. But that's not why I'm telling you that what I'm telling you that is because although I appreciate it, why I'm telling you is because that when she is impacted with hormones, the run up to her period, for example, her deviation jumps up to 45. and N are no holes aren't different, her meals don't vary. It just, she needs more insulin. And it sometimes takes a couple of days for you to realize that that's happening. And then once it's happening to remember, it's happening to remember, like, you know, oh, you know, my ratios are telling me this much insulin, but it's four days before I'm gonna get my period. So it needs to be more, it's difficult to recall all that, you know, constantly. But it's fantastic. It's interestingly fantastic to see because if Artem was a boy, I think I would have a son with a with a standard deviation, pretty consistently within 24. Until they hit I'm assuming puberty as well. But you as a, it's just very interesting to look at your 30 day chart here. You're I know we're talking about so you don't mind, but your standard deviations 42. And you're saying it's not where you want it, but it's also not terrible, like people are trying to understand on the outside, what's the number that keeps them healthy? And what's the number where they think, you know, something else is going to happen? It is very simple in people's minds when they think about these numbers, like what am I gonna hit? How do I get to it?

John Welsh M.D. 48:03
Oh, yeah, yeah, and I think if the the more useful number and I think the one that is very convenient to have as a as a goal, and is is the coefficient of variation. And that's just a ratio, it's the standard deviation divided by the mean. And aiming for something less than 36% would be would be a reasonable would be a terrific goal. And if I were still seeing patients, I would say, Here's your, your coefficient of variation is 40%. Let's look more carefully at the trajectories or the, this is called a modal day plot. And I'm sure your audiences has seen this, it lays out the clock time here on the bottom axis, and then the glucose values on the vertical axis. And you can see the median value here and the bold line right in the middle. And then you can see the shading here, the blue shaded area covers 50% of the values and then the area in between the dotted lines covers 90%, or I'm sorry, 80% of the values. So what what I'm looking for what I wouldn't be looking for if I were looking at somebody else's plot is a smooth ride. And sometimes you can identify parts of the day where the ride is pretty bumpy. For example, after lunch, if you're having lunch at your desk and you're not going for a walk and you're having the third slice of pizza, you might see spikes after lunch or dinner. Or you might see plummeting lows after breakfast if you gave yourself too much insulin for breakfast, and fun to go with breakfast. So I'm not the standard deviation. If you're always cruising around a relatively high number like 170 The standard deviation is going to be bigger than if you're always cruising around at a much lower number like 100 And so, um, the number that I think is more reasonable to target as a therapeutic goal is the coefficient of variation.

Scott Benner 50:09
Okay? Under 36.

John Welsh M.D. 50:13
Yeah, that's, that seems to be the magic number. And that's the consensus and, and it's, it should be achievable if you if just pay attention to parts of the day where you might be having a bumpy ride, you can look at your behaviors, look at your response to your behaviors and say, You know what, I think I will, instead of having three slices of pizza, maybe I'll just have one. So CGM can be a wonderful motivator. It can inform people it can motivate and reward good choices. So I'm you can tell I'm a huge fan. I love evangelizing this stuff, but you can learn from, you can really learn a lot from the numbers. And the numbers can tell you, if you pay attention to him, to the numbers themselves, and also to the summary statistics, like the standard deviation, you can learn quite a lot from him.

Scott Benner 51:03
I'm a huge fan, I don't understand that, obviously, nearly as well as you do, but I know what it tells me. So for instance, after Ardennes, my, my poor daughter, one day is going to listen back to this and be like how much did they talk about my period on that podcast, but after so the lead up to her period, there's like three or four days prior to it, she gets, you know, all of a sudden, she needs way more insulin. And then in the first day or two of it, it happens still, but then there's a moment where it levels like whatever happens is done. She's still the periods still happening, but the hormonal impact seems to be going out of her body. So let me give you an example. Because it just happened yesterday for the last 24 hours. Arden's estimated a one, C is five, and our standard deviation is 24. Per average, blood sugar's 98. But if I just go back seven days, through her, you know, through this lead up to this period, estimated a one C 5.8, standard deviation 43, average blood sugar 119. It's an it's just the hormones, it's the lead up to her period. And so it's fascinating and not that you don't know but and then there's another time of the month where it happens again to her for four or five days. But just those just that week, and then that other chunk. So basically what I think is about 789, probably 12 or 13 days of the month, takes what would normally be I think, an SD and like I said in the mid 20s and an A one seat closer to five than six, and it moves her agency more towards like Hurray, once he pretty much sticks at like 5.6 it doesn't move very much. Okay, it's just very, I don't know, like I don't know what I would do before this information like no lie prior to it. I wasn't a different person. And we were not good at this at all. You just diabetes in general her hurry once these were in the eights and I finally got them into the sevens just by having, you know, better tools and insulin pump and a glucose monitor. But I still didn't understand that enough to turn it into real, like success, you know, like, like the idea of knowing when to Bolus and that sort of thing. But I know all that from this data now. And it's sure incredibly beneficial.

John Welsh M.D. 53:27
Absolutely. Well, I'm, I'm with you 100% on that. And I think for my own my own experience was in the bad old days before CGM, I was poking my finger and making a lot of guesses. And it really got me interested in how the body works. And it was a great, great motivator all through college. And that was part of my story when I was applying to medical school and I'm not alone. There's a lot of a lot of physicians who specialize in in Endocrinology and Metabolism who also have type one diabetes. So my own story is, is hey, this is really interesting. I want to learn about it. And I want to go to medical school and what do you know, the medical school here in town said all right. All right. Coming to medical school, and you can learn you can learn quite a lot in medical school about about the disease itself and about how you measure how you measure sugar and measure all the other important things that we care about in metabolism. So it's for me anyway, it was not just a life changing event when I got that diagnosis but it also sort of defined my career path toward a toward becoming a physician and also to to working here at Dexcom

Scott Benner 54:40
Yeah, so that's fascinating and I'm afraid I'm gonna start talking to you and then lose track of what we're supposed to be doing because questions I almost answered ask them and I was like, No, don't do that. What cut when you when you when this data is pulled together, given that there are you know, Blood Sugar legs and meters aren't perfect and nothing's perfect. What? What's built in to deal with the error? Like, how does it come to the number and? And take the the imprecise pneus out of it? Is it like, like looking at yours? For example, your standard deviations? 42? What if if if a Dexcom was absolutely perfect if there was a you know, if it wasn't technology, but it was it was your, you know, I don't know, something organic that could know 100% For sure. What all these measurements are on your glucose all the time? How far off? Do you think that number would be? If you had perfection? Does that make sense?

John Welsh M.D. 55:41
Oh, yeah. Yeah. You're You're hypothesizing that there is some there's no real answer. Yeah, there is. There does exist some true number. And we're always trying to become more more accurate and getting closer to that true number. We are, we're never going to get there. You have to stipulate that we're always going to have some, some wiggle and some imprecision. And that's, I think true. Because nothing on this planet is perfect. And we have to, if we get to heaven, and then everything is perfect in heaven, if we ever make it there.

Scott Benner 56:17
That'd be my first question. When I get there, I'll be like, what was my kids? Really?

John Welsh M.D. 56:24
Yeah, so that's a whole nother line of inquiry. But we're probably certainly within 10%, I think I'm confident that we're within 10%, I'm less confident that we're within 5%, I wouldn't be surprised if we were within 3%. And I would be really astonished. If you told me it was within 1%, I would be astonished. So I've got some confidence, the for the 10% precision. And I've got some optimism that we can usually get within 5% of the true value. Those are just speculative numbers. Because there's no such thing as a perfect value, even if even if you use the gold standard. We could quibble about any reference instrument. And this is one of the things they drilled into us during my residency training in laboratory medicine, which is, is there such a thing as a perfect measurement? No, not until we all die and go to heaven. While we're living on this earth, you have to deal with imprecision and uncertainty. But I think we're pretty good. And just for purposes that we care about managing managing diabetes and living a long happy life, I think we're we're well within the realm of of good enough.

Scott Benner 57:40
And outcomes are good based on what we noticed. Does that mean, from what you just said, if at a 42 standard deviation? Is it possible that your standard deviation is somewhere like 36? Or possibly like, I don't know, 48 or 47? Or is it more likely it's lower? Or more likely, it's higher? If it's Is there a likelihood that it's more one way than the other?

John Welsh M.D. 58:05
Oh, yeah, the standard deviation just tells you how, how spread out the distribution is. And the the true standard deviation could be higher or lower? Because all the numbers that the standard deviation depends on could actually be incorrect. So I think, yeah, that's a tough one. Let me let me think about that. Yeah. I'm looking now at this. Looking now at the standard deviation and this famous bell curve, the you know, what the, if I'm understanding your, your question correctly, could the standard deviation be something different?

Scott Benner 58:51
You use me as an example, in my situation, right. Now, if I put on a new CGM, every 10 days, I wear three sensors a month, nine sensors over a three month period, if I look back at my 90 days, my standard deviation, if my if my sensors reading just 10 points higher for the first, I don't know, just say 36 hours of every one of those things. Am I more likely to look higher than I am? Or lower than I am? Because of that? Higher right?

John Welsh M.D. 59:19
Oh, yeah, I think I think you would have a high. It's called a high bias. But your earlier question, could the standard deviation be something other than the calculated result? I? I think the answer is no. If if you give me the numbers from one to five, could the total be something other than 15? And I would say no, the total of the integers from one to five is 15. And if you give me a set of numbers, I can calculate the mean and the standard deviation. So I think the calculation that we've done here, resulting in this standard deviation of four 32 If we did the math correctly, then the standard deviation is 42

Scott Benner 1:00:04
is the I'm sorry, there's the algorithm that's making this decision. Does it scrub anything? Like, you know, like a compression load? Does it see that and go, we're not going to take this into account, does it do any of that kind of stuff?

John Welsh M.D. 1:00:18
Oh, yeah. And that's, I think that's true. That's got to be true for Medtronic, it's got to be true for Abbott, it's got to be true for sensing Onyx. And also for Dexcom, we've got, we've got algorithms, the signal that we are measuring is actually a voltage. It's a, it's so I'm sorry, it's current. So the current is very low. Current, usually measured in amperes. And we're dealing with billions of an ampere, I think, nano ampere, or Pico amperes. So incredibly small currents. And the challenge for the engineers is to take that very small electrical current, and translate that into a number that makes sense and number of milligrams per DL. So that requires some, some engineering talent. And it requires an algorithm. And I think that's part of the secret sauce that we have here at Dexcom. Medtronic, I'm sure they have a algorithm, which is similar, but slightly different. And the same for Abbott. And the same for sensing Onyx. And that's true. Whenever you're measuring something and saying what you're measuring, you know, for the example of your oven, if you're cooking, you're making your cookies, you're measuring temperature, what you're really measuring is the height of the mercury in the thermometer. And the trust is that that's a good representation of your temperature. And then going back to the bicycle speedometer example, what it's really measuring is how fast the wheel is turning in, you're translating that revolutions per minute into a speed. So it's a challenge to take a very small electrical current and turn it into a glucose value. And but that's, that's what we do. And I think that's what all the manufacturers have to do.

Scott Benner 1:02:07
It's amazing. And listen, we're one rabbit hole away from wondering if we live in a simulation. So let me ask a more concrete question. Ready, John? John, in 30 more minutes, we're going to be like, we're probably in the matrix. So just a real quick when Canadians or people who are using other scales, did they multiply their standard deviation by 18? To get their answer? Like, this person gave me an example so that their last standard deviation in Canada was 1.62. They multiply that by 18. To get the number that the way we're talking about it right now.

John Welsh M.D. 1:02:46
They sure would, yeah, so the the units for standard deviation, the standard deviation here in the US as milligrams per deal. outside the US, the standard deviation is millimoles per liter. And the conversion factor is is 18. So the standard deviations would be less by a factor of 18. In places where they use millimoles per liter, the end and that's a good point, thank you for bringing it up. And the point is that what would not change is the coefficient of variation. So if you were to take all my numbers, or if I were lucky enough to be a Canadian, and measuring my sugars and millimoles per liter, I would still have this coefficient of variation of 31.3%. That would not change, because you're dividing milligrams per DL in the numerator, milligrams per DL in the denominator, and those units would would cancel them out coefficient of variation. There's no units for that. It's just a percentage. I'm

Scott Benner 1:03:50
glad you said that, or some person, Saskatchewan was gonna take their coefficient and multiply it by 18. And that's great to know. And thank you for knowing it. By the way, when I asked the question, I appreciate that.

John Welsh M.D. 1:04:04
That's a good one. You know, if you got to, if you were to travel to Japan, you would trade your dollars for yen and you would find yourself 100 times more wealthy. Because you can buy you can buy about 100 yen with $1. But wait, everything's 100 times more expensive so

Scott Benner 1:04:21
well, so let me make sure I'm understanding exactly. So coefficient of variance, or variation we're talking about under 36 Really lessens your possibility of low blood sugar's standard deviation shows us how much stability we have, right like by keeping our variability lower. What is the measuring?

John Welsh M.D. 1:04:45
Oh, in terms of our health Oh, yeah, a one C there's I love a one C I want to strangle it and drown it in a bathtub. i A one C has been with me for a long time. It's about biomarker, it's hemoglobin obviously is the protein that fills up your red cells, it's got the red color, because it's got iron in the middle of it, it's got an iron atom. And it's the same color as rust. The hemoglobin a one C, the a part of it refers to the a chain. There's an a chain and a B chain. The hemoglobin a one refers to the first amino acid in the a chain of hemoglobin. And the C refers to the isoform, if you want to know refers to the isoform, of altered hemoglobin that travels on chromatography. Anyway, that's that's the long answer. The short answer is that hemoglobin a one C is a abnormal form of hemoglobin that has a sugar atom stuck onto it. And having that sugar, I'm sorry, sugar atom, it's a sugar molecule stuck onto it. And it's a nice indicator of how your ambient glucose concentrations have been going over the past two or three months. The downside of having a high a one C is that hemoglobin a one C molecules behave a little bit differently. And they're also markers that things are going haywire in other parts of your body, other proteins in your vasculature in your kidneys, and your liver might be getting decorated with sugar molecules when they really shouldn't be. So having having a very high hemoglobin a one C number tells you that quite a lot of your hemoglobin molecules are traveling around with this kind of gooey sticky sugar molecules stuck onto them. As I mentioned earlier, I it's it's not my favorite biomarker. What's your favorite biomarker, John, there's there's ways that you can fool the hemoglobin a one C test, and we can talk about those. There's some some people have problems with red cell production or red cell destruction that would throw it off. So you can really be misled by an A one C number, it can be too low. And you can say, Ah, you're doing just fine. Your a one C is in the normal range, when it should be much higher. And then on the flip side, you can see in a one C, some people have a one c values that are unexpectedly high compared to what their average glucose values are. So it can it can mislead you in a couple of different ways. I'm a much, much more enthusiastic about just using the average glucose value that you get from a CGM system to assess the adequacy of your glycemic control.

Scott Benner 1:07:50
Is that okay? You know, it's interesting, you made me think of last year I suffered, I had my ferritin was very low. And it's it. You know, at first everyone, the doctors thought I had cancer and we did all these things. And it turns out, I just had low ferritin. And so I got an infusion of of whatever they call it, it's I can't think of it now sit iron and it's a it's a mix, it looks like a rusty bag of water and back up, but during that time, what I was told was we can't trust your Awan see right now, because of your low ferritin. And I was like, huh, dig too deeply into it. But it's something you just said now made me think of it again. And then it made me think about how, you know, measurements, right? And you always get, you could use anything. Here's an example. My daughter has hypothyroidism. But when we first figured it out by her symptoms, the doctor's office looked and said, well, she's low, but she's in range. We don't want to do anything. And we made them give her the hormone, then because we had an experience with my wife who was low in in range, and they would never help her and it really hurt her over time. And so it made me wonder, especially for, you know, women in the menstruation age, is it possible that they have an A one see that looks better than it is if they have lower ferritin just like,

John Welsh M.D. 1:09:14
there you go. There you go. There's that's another of all the ways that a one C could be misleading. That's, that's, that's one of them. And I'm thinking, my own experience, I used to be a really avid blood donor. And I thought, oh, you know, what if I if I were to donate two units of blood, and then wait around for a couple of weeks and then get my a one C measured, that would falsely lower the a one C because as soon as I donate two units of blood, my my bone marrow is going to wake up and say, oh my gosh, John, you did something either stupid or crazy or really altruistic. By donating those two units of blood. We have to ramp up production, and we're going to flood your system with brand new red cells. So after two weeks after donating the blood, I would have a population of red cells, which were relatively young and had not had a chance to get glommed on to by the sugar molecules. And my agency would be falsely low. And I say, Yep, I can sure game the system that way. And that's the same for people who undergo acute blood loss, the A one C would be falsely decreased within a couple of weeks, once the red cell production line kicks into gear. And then people who have shortened red cell lifespans, there's there's some conditions, a lot of syllables, but hemoglobinopathies, if your hemoglobin, if your red cells are, are not up to the task, and if they're prematurely destroyed, you would have a very low a one C, and it would be misleading if you were trying to manage diabetes based on that.

Scott Benner 1:10:55
Okay, so Okay, so you as a person who's had type one for a long time, and is a physician, and I think we didn't really dig into it. But it sounds like you used to help people with type one as well, when you were practicing, is that right?

John Welsh M.D. 1:11:09
Oh, you know, indirectly I specialized in laboratory medicine and also anatomic pathology. So I would, I would look at disease, and I would measure disease and then I and then I went to anyway, so I never directly took care of people who were who needed insulin management.

Scott Benner 1:11:27
But for yourself, then let me just ask yourself that I guess it makes more sense. With your background, and how much time you spent digging around in this data? How do you measure your success? Like which one of these? I know there's going to be a grouping of them here. But but can you tell me what you look at every time you look at your data, just when you want to look and go, oh, I need to do a little more a little less? Like, what what is it your? Where do you focus? And is there any way to put them in descending order?

John Welsh M.D. 1:11:57
Oh, um, well, I am I'm getting old, every if you wait long enough, everybody's gonna get old. I used to worry quite a lot about my agency. And now I I really don't care I what I focus on mostly is the average glucose. And the the example that we're looking at now is 133, which, which is wonderful. And beyond that, I try not to rank myself, I try not to compare myself to my peers. Here at Dexcom. We've got some, some very talented folks with type one who are even more dialed in than I am. If it if it seems like I know what I'm doing, there's people down the hall who are even better. And then there's people in the community who who are need some advice. And that's the mandate, I say, You know what I'm I'm doing fine. But let's, let's see if there's problems that I can address. So I look at my average sugar, I look at the time high and low time and range. And the example that we're looking at 85.9% is pretty good. And then I also look at the the amount of trouble and strife that it causes me and I try to minimize that. I try to settle in on a good routine. That doesn't cause me too much trouble and strife. And finally, after 45 years of I think I've found a good routine for managing my own diabetes. That's

Scott Benner 1:13:23
amazing. That's I think what people need to hear too, it's funny, as you were saying all that I was looking at, at my daughter's nine, like I went to 90 days on her information, because you said average blood sugar. And her average blood sugar has been 115 over the last 90 days within an estimated a once a 5.6. But her standard deviation over that time is like I said, it's it's 45. And is that should I be more concerned about that?

John Welsh M.D. 1:13:54
Well, here's, here's an important question. And it relates to the time that she spends really low and I wonder if there's numbers for either time less than 70 or time less than 54 because because those are those are things that can cause trouble in a hurry. Being being less than 54 is kind of dangerous.

Scott Benner 1:14:14
I have I have her range set as 65 to 120 She's 9% low 54% in range and 37% high but she does not get for the most point we don't go over about 180 ever and under 55 I don't think happens twice a month maybe for long periods of time not like under 55 and falling where people are running around the house you know looking for the will and stuff like that just you know like a dip down that you caught a little too late and and it'll go to 55 and hang and come back up but we don't let her sit under that number. But I look at her standard deviation all the time and I I'm always just like, ah, that's where I need to do better. But like I said, you know, for half of the month, that standard deviation is 24. And then during her, you know, her hormonal times throws throws that number off, like, is that number less scary? Because she's a girl than it would be if she was a boy. I know. That's a weird question. But you don't I mean,

John Welsh M.D. 1:15:25
well, I, I don't know if I'm, I'm gonna take issue with your premise. I, what you told me was, is that number scary? And I? I don't think so. I don't think that's a scary number at all. Just based on the fact that she is so dialed in, and that she has almost continuous awareness of where she is. And she's got good access to to her family and to you and good access to to Kandi if she needs it. So it doesn't sound like she's in harm's way at all. The thing that you know, there's there's some things that are absolutely dangerous. One is one is going low, and finding yourself waking up with a crowd of people trying to resuscitate you is a terrible misadventure. Because you, you went low and you ignore the symptoms. And guess what, you had a seizure, you lost consciousness, you bumped your head. And now the EMTs are out. That's a scary misadventure. So I think if you told me earlier, she's, she's had it for quite a long time,

Scott Benner 1:16:34
she was diagnosed, too, and she's going to be 16 next month. Okay.

John Welsh M.D. 1:16:39
So 14 years, 14 years into it. Hopefully all the autonomic counterregulatory hormones are intact, and I hope they stay that way. So the hypoglycemia awareness, I hope is fully intact, and the counterregulatory hormones that that would kick in to bring her sugar back toward the normal range, I hope are intact. The, the coefficient of variation, you mentioned earlier, the standard deviation for your daughter and remind me of the coefficient of variation.

Scott Benner 1:17:11
Oh, let me get it for you. It does similarly, change with, with what's happening in her I have it at 90 days as 39% in the last 139, in the last week, 36%. But if I go into just the last three days, where like I said, the impact from the hormones is gone. It's 30%.

John Welsh M.D. 1:17:35
Okay, wow. So sometimes, sometimes it gets above that arbitrary number of 36%. So there's some stretches of time where the variability is, is in excess.

Scott Benner 1:17:48
And it's, it's important to note that so my daughter now for over six years has had an A one C between five two and six, two, and we don't restrict her diet in any way. So she'll have pancakes, you know, for breakfast on a Sunday morning. Just as easily as this morning I said she had, you know, an avocado, avocado toast. And so you know, she she's all over the place with what she eats. So we'll have nights where she just has a big salad for dinner, and nothing else. Last night, she had some turkey and small amount of potatoes. But when dessert came out, she wasn't interested. And so she's I call I would call her eating healthy and varied and not excessive. She's not a sweets person, like she's, she'll Trick or treat, but that's the hangout with our friends. And she comes home and doesn't know what to do with the candy. But you don't like that. That's sort of an idea. But, you know, I'm trying to talk through her to everybody so that everybody can kind of get a feeling for how they should feel about this information for themselves personally. Sure, yeah.

John Welsh M.D. 1:18:53
Well, there's, there are some things and we've we spend a lot of time looking at data here we've got some data science, people who built our career on looking at data, there's a couple of comments that might that might be helpful and one is to to look for opportunities to lower the standard deviation lower the coefficient of variation. One is to see if there's any evidence of overtreating highs or lows. And sometimes those really jump out if you look at the, the hourly plot, we call it the modal day plot. Sometimes you'll say, Oh, here's here's something where I know I know where I went sideways on this. I know I had the the big snack after lunch. I shouldn't have oh, there were free doughnuts in the conference room. I should have said no to those doughnuts. So sometimes there's opportunities for looking at your data, not the numbers but just looking at the the image of the 24 hour stretch of daytime you say wow, there's a big spike there. In the early morning hours, maybe I had too much snack before I went to bed. Maybe I have too much my own case, I had a habit of taking too much fast acting insulin to cover breakfast, and I would always go low around nine o'clock in the morning. So being looking at the data, not just as numbers, but as a graph can be very helpful. And it can reveal opportunities for making adjustments. And if if the standard deviation is in, in the high range, if the coefficient of variability is in the high range, then it deserves some some careful consideration about Wow, this is a bumpy ride, are there any particular times of the day that you would like to address with your end might be really amenable to making thoughtful changes?

Scott Benner 1:20:51
Can I ask, given how the numbers are calculated? If? How much is that? What's my question? Are any of the numbers based off of the the range that I've set up? So keeping in mind that my daughter's range is on my phone, it's 65 to 120. On her phone, I think it's 70 to 130. And so on her phone, which is the one that you know, her clarity accounts connected to and everything, if my daughter's blood sugar is quite literally, between 75 and 110 for two thirds of the day, but she has two big meals that spike her to one ad. But she's not more she's not at that one ad for more than an hour and comes back down without getting low. Do those numbers look artificially inflated? If that's how it works for her sometimes?

John Welsh M.D. 1:21:48
The I think your question is, what are the numbers that you see in the clarity report or the clarity, summary. And the time in different ranges? You can, you can set those you can customize the ranges that you want to see for and you can do that in the daytime in the nighttime ranges.

Scott Benner 1:22:08
If I changed her range, this might be a stupid question. But if I pushed my daughter's high number up to 180, would her standard deviation fall?

John Welsh M.D. 1:22:18
Oh, no, it would not know the standard deviation doesn't care whether a number is in the range, the range that you set is pretty arbitrary. You can you can turn that dial up or down. The the range that you set within clarity just tells you when are you going to get beeped. And what are the summary statistics for time and range?

Scott Benner 1:22:40
The data is based off of those ranges. Got it?

John Welsh M.D. 1:22:44
That's right. That's right, the standard deviation coefficient of variation, those numbers are those are not subject to change by just changing the the alerts or the target ranges.

Scott Benner 1:22:57
Okay. And they're based off of what quote unquote normal would be. Is that right?

John Welsh M.D. 1:23:03
Oh, actually, not the the normal range I mentioned earlier than the normal range is no more than 120. And at the moment, I'm just leaning over and checking my sugar right now is it's 109. But for the most part, having having a sugar of 150 would not be concerning. I don't think for any endocrinologist, if you were to cruise around at 150, all day, every day. The endocrinology community would say you're doing a good job, you're a one C is likely close to 7%. And your risk of long term complications is close to baseline is close to what the non diabetic population would have. So that'd be very reassuring. Even if you're having a abnormally high glucose numbers. I got a I got a call once I did some lab tests and for a different occasion, and the nurse called me up and said, John, I've got some very concerning news. Your your glucose is 123. And I thought, well, what's concerning about that? And she said, Well, it's higher than normal. And I said, Well, I have type one diabetes. And and as soon as she heard the fact that I had type one diabetes, she said, Oh, well, you're boring. Have a nice day. Goodbye.

Scott Benner 1:24:24
You mean, my daughter had to give urine one time and I left the room or I dropped off and didn't tell the nurse she had diabetes. And I walked halfway down the hall and ran back because I was worried for the nurse and she was running out of the room at the same time. And I looked at and I went she has type one and she goes Oh, okay. And then she she goes back in the room. Let me re ask my question because I have it in my mind and maybe I might ask another dumb question here. Trust me. It's very boss. I'm ready. So So Arden's blood sugar does sit in the 80s for most of the time, but sure, and and like I said, Sometimes she'll hit one ad on a call couple of meals. What if her blood sugar always sat at 120? And sometimes hit those 180s? Would that make her standard deviation lower?

John Welsh M.D. 1:25:14
I don't think I don't know, I don't think you've given me enough information. To ask that question we could we could do some numerical simulations, which would be interesting, but maybe a quite a digression. I don't think we can tell for sure, just based on what you told me. So it's, it's a big question mark, right now, I'd have to punt and say, I don't know,

Scott Benner 1:25:39
that's fine. I'm trying to I can't wrap my head around my own question, which is frustrating, as you may imagine, and a limitation of my intelligence, but I'm trying to, I'm trying to decide how, you know, so. So you don't, I know, you've heard a couple episodes of the show, John, but you don't listen to the show. And I actually would like to send you a short list of episodes, and let you listen to them and hear what you think of them. But most of the people who listen to this podcast, I would assume having a one C in the fives, or I would think over six and a half, for somebody who's been listening more than three months would be uncommon. And the basic tenant of the podcast is that you don't, you don't stare at a high blood sugar, you get it back down, without causing a low and there's ways to use insulin, you know, with the data that that makes that work. So we, you know, we're pretty heavily talking here about make sure your Basal insulin is right Pre-Bolus Your meals, don't stare at a high blood sugar, you know, don't cause a low bumping nudge with insulin, you know, if you after a meal at a meal time, you know, 45 minutes after you eat. If you're 136, diagonal up, we bump it back down. Again, if you're 85, diagonal down, that turns into 80 that you think this is going to keep going, you don't wait to see a 60 you take in a few carbs, and nudge that that blood sugar back up again, it's like driving between two lines, you know what I mean? Like you don't want to swerve, you just want to kind of try to stay as steady as possible. And we talked about a lot about how to use insulin, temporary Basal rates, both positive and negative, and food in ways that keep those swings from being crazy. And yet, there are people who come back with amazing a onesies who don't get low very often, but have a couple of spikes with larger meals. And these numbers that everyone's telling them, they're super important, you know, standard deviation, they can't seem to get into the space that they want. And then they start thinking about limiting food to make that happen. And I, I think that I think this podcast has a lot of different goals. But one of them is for you to understand insulin enough that you can eat what you want to eat. And I'm not saying that everyone should run out and eat those doughnuts at the conference table. Like, that's not my point. My point isn't, I'm not a person who says, Oh, you have diabetes, you know, don't ever think of you know, don't ever think about your your health, just eat whatever you want, because insulin can take care of it. My point is that if you understand how to use insulin, then you can go off into the world. And with a diet of your choosing, keep your blood sugar's in a more normal range and extend your health. But I'm baffled a little by my daughter's standard deviation. All the other numbers make sense to me. But that one number, I can't wrap my head around.

John Welsh M.D. 1:28:28
Yeah, and and you mentioned, you mentioned the hormonal changes that come by every month and and sometimes the good control becomes more of a challenge, obviously. And the coefficient of variability goes up. And and then unfortunately, the having a high coefficient of variation gives you a higher risk of symptomatic or potentially dangerous lows. But but so it's it's especially important to have that awareness of misadventures on the low side, especially during that time of the month where the swings are, especially high amplitude. The but the goal is, as you said, I think the goal is to spend most of your time out of harm's way. And to live a long happy life where your retinas your retinas last your whole life and your kidneys are going to last your whole life and you're going to die with all 10 of your toes where they belong at the end of your feet. So it sounds like she's well on the way and especially the education that you've been giving her and the insights that she's been getting from from CGM. Sounds like they've been tremendously helpful.

Scott Benner 1:29:40
I appreciate John I just did something that I'm so I feel badly about that because you're sharing sharing your screen. I can't see my screen. And I just realized that I've had you on for an hour and 20 minutes I'm so sorry. I didn't even I didn't really enjoying this and I didn't I didn't recognize about the passage of time. I hope I haven't kept you from something here. not just being polite to me.

John Welsh M.D. 1:30:01
Oh, well, let me You know, I think I had something that I did have something else on the calendar and I hope I'm not. I mean, check my little outlook here. You can see my calendar, there's something coming up at noon, so maybe we ought to

Scott Benner 1:30:15
go is what I was gonna say, yeah, 100% I, I just looked at my phone to look at something about art and to save you. And I was like, Oh my gosh, they're gonna crucify me. I've been I've had you wait too long. Listen, this was incredibly interesting. And I can't really thank you enough for doing it. Because, you know, it's not something everyone jumped up to do when I say can I get somebody who really understand standard deviation talk was a long line of people with their hand up, you know, so I really, I genuinely appreciate this. And I have to tell you, it's gonna go right out tomorrow. I don't usually put stuff out this quickly. But if this fits right into my schedule, so you'll be able to hear yourself and be horrified by your own voice in probably 12 hours or so.

John Welsh M.D. 1:30:57
Well, that's great. So you can I hope you cut out the obscenities and the screaming and and the lawnmowers. And

Scott Benner 1:31:03
all that horrible stuff you did will be cut out now people will just hear you say that and wonder what it is that we

John Welsh M.D. 1:31:10
Scott, what a pleasure, I enjoyed speaking with you, thank you for thanks for reaching out, and I'm a dew point. Dexcom is great. I'm just surrounded by really smart people who love who are really bought into the mission. It's a good company, it's a good product, it's a good mission. And I it's nice hearing about your own experience and your daughter as well. I hope you have a long happy life with with this thing that nobody wants. But we're doing the best we can with type one diabetes, you're very

Scott Benner 1:31:39
nice, John, but to think that you're not going to get drunk back on this podcast at some point is, is not reasonable. I'm gonna get you back here at some point, we'll find out more about you and your diabetes one day. I really appreciate this. I'm going to be incredibly humble all day long after talking to you just so you know.

John Welsh M.D. 1:31:56
I realized you've got to You're the God of podcasts, though. You can go have some podcast swagger, and brag about having a wonderful podcast.

Scott Benner 1:32:03
I'll have to lean on that since I couldn't get out of algebra in sixth grade. So thank you very much.

John Welsh M.D. 1:32:08
Okay, cheers Have a good rest of the afternoon. You too.

Scott Benner 1:32:13
I know that was a denser episode than you're accustomed to on this podcast. But I just thought that having someone like John walk through these ideas was important. I took a ton from it. I'm going to listen back to this a couple of times, because I am I'm not as smart as I need to be sometimes about some of this stuff. But John made it understandable and complete. I was really thrilled to have him on I'm going to have him back someday and just talk about him and his diabetes and try to learn his story. I wish you could have heard the conversation I had with my Booker when I was like, hey, I need somebody from DAX calm to talk about standard deviation, like, really deep dive. Is there somebody over there that can do that? And she was like, I'll find out. And boom, John Walsh comes out of nowhere. Really lovely. Man. I want to thank you for listening. I mean, especially if you're still here, an hour and a half into this, you are a major geek about diabetes data. And I love you for it. Thanks so much to on the pod touched by type one, the Contour Next One blood glucose meter, and Dexcom for sponsoring this episode of The Juicebox Podcast. Please again, go to juicebox podcast.com. For those links, or look right into the show notes of your podcast player. You can clicky clicky on him right there. One way or the other. If you use my links, you'll let the sponsors know that you came from the Juicebox Podcast and I will of course really appreciate that. Hope you're all well, especially in these times. I'm thinking of all of you, and I'll see you soon.

I hope you enjoyed this episode of Best of data. Data. Duda data. People love diabetes data. This is a all time favorite episode of the people. Would you like to save 35% on this sweatshirt that I'm wearing here? Are these silky joggers? Am I rubbing my legs while I'm saying it? I'm not gonna tell you because it sounds creepy, but they're super soft, cozy earth.com Save 35% at checkout with the offer code juice box. And of course you can get 10% off your first month of therapy@betterhelp.com forward slash juice box just by going through that link. It's all you have to do. Thank you so much for listening. I'll be back very soon with another episode of The Juicebox Podcast. If you enjoy this conversation and you're not in my private Facebook group, it's absolutely free and I think you would love it Juicebox Podcast type one diabetes on Facebook private group 35,000 Plus members. That's over 35,000 members, tons of conversations, opinions, perspectives, and great conversation absolutely free. Go check it out.

---

Please support the sponsors

The Juicebox Podcast is a free show, but if you'd like to support the podcast directly, you can make a gift here. Recent donations were used to pay for podcast hosting fees. Thank you to all who have sent 5, 10 and 20 dollars!

Donate