Machine Learning for Blood Chemistry Interpretation [transcript]

Written by Christopher Kelly

Jan. 25, 2018

[0:00:00]

Bryan:    Hello and welcome to the Nourish Balance Thrive podcast. If you are a regular listener, you may have looked down to your computer to see if you downloaded the right file because you may have heard that I don't have a British accent. That's because I'm not Christopher Kelly nor am I Dr. Tommy Wood. But in fact, I'm Dr. Bryan Walsh. And we have me as the host of this podcast today because we have a very special guest that you all know and love. And we thought it’d be a little weird if he interviewed himself. So, we also wanna do something a little bit different. And we’re gonna disprove today that an old dog can't learn new tricks because in fact they can especially when it's presented with enough evidence, compelling evidence, that new tricks need to be learned. So, without further ado, today, I will be interviewing— we're gonna switch around the seats a little bit here— we’re gonna interview the one and only Chris Kelly. So, it feels a little funny to welcome you to your own podcast, but, Chris, welcome.

Chris:    Thank you, Bryan. I’m blushing over here. This is truly an honor someone that has been so instrumental in my own health education interview me for my podcast. It’s a bit weird, but I am delighted to have you. Thank you.

Bryan:    Well, and listeners might wanna know as well. You wanted to know a little bit about the questions I might be asking you and I told you that I wasn't going to tell you because I wanted you to be in the hot seat and squirm a little bit.

Chris:    Well, mission accomplished. I’m squirming.

Bryan:    I got you covered. So, for those of you who don't know, Chris Kelly has a degree in computer science from South Hampton University if that's correct, worked for Yahoo over in London for a little bit, moved over to the Bay Area where he worked as a software engineer for a hedge fund company, found his health in this process deteriorating. He wasn't feeling really good; went through conventional medical circles; didn't find the care, the answers that he was looking for; being an engineer decided to take matters into his own hands and spent a lot of time and a lot of money learning about health, learning about different diets, things that he could do; and has today, I think, probably achieved a level of health that he probably hasn't ever had before. And because of this and he knows that he's not alone with this, he decided he wanted to help other people. And so, he is the cofounder of Nourish Balance Thrive. Is that a fairly decent introduction, Chris?

Chris:    That is perfect. That’s perfected. I'm tempted to use that again, Bryan. Thank you. Cool.

Bryan:    Do I have a new job as the podcast—

Chris:    My voiceover man. Yes.

Bryan:    Anyhow. So, today, I'm actually really excited. And like I said, your listeners may know. I have a long history and passion of being a student of blood chemistry. What we’re gonna talk about today is something that I can honestly say I am more excited about in a clinical practice that I probably have been in over a decade and Chris has been instrumental in putting this together. And I'm very excited about that. We thought there was no better way than to have this information get across and to ask the man himself how he puts some of these things together. So, I wanna start out, Chris, tell me your personal experience and journey has been with blood chemistry just in general.

Chris:    Ooh. Yeah, that’s a good place to start, isn’t it? I mean, so humans are very driven by problems, aren’t they? And I never cared about blood chemistry until there was something wrong with me. Right? Why would I look at that stuff? I don't think I’ve really done a blood test at all until 2011 and then suddenly I’ve done tons of ‘em.  Tons and tons and tons of them and I didn't know what any of the results meant. That was scary to me. Now, I work with numbers in my day job. And not knowing the most important numbers of all seems wrong. And so, I started trying to understand my own blood chemistry and that was what led me to your teaching material. And you know, the rest, as they say, is history.

Bryan:    So, you had all these numbers. I am assuming probably that's how you found my stuff as you were looking them up online. You’re trying to make some sense of it yourself. So, conventional doctors, how does a conventional doctor look at blood chemistry tests?

Chris:    Yeah. So, I had that experience. So, you keep asking why to your primary care doctor. And eventually, they'll show you the decision tree. They actually showed me the decision tree and they said, “Look, there’s nothing out of range. There's no way to go on this decision tree. I don't think there’s anything wrong with you. I need to refer you to a gastroenterologist.” The whole conversation seemed to be leading towards getting me out the door. You could tell that that’s what they really wanted.

Bryan:    Right. Yeah. And I should just quickly say— and this will be in the show notes— we just recently did a podcast where we talked a little bit about reference ranges and the utility of a blood chemistry test. So, if people wanna go back and listen to that, I think it was a pretty good little talk. So, that's conventional medicine. They look at a marker. If it's not out of range, then they pretty much don't know what to do. Now, juxtapose that to let’s say someone in the functional medicine community, how would somebody that's a little bit more informed about blood chemistry— somebody like Dr. Wood like Tommy— how would he look at a blood chemistry a little bit differently than a conventional medical doctor?

Chris:    I think he’s looking for patterns. I don't think he's following a set of rules that are written down anywhere.

Bryan:    Describe what you mean by patterns.

Chris:    I hate to talk for Tommy like this, but here’s what I think he's doing. I think he's thinking about different areas of problems and then he’ll work through trying to either confirm or deny the existence of that problem. So, a really good place to start would be blood glucose regulation. So, maybe he would look at the fasting blood glucose, and then the hemoglobin A1c, and then he would look at C-peptide, and then he would look at the fasting triglycerides, and then he would look at GlycoMark on the test, and then that would start form his impression.

[0:05:11]

    And of course, by the time we got to this point, he will have a detailed history. And so, he will know what the person is eating. And so, all of this information is coming together like evidence in a case of law.

Bryan:    Right. And so, if I could even take that a little bit further, he might even then say, “Well, we understand that this person's glucose is dysregulated, but why?” And then we look at things like cortisol. If cortisol was high, then that might be what's driving this whole thing in the first place or if there's thyroid imbalances. And so, that's what you mean by looking at patterns. So then, as an engineer and as somebody that's been through this, what in your opinion are some inherent problems with this or are there problems? And if so, what are they?

Chris:    So, we talked about the problems in the last blood chemistry podcast. Right? So, if you’re within that 95% of all the people that did the blood test, then you're not gonna be outside the normal reference range. So, the doctor is not gonna say anything at all. And of course, you only have to walk down the street today and look at somebody who is supposed to be normal and know that that’s not right. I read a paper this morning that said they were facing a pandemic. That’s a fancy word. Isn’t it? Pandemic of diabetes. Do you really wanna be part of that pandemic of diabetes? No. Not really. So, I think there’s a problem with being normal today. And Ken Ford talked about that in my podcast as well recently ‘cause I think there’s a real problem with being normal.

Bryan:    Yeah. Go ahead. Go ahead.

Chris:    I was gonna move on to the problems of the way that Tommy does things. And the main problem that I see here is that there’s a limit to the patents that a human can perceive in numbers. I think we’re very, very good at recognizing visual patterns. Right? In fact, if I was to show one of your kids a picture of a cat and a picture of a dog and I asked them to label those 2 images, I think they would do a fantastic job. Even a 3 or 4-year-old kid could do that very, very well. But if I was to show them a blood chemistry and say show me the guy with diabetes, I think they would struggle. Right? So, I think humans are really good at processing visual information, but maybe not so much the numbers.

Bryan:    So then, when it comes to pattern recognition, what you're saying is that it's finite. You described this experience. So, I have created a lot of these things myself as a student of blood chemistry trying to figure out better ways that we can interpret blood chemistry, far better than what conventional medicine typically does and far better than I had been taught. And I had created these— I think you call them decision trees. Describe that experience too in terms of looking at a zinc deficiency for example.

Chris:    Yeah. So, the decision table is perhaps a better descriptor.

Bryan:    Right.

Chris:    Yeah. So, when I went to your blood chemistry seminar in Denver earlier— It was in 2017. [Talk Out of Context]. That was nearly a year ago. And so, you presented these wonderful decision tables. So, if you could imagine a table, right, think of a spreadsheet, there are columns and there are rows. And then in each column, you described a problem, say zinc deficiency. And then in each row, you described what would happen to a particular biomarker in the existence of that nutritional deficiency. And you had put that table together based on your understanding of physiology and stuff that you've seen in the literature. And I got very, very excited about this. I thought “Blimey, a decision table. That's not gonna be hard to implement in any programming language.” And there are some very elegant solutions to implementing decision tables that any program I can use. But when I try to implement them, it didn't work. So, you take, say, a group of 2,700 people who have done a blood test and they have all of the markers that you described in the decision table and then you try and you implement it, right, you try and find people that have low alkaline phosphatase and low neutrophils and all that kind of stuff. You can join all these things together and typically you're left with 2 people. Right. It just never happened. So, it makes perfect sense in theory. But in practice, it just never happens.

Bryan:    Right. So, somebody with a true only zinc deficiency, that was the only thing that they had going on, which is very unlikely quite honestly, that that decision table would probably be accurate for that person. But like you’re saying, so somebody has the zinc deficiency, may show up with low red blood cells; low hemoglobin; low hematocrit; low MCV, MCH, MCHC; and a high RDW, but they also might be dehydrated or they might have hormonal imbalances and they might have so many other things going on that all of a sudden not all those markers show up exactly as it would in somebody. So, it's almost like a zinc deficiency in a petri dish of what somebody would look like. And like you said, there's only maybe 2 people that would actually show up with that. So then, that leaves us with a problem. So, we kind of suck at numerical pattern recognition. There are limits to this because there's so many different confounding variables that goes into some of these biomarkers found in a blood chemistry test and it doesn't happen in a vacuum or a petri dish. You wanted to find a solution to this.

[0:10:00]

    So, we had a conversation, which I told you it. It was probably a year and a half ago where I told you I was sitting on tons of biochemical data like 170 biomarkers per lab test and I knew there was some way of doing some kind of pattern recognition so that I could get more information out of a blood chemistry test than a human typically could. And you started talking about all that stuff. And we started going all techy. Honestly, I don't follow you real well. But you started talking about machine learning. So, tell me a little bit about— just globally for me since I'm a neophyte with all this— What is machine learning?

Chris:    Machine learning is concerned with pattern recognition. So, trying to model data. We’re trying to recognize patterns. And we're doing that by example is my shortest, simplest description of the problem.

Bryan:    So, give an example of this. it doesn't have to be a blood chemistry, but just how would— I mean, in real world terms, how had machine learning actually worked using an example of something?

Chris:    Let me think of an example. Well, can we go straight to the example that we’re using?

Bryan:    Sure. Yeah. That’s probably easier. I didn’t know if you could use that with the dog and cat example or you wanna try and just go straight in.

Chris:    Yeah. Let’s talk about that. This is really interesting. So, the dog and cat example. So, it’s starting to become almost a classic problem in machine learning. And up until recently, it was very hard for a computer to solve the problem of is this a picture of a cat or is this a picture of a dog. And recently, there's one particular flavor of machine learning, the deep convolutional neural network knocks that ball out of the park. It does that now with 99 point something percent accuracy. And that particular algorithm is really trying to model what goes on inside of a human skull. Right? So, we already said that humans— Even my 4-year-old daughter could do a really good job of recognizing the difference between a cat and a dog. Well, this algorithm is and it's that model of what's going on inside of our head. It's not perfect obviously. It’s in fact quite crude, but it's good enough to solve this particular task. It makes sense. Think about this problem. Let’s say that we had a way to somehow break down an image of a cat or a dog into its component parts. Okay. So, I’m looking at my dog right now. These are the ears. Here are his eyes. This is nose. This is his teeth. These are his paws. And you think about it. Even that, the assumption that these tools exist to be able to breakdown an image like that, even that is quite difficult when you think about it. What is an image? It’s just a sequence of numbers. Each number represents a pixel and it's just a sequence of numbers. So, even being able to recognize those shape will be quite hard. But then think about what it would be like if you were to try and hand-roll an algorithm, if we were to come up with a set of rules that would explain whether or not this is a picture of a cat or a dog. I'll give you one example. Okay. So, if the ears stick up, then it's a cat. Otherwise, it’s a dog. And that works for most dogs. Right? So, I’m thinking about Tommy's dog. He's got a boxer. The ears are floppy. The algorithm works. And I'm thinking about my friend, Taresh. She’s got a bloodhound. The ears are floppy. The algorithm works. But now, I’m looking at Kipper and his ears are pointing up. Oh crap, now my rule doesn’t work anymore. And it gets worse because Kipper is looking at me right now because he’s excited and his ears are pointing up. But when he’s done something bad, his ears are down.

Bryan:    Right, right, right.

Chris:    So, the same is true. When you try and hand-roll an algorithm, when you try and construct these rules— You know, if alkaline phosphatase is low and neutrophils are low, that’s like saying, “Oh, if the ears are pointing down and the teeth are sharp—” Do you see what I mean? You just hand-roll an algorithm like that.

Bryan:    You were describing exactly what I have been spending a good part of a decade doing, is trying to come up with the best possible algorithm that I could for different markers and looking at different things that might be showing up on a blood chemistry I love the example that you used because you're using a very simplistic example, but therein lies the problem with creating an algorithm for blood chemistry interpretation that there's no way that we can look at all those variables simultaneously. So then, enter in machine learning. What have you done with machine learning as it pertains to solving this problem that I just talked about when it comes to blood chemistry interpretation?


 

Chris:    Okay. Yeah. So, the cat versus dog example, you can’t do it. Right?  You can't craft a set of rules that would describe whether or not is that a cat or a dog, but what you can do is you can take a million images and then label them, hands label them whether this is a cat or a dog. And then now, algorithms exist today where you can show the algorithm lots and lots of pictures of cats versus dog and then have the model learn by example how to correctly identify the label. So then, when you show the algorithm a picture of a cat or a dog it’s never seen before, then hopefully it’s able to identify it correctly. And I say hopefully. Like I said, the accuracy on that is now 99.97% I think is the academic state of the art maybe. And so, we can do the same with our blood chemistry interpretation. So, imagine you had a dataset where there was some ground truth established.

[0:15:00]

    So, I’ll talk about another dataset actually. There’s a very famous public Pima Indians dataset. It’s part of the UCI dataset. And so, they took this group of people and the task was to try and predict the onset of type 2 diabetes in— I believe it was the next 5 years. And you’ve only got 8 biomarkers that you can use to do it. And it's all stuff that any physician can measure in his office. Things like BMI, blood pressure, fasting glucose. I think an oral glucose tolerance test at 2 hours is in there. So, a skinfold calipers on the triceps. So, the standard stuff you can measure in any doctor’s office. And so, what this dataset now allows the data sciences to do is learn by example. There’s lots and lots where they did this test on thousands of these Pima Indians. And then a data scientist can build a model by training it by examples. Oh, this one got diabetes. This one didn't get diabetes. This one got diabetes. This one didn't get diabetes. And so, the model, as it sees these examples, it builds this very sophisticated statistical thing that allows it to identify future instances of diabetes given those same 8 input markers. Now, this was the problem that gave me the inspiration you see. When I saw those columns, I saw these 8 biomarkers, I thought “Well, what would happen if you were to switch out BMI for your sodium level in your blood and then what would happen if you whip out the skinfold caliper measurement with potassium?” And so, it goes on. And so, you’ve got 38 markers from a basic blood chemistry. And the ground truth, maybe we're not just interested in predicting the onset of type 2 diabetes, but then something else as well like zinc deficiency. And we can define zinc deficiency as being a specific level of plasma zinc. So, the value proposition that we've created here is now you can do a basic blood chemistry and find out what the value of the plasma zinc would be without actually doing the test.

Bryan:    I can say by training— If you have listeners that are as dumb as me when it comes to this, by training the machine learning algorithms and saying here are— let's just right now say 1,000 blood chemistry test that have zinc in it— Is that so far so good? You enter that in and you say this is what a zinc deficiency looks like with all the other markers and then it learns, if you will, what that pattern actually looks like so that if you have another now, the 1,001 test that you give it that doesn't have zinc on there, based on all that other stuff, it can predict what the zinc would have been on that individual had that been run. Is that right?

Chris:    That's exactly right. Yeah. So, it's pattern recognition that's been learned by example.

Bryan:    I wanna find out a little bit more about what it is that has actually been created here in terms of software program, but then what problem does this— I mean, what benefit does this have to clinicians, to nonpractitioners or nonclinicians? I mean, what problem is this solving in my opinion for the first time ever?

Chris:    That's a really good question and you could just say just run the plasma zinc. Like who cares, right? It’s a 5-dollar test. And that may be true. I’ve not looked at the price of plasma zinc, but we have this expensive baseline panel that we do for members of our elite performance program and that’s a 10,000-dollar program. And the reason it costs $10,000 is because most of the money that I collect from our clients, I pass it right through to the labs that perform that test.

Bryan:    The laboratory.

Chris:    Right. Yes. So, my accountant is like “Wow, you're doing great.” And then he gets further down the balance sheet. He’s like “Oh wow, your company, it looks kinda like kayak.com where they’re taking a lot of money upfront for airline fees and then they just pass that money right through to the airlines that are performing the tasks.” And so, this stuff adds up really quickly like we talked about in the previous podcast. You know, you can add 1 or 2 expensive markers. Maybe I add a GlycoMark and that only adds $60, but that's just one marker. If you end up adding 100 markers, before you know it, you've got 1,000-dollar blood test. And in fact, Tommy and I went to the Buck Institute for research on aging and we did their [0:19:06][Inaudible] panel. We did his training and I put together is panel when I got home. And the guy from the lab, he came back with a quote for $2,000 for that blood. I’m like “Holy cow, I got charged $2,000 for a blood test.”

Bryan:    It adds up quickly. I mean, something like ceruloplasmin is a few dollars. Adiponectin is 300. And then the more advanced markers, I mean, you’re looking for things like aging, it can really add up quite quickly. And that's just with the blood chemistry test. So, who do you screen or run a test for H. pylori, or some kind of infection, or some kind of nutrient deficiency, or some kind of toxicity issue? And so, to continue telling me what problem this solves, I mean, what is— Listen, I have my own ideas on this if you don't wanna answer. What is the utility of this? What is the game changer about this, the revolution about this, and could this change clinical practice?

Chris:    Yeah. I mean, so, there’s 2 big things here. The first is the cost. Right?

[0:19:59]

    We’re bringing down the cost of doing a very high quality of investigative medicine. Right? We’re not having to do 2,000-dollar blood test. Another is the geographical availability. Right? So, the elite performance program is all fine and dandy for people who are in the U.S., but getting a stool test done when you live in Jordan is actually not that much fun to do. You know, if you don’t have FedEx and then even if you do, it's expensive and it's challenging. Right? And so, in an ideal world, I would do all of the tests on all of the people. But in practice, that isn't possible. And so, what I need is at the very least a navigational aid. Right? So, I think of this thing a bit like my GPS. I can do the predictions and then I could know in what direction to go next. I'd love to hear your thoughts.

Bryan:    Yeah. Listen, I'm still asking you the questions here. You're right. When you said lowered the cost, how many— And just give an example of some markers that are required in order to make accurate predictions and we're still kind of skirting around this whole topic that I wanna eventually get to here. But how many markers do you need in order to predict nutrient deficiencies or toxicity levels or, you know, you mentioned GlycoMark, without running GlycoMark or hormones, how many markers are actually required?

Chris:    So, it’s just 38. It’s a very basic blood chemistry.

Bryan:    And give me some examples of what these 38 markers are.

Chris:    These are the tests that everybody either has done or should be doing. So, it’s normally regarded as or named a comprehensive wellness profile or comprehensive metabolic profiles. So, the glucose and sodium, potassium, albumin, globulin, ALT, AST, GGT and then we do an iron study. So, we include total iron binding capacity. We include ferritin and then we do a lipid panel, so total cholesterol. Tommy and I talked about all of that in a recent podcast. And the complete blood count of course that we talked about.

Bryan:    CBC. Right.

Chris:    You know, I was looking at a paper just yesterday that said having an elevated red blood cell distribution width— Remember we talked about that the last podcast?

Bryan:    Oh yeah.

Chris:    Having an elevated red blood cell distribution width was a similar risk to being 80 years old.

Bryan:    Yeah. Yeah.

Chris:    Holy cow. Do I wanna fix that?

Bryan:    It’s associated with Alzheimer's mortality and a whole bunch of things. And as you know having been a student of blood chemistry yourself is it's never talked about. It happens to be on a CBC and someone will say it needs to be under this range, but never talk about ‘em. That’s a problem. Anyhow. So, it's 38 markers. There's glucose, electrolytes, wastes, proteins, enzymes, lipid panel, and a CBC with differential. And what's the cost for somebody just to run that out of pocket?

Chris:    Oh, the out of pocket cost in the U.S. is about $65.

Bryan:    So, $65 for that panel and using those markers, which most people will probably have most of those markers from their doctor. Ferritin, TBIC, GGT. Sometimes those are a little bit more obscure and aren’t often run. But what are some things that you can predict? Well, what have you created? First of all, just tell me what we've created here.

Chris:    Yeah.

Bryan:    What's it called? Give me the name. I wanna just throw it out there.

Chris:    It’s the blood chemistry calculate.

Bryan:    The blood chemistry calculator. And it's a software program that you can basically enter in those 38 variables, age, and gender. And what are you able to predict and with what accuracy?

Chris:    Oh sure. So, we’ll be able to predict all kinds of things. We’re able to predict nutrient deficiencies, toxic metal exposure, nutrient element overload. So, iron overload is a real problem in the people that we see. Xenobiotic exposure. And we got some infectious stuff there that we were able to predict. So, you mentioned H. pylori. We can also predict some worms, toxoplasmosis. What else? Oh, hormones as well. Testosterone for men, estradiol for women, T3 system thyroid hormones, and some inflammatory markers. Right?

Bryan:    And with what accuracy?

Chris:    C-reactive protein. Oh. So, the accuracy. In data science and in medicine too, they use these 2 markers that are called sensitivity and specificity. And that enables a clinician or data scientist to know exactly how good the test is and everything we've done— like I haven't included anything that I couldn't get to about 90% sensitivity and specificity. So, this is approaching as good as doing the test. Right? So, when you send a stool sample into the laboratory, that’s not a perfect test either. Right? There’s difficulties and challenges with doing microbiology especially if somebody is looking at something under a microscope. They might have had a bad day. They might miss something. Maybe the culture didn’t work properly. And so, even those tests are not perfect. Those tests are usually around 99 and 99 for most of the stuff that we test on our stool panel for example. Even just without doing the test, we’re still approaching the gold standard for sensitivity and specificity.

Bryan:    Yeah. I'm glad you said that because in the world of medicine, if something is 90% both in sensitivity and specificity, it's considered to be the gold standard of a way of testing.

[0:25:08]

    So, sensitivity and specificity is basically testing what it's supposedly testing and not identifying what it's not supposed to be identifying. So, the higher the number, the better. So, let me just get this straight. And by those 38 basic markers, a 65-dollar panel, give me a break. With those basic markers with 90% accuracy or above, you're able to tell me if my hormone level of testosterone is high or low, or cortisol, or estrogen, or xenobiotic. Like you said, environmental pollutants. I think we did a talk earlier about how you can't test those. There's really not a really accurate test for this. If I were to give you those 38 markers, you could tell me if I was deficient in zinc, or copper, or selenium or not. You could tell you what my thyroid hormones would be. You can tell me what my sex hormones would be, whether I may have had an infection of some kind, if I have toxic exposure to metals or toxic exposures to xenobiotics whether phthalates, bisphenol A. You could tell me all those things just from those 38 markers.

Chris:    Yeah. Amazing. And then also the blood glucose stuff that we care so much about as well. Right? The oral glucose tolerance test, and elevated plasma levels of free fatty acids, and elevated hemoglobin A1c, and low GlycoMark which we talked about in the previous podcast as one of the earliest signs of type 2 diabetes. So, that’s a lot we can do with just a basic blood chemistry.

Bryan:    So, you can tell me what my oral glucose tolerance test would be just by using those 38 markers.

Chris:    Yeah. And this is an important point. So, I’ve turned the problem into what they call a binary classification. Right? Is this a cat or is this a dog? I try doing what they would call a logistical regression. So, that would tell me the exact value of the biomarker that we’re trying to predict. So, maybe I’d tell you that your GlycoMark is exactly 13 and then you do the test and you find it what it actually is. But in practice, that isn't terribly helpful. So, you're not going to do something different if my GlycoMark is 10 versus if it’s 8. Technically, it’s not important. So, you know, I thought it was easier, simpler to understand. We’ve already wandered into quite difficult territory here. We’re thinking improbabilities. And so, I thought it was easier if I turn this into a binary classification problem. Do I have a problem with GlycoMark? Yes or no?

Bryan:    And I'm glad you did that. And I'll tell you that that's the beauty of being a computer engineer and clinician because you could have gotten hung up and it's very sexy to come up with the specific numbers. But in reality, what does a clinician want? Is your testosterone low or is it now? Are you zinc deficient or are you not? Because if you are, let's deal with it. If you're not, let's move on to something else. I’ve joked around you guys a couple times about this. I think I've just been at this longer than you have because the level of excitement that I have and I'll say this again that I can give you those 38 really basic, really inexpensive, very old in terms of science markers. And you can tell me what my oral glucose tolerance test not quantifiably, but you can tell me if it would be higher or low, abnormal just by running those. I cannot tell you what this is going to do to a clinical practice. Well, I'll just give you some examples. I have some really challenging patients that are quite honestly I don't know when they’re a head case or when there's something going on. There's only so many things that can go on with somebody to cause them to feel miserable. Maybe it's an infection. Maybe it’s immune systems. You know, nutrient— I have head scratches, man. I just like what the heck is going on? This tool can help the clinical decision making for these really challenging patients to say, “All right. Well, what direction should I take first or what might I turn up that I didn’t even think of looking at?” If that makes sense. You know, we all go to seminars. And so, I go to the GI seminar. And so, I go to the GI seminar. I learn about all these bugs, and stool testing, and probiotics. But then a couple years later and a few seminars later, I’ve kind of forgotten that stuff and it’s not on the tip of my tongue and not on the top of my decision making. So, something like this, somebody can quickly see what may be going on that's an issue and a priority and what's not. And for a clinician and even for a nonclinician quite honestly, that's incredibly, incredibly valuable. I can go ahead and share. I shared this with you once, but I think listeners need to hear this as well. So, we've known each other for a while. I believe you. I don't know how I feel about computers and I trust them, but they’re a little bit scary as well. And so, you were coming up with this stuff and that was all well and good, but I'm a bit of a skeptic. And so, what I did is I took 3 really comprehensive panels that have all the hormones and then some. It has all the glucose markers that you just talked about. It has tons of markers on there. And I didn't send you the whole panel. I just sent you those 38 markers and then you send back the probabilities, and the predictions, and the forecasts of this. And I gotta tell you. Man, for all three of those, it accurately— The higher the probability or predictability of a given marker was spot.

[0:30:00]

    Low testosterone, abnormal reticulocyte which nobody runs. I’m trying to think of some of the other ones, but a C-peptide I think was on there. I could not believe— I have talked for years about the crystal ball, the clinical crystal ball. And I have to be honest with you. This is the closest I've ever seen anything come and it's really close to being that crystal ball. And so, I wanna commend you for all the work that you've done in putting this together. So, it’s called the blood chemistry calculator. Who is this for? Who can use it? How does one find it if this is what they wanna do?

Chris:    I mean, it’s definitely for geeks like me, right, that have already run some blood chemistry and interested in knowing some possible areas to investigate next. I know that not everybody listening to this will be running blood chemistry regularly, but I think that everybody listening to this would have done some blood chemistry at some point in their life. And if you haven’t and you’re interested in doing some blood chemistry and using the tool, then we can order a blood chemistry for you as long as you can get to Quest Laboratories in the U.S. And they are a high street franchise. You can find them in almost every major and even minor city as well. I mean, I think the other person it’s for is for me right now. It's interesting to think about the history like did I need to be sick and then work as a coach and to know what software I needed to build, right? That’s the really interesting thing for me to think about. Would I have been able to build this software whilst I was still working for a hedge fund before I’ve been to any of the health stuff? And the answer is I think I probably could. And so, that maybe is what makes me a little unique in that regard. Yeah, that’s what I’m thinking, is people who are working in a similar business to me. Someone who is either a coach maybe or a personal trainer or maybe a medical doctor who is thinking about starting a concierge practice.

Bryan:    That brings up a good point. So, let's say somebody is a fitness professional. He’s a personal trainer. I mean, they're seeing clients that likely have some things going on physiologically. You’ve run enough labs on “healthy people” that definitely have abnormal things going on. So, let’s say there’s somebody that they can’t diagnose that’s outside of their scope. And in fact, it’s illegal. So, let's say a fitness professional. How could a fitness professional use this and how would he or she use this?

Chris:    Sure. Yeah. Sign up for the monthly membership. I really want people to sign up for the monthly membership. And the reason is that gives us a steady reoccurring income so that we can do things like more software development, more research. We're gonna do monthly webinars where we teach blood chemistry analysis and then also help people understand some of the full cost that’d be made by our tool. Yes. So, you can sign up for the membership. You could either take your client’s existing blood work or you can order a test for them. And then you can run the results through the calculator. There’s no data entry for anybody to do by the way. Even if you just sat on a JPEG image that you took with your cellphone, you can still upload that to the calculator and we will take care of the data entry and then we're gonna generate a report, which you can use as a navigational aid to decide what to do next. Right? So, if you’ve got a client that has some sort infectious thing going on or maybe they’re not handling carbohydrates well or maybe they have some sort of nutritional deficiency, or xenobiotic exposure, or maybe heavy metal exposure, then no amount of shouting at someone to do more burpees and hoping that they show up at 6 o’clock in the morning before work on Monday is going to get them the results, right, as you have talked about, Bryan.

Bryan:    I thought burpees fixed everything. No?

Chris:    No. Unfortunately. Yeah, I mean, we talked about it. We talk about this all the time with clients. We’ve just found some sort of protozoa and parasite infection, you know. Like the reason you had to drop out as a pro cycler is just because you had Giardia and no one was able to diagnose it. And the fact that you started meditating was probably really helpful, but no amount of meditating is gonna have the Giardia leave. Oh my.

Bryan:    No. Right. Right. So, for a fitness professional, if they wanted to use this because they can't diagnose, they can't treat, they really shouldn’t be even playing around with too much nutritional intervention, though they do, and you know that's because there's not a lot of good resources for their clients to go to, but so, if they were to do something like this, then it would show them probabilities of things that might be out of balance that they're not gonna do anything with. But what could they do that with that? They refer their clients to a physician or some kind of functional medicine practitioner to look a little bit deeper into it?

Chris:    Yeah. Absolutely. And so, we do that all the time. You understand the limits of your knowledge. And when you realize that you can count on something that you don't know how to deal with, then the next thing you should do is like find someone that can help and make a referral.

Bryan:    But what this fitness professional might do is actually uncover something that's fairly significant that's maybe been going on for a while that honestly practitioner or a doctor may not have ever found if not for a while. So then, how would a licensed practitioner— And by that, I mean it could be a naturopath or maybe a medical doctor, even a conventional medical doctor, how would a licensed practitioner use this?

[0:35:04]

Chris:    Yes. So, I think, again, it’s that navigational aid. So, I think that you're not going to be able to run a whole bunch of expensive tests and have those covered by insurance. I've talked to doctors recently that are getting a hard time because they’re running things like 25-hydroxyvitamin D. The insurance company saying is just give them the supplement already, people who live in Seattle. And they're saying, “Oh, reverse T3 is not a fully verified marker. You shouldn’t be running be running that. We’re not covering it.” So, it's basically the insurance company is running the show when it comes to your health. And so, the licensed practitioner is not able to do all of the testing that they would like to do, but this, again, could act as a navigational aid and then make your decision that you’re gonna do another test really, really solid.

Bryan:    Yeah. And so, that was a piece that I wanted to say earlier, is when you get stuck with a case, this can be really, really helpful, but I think that's the other piece to this. I mean, my follow-up  question to you about this after or before I say this is does this wipe out the need for these other functional medicine fancy expensive tests?

Chris:    No. Not at all. I don't think. Like I said, it's a navigational thing. So, I'll give you my own personal example. Right? So, the first time I run this thing on my own blood chemistry, it says that I have inorganic mercury exposure, which comes from mercury amalgams. And I know that I have a mouth full of mercury amalgams. Right? And I know that we have the quicksilver scientific mercury try test that we do in our practice all the time and I've seen the before and after when people go through a biological dentist and have those pulled out. And I've even seen a before and after after people have done Tommy's detox article that he researched and put together. Even without doing the dental work, you can still fantastic results in blood levels of mercury being reduced. But for some reason, I've never done the— Why have I never done that? I don’t know why I’ve not done. So, the moment I saw that, I’m like “Okay, Tommy, can you send me mercury try test kit please.” I'm not gonna start go doing $10,000 worth of dental, although arguably you could. Like that's not good. You should maybe just go do the dental work anyway. But really, that’s how I think. It’s like, okay, what’s the next logical thing for me to do? It’s like to confirm or deny this forecast.

Bryan:    And that’s what I was gonna say. So, for a licensed practitioner, this is the first test that they should run because in essence what it’s going to do is, first of all, it’s a lot less expensive for the client or the patient, but it will sort of dictate to you which direction you should go. I mean, if mercury is off the chart according to this test and I've seen these tests, mercury led, arsenic, all these things, they might be coming in with difficulty losing weight or maybe— I don't know— a little bit lack of energy or some skin issue, but then you see this. You’ll go like “Well, listen, these are off the charts right now. I think we need to address this and let's see if your skin condition improves by way of doing this.” So then, you mentioned doing some follow-up testing based on the results of this. But because of the machine learning and looking at these patterns, is it conceivable that somebody like yourself does the 65-dollar 38-panel test then do something like Tommy's detoxification protocol for you and then 3 months later redo that 38-panel test and see if the probability or predictability of mercury came down? Is that something that's possible as well? Do you even need to do that follow-up test initially?

Chris:    Yeah. That's a really good point. You could totally do that. When you look at Tommy's detox protocol for mercury, the main downside is financial. Right? It’s not like you’re taking a bunch of hardcore drugs that have potential unwanted facts. You could totally do that. You could just do the 65-dollar test. You could do the detox protocol, right? You’re getting all of that good stuff and then redo calculator report and then see how the probability of mercury has changed. That’s a really good point.

Bryan:    And at that point— ‘cause quite honestly, that's how I would use it. If it didn't come down after that 3-month period of time in doing Tommy’s, then something is being missed and maybe doing a follow-up test at that point would be more reasonable if that makes sense. So, you come in with low testosterone. Rather than actually running testosterone, let's play around your diet, your lifestyle, your nutrition. Take some supplements. Try some things. Redo this freakishly inexpensive 65-dollar test. You could even do it 1 month later, 2 months later, 3 months later. For $65, why not? And then because of the way the machine learning algorithm works, if other things improve, then you could see the probability of low testosterone to come down indicating that their testosterone is probably improving. Is that right?

Chris:    That’s right. And it doesn't preclude you from looking at that blood chemistry in the way that you always have. Right? If you're a doctor that’s used to looking at blood chemistry, you just have to look at the blood chemistry. Right? So, we've done that with our health assessment questionnaire. We've used a similar technique in the past where we've taken subjective questions given to athletes and to use subjective questions to predict the results of some of the other testing that we do in our practice.

[0:40:00]    

    And that has worked really well too, but that doesn't stop us from looking at the questions subjectively. We don't just look at the output from the machine learning algorithms. We also look at the were you satisfied with your sleep like that's really, really valuable information. So, you can still look at the blood chemistry qualitatively or using— So, this is the other made feature of the blood chemistry calculator, are optimal evidence-based, health outcome-based reference ranges. So, you can still look at the fasting glucose and say, “Okay, that's now 83. That's a really, really good thing. It was 103. That's a huge improvement.” I mean, I don't care what the algorithms say like I'm gonna be super happy with that result regardless of what the algorithm says.

Bryan:    Yeah. And I don't think we have time to jump into that whole optimal reference range and the evidence base, but this is why this is— I have not been this excited about anything in a clinical practice and I honestly can't tell you how long because not only does this do the predictability and the forecasting that we've been talking about, the machine learning algorithm, but there's a set of optimal reference ranges that is based off the scientific literature, which is the first time in this industry that's ever happened. And so, not only do you get their predictions, but you can see if something might be slightly out of range. Even though a doctor will tell you that it's normal, it's outside of our optimal reference range, which is based on scientific studies to see if there may be something going on a little bit earlier before it becomes more of an issue. So, if somebody is interested in it— And so, nonpractitioners, nonclinicians have the ability to do this as well. So, where do they go? Is there a website?

Chris:    There is a website. Bloodcalculator.com.

Bryan:    So, bloodcalculator.com. And then what's the process? They go there and then what happens?

Chris:    Yeah. You click on the button on the front page of that website and you can make a decision right there. If you know you're gonna run a bunch of blood chemistry through it, then you may as well join the monthly membership. Because when you join the monthly membership, that brings the cost of each report down to $5. And so, if you’re a practitioner and you've got a big stack of blood chemistry you wanna run through it, then it makes sense to join the membership and then you can also attend the webinars. If you just want to run a single report, then you can also do that. The price is $50. And if you need us to order a blood chemistry, then we can do that for you. And the cost of—

Bryan:    That was in the U.S. Right?

Chris:    That’s in the U.S. And the cost of the blood chemistry is $65. Personally, we have our athletes do blood chemistry at least twice a year. I think it's a really good way of tracking progress. You know, you put so much time— You know, for people listening to this who are all athletes, right, you put so much time into your training and racing. Like why would you leave something like that on the table? Do you think the world class athletes are not doing blood chemistry on a quarterly basis? Yes, of course, they are. Listen to my interview with Jeremy Powers. He would tell you that he does blood chemistry every quarter and that's because he's performing at the very highest level.

Bryan:    But I would even argue at 65 bucks and with all the predictions and things that are being predicted and therefore somewhat tested for but aren't normally tested, you could even do this every couple of months quite honestly to see if there's any change in particular things. I don't see a reason against that because it's such a small blood draw and it's so expensive. So, they go to the website. They can do the monthly then they also have the ability to upload their own report or in the U.S. order their own. You mentioned webinars and you haven't talked about this. Is there anything else in there like— I don't know— a fatty liver index or atherogenic index? What else is coming down the pipe? What else can this thing do which is revolutionizing as I said? And I don't use that word lightly. How we're gonna enter blood chemistry test.

Chris:    Absolutely. I’m just so excited at the moment about being a software engineer. There’s never been a better time to be a software engineer. The work that we’re doing over the last year has just been pure delight. The things that used to be so difficult to do. Like remember when I used to work at Yahoo and I would go get coffee whilst I was waiting for this 200-megabyte shared object to finish compiling. So, I’m using computer talk now, but you know what I mean. You’d have to like go away and come back and wait for the computer to finish just to see the change that you had just made to actually have some effect. And now, everything is so false and you've got all these amazing tools like the machine learning algorithms. I think that's important to point out here actually, is that I didn't create this algorithm. I just used somebody else's algorithm. And the algorithm is called XGBoost. And I can link to a paper on XGBoost in the show notes for this episode. Really what I’m doing here is standing on the shoulders of giants, which I think is really important. But yes, so what we got coming down the pipeline. So, you, and Tommy, and Meghan are doing this amazing job where you're sending me these papers saying, “Hey Chris, could you implement this?” I’m like “Oh yeah, I could implement this.” You think it’s important and you say, “Hell yeah, I think  it’s important.” And so, we've got this really nice team thing going on at the moment. So, the fatty liver index is something that's already implemented in the calculator. Bryan, do you wanna say something about the fatty liver index before I move on?

Bryan:    Well, just the super short version is that the gold standard for identifying fatty liver is a biopsy. Unfortunately, it’s a little invasive. A researcher came up with this fatty liver index using some really basic biomarkers, very accessible biomarkers, that when put up against people that had fatty liver did exceptionally well in identifying fatty liver just using these basic biomarkers.

[0:45:09]    

    So yeah, that's the fatty liver index. And you’ve included that. They can predict whether or not they might have fatty liver. There is I think the atherogenic index, which is another very basic, but validated, calculation looking at the likelihood that somebody might have atherosclerosis. I’ll finish your sentence for you, but calculations for viscosity, looking at the likelihood of things like dehydration and maybe why somebody is  dehydrated, osmolarity. There’s the wellness score that you're looking at that I personally am extremely excited about once we figure that out. Anything else coming down that people should get excited about?

Chris:    Yeah. The main thing I wanna predict is mortality I think would be revolutionary. Okay. So, we've already used that word. So, I’m not sure what word I’d use for being able to predict mortality. And I think I might have it working today. That's how fast things are moving. I think I might have it working by the end of today. So, there's a paper I can link in the show notes that describes the intermountain risk score and it uses as input all of the markers that we've been talking about and then it predicts 30-day, 1-year, and 5-year mortality based on these really basic—

Bryan:    Super basic. Yup.

Chris:    It’s been validated in the scientific literature. It’s been validated against the [0:46:23][Inaudible] data. And so, what I'm hoping this will be able to do is you put in your basic blood chemistry. I show you your risk score, which is basically a proxy for your risk of mortality, right, when am I gonna die and then you make some changes. You run it through the calculator again and you’re like “Oh shit, I just started 10 years in my life. That’s pretty freaking awesome.” What better motivation could there possibly be than that? And I think the crucial thing that we're changing in health right now is closing that loop between the change and the feedback, right? You don't normally get that. When I'm driving a car, you turn the steering wheel and you find out right away what's happened. Whereas in health, it’s not like that. You know, we do work with some people that have some almost overnight success or overnight wins with improving their health and performance. But for the most part, it requires a lot of delayed gratification. Right? You have to kind of have faith that this is doing something for your health span. And you may never find out. Right? And there’s no counterfactual. How can you know that all the work you put in in your 40s meant that you didn’t slip up in the shower and break a hip, right? You didn’t lose your fast twitch muscle and that’s why you’ve been weight training all those years. You can never know that. And so, I think closing the loop is gonna be the important thing in improving health.

Bryan:    Yeah. Yeah. And back to that intermountain health score, there's those few papers that you come across into your perhaps career that when you do, you’re like thank God I found this thing because this is more awesome than most other scientific papers. And what I think is really unique about that and speaks to what you were just talking about is that somebody— I mean, we’re talking about mercury. Let’s say that mercury didn’t come down and that could be really disheartening. They just did this Tommy’s mercury detox protocol for the past 3 months and spent how much money on it, but their mercury didn’t come down, but their mortality score improved. And that is significance. And without that and without that built into this software program, what are you left with? As a clinician, what are you left with? You’re left with a patient or client that may not stick with you anymore because you're not getting them anywhere. It feels like they spend a bunch of money and now they're frustrated because they can't look this thing. But in fact, if you have other biomarkers and this incredible intermountain health score or mortality score, if that improves, you know that you're making progress. Yeah, your mercury may not have come down, but something's happening. You're getting better. And maybe if we stick at this, then your mercury is gonna come down. I'm extremely excited about that going in as well. Well, I had one other question. So, when you say Yahoo, you put the emphasis on the second syllable. Is that how the Brits do it or is that just you? Over here in America, we say Yahoo.

Chris:    I think you’re probably right. Yeah.

Bryan:    I heard you say Yahoo. i wasn't so sure about that. I think they're both right. Any other final thoughts?

Chris:    Yeah. The final thought, you know, something we didn't touch on, which I think is really important, has come up a couple times since we launched the 2 questions that people have asked. The first question and this is a really great question. How’d you know the sensitivity and the specificity like how do you know how good these predictions are? Are you just relying on people like Brian doing these held out data test and then confirming or denying? Obviously, there could be bias there. And it’s a fantastic question. And the answer is that when we train our models, we hold out a quarter of the data for testing purposes. And this is very standard operating procedure in data science. So, imagine I have— Not imagine. I mean, this is what really happened, is we have 36,000 blood tests in a giant spreadsheet.

[0:50:00]    

    So, the first thing I do is I take 9,000 rows of that spreadsheet and I cut them out and I save them in another file and then I don't touch them again until I'm ready to test my models. And so, that’s how we know the sensitivity and the specificity. Another common question that's come up recently since we launched the tool is what do I do if I don't have all of your required markers? And one of the beautiful things about XGBoost is its ability to handle sparse data. So, what that means is it handles missing things really, really well. Now, if you're missing some of the markers that were most predictive in the problems that we're trying to find— So, we talked about red blood cell distribution width. It’s really, really important for predicting stuff. GGT also seems to be very predictive. Eosinophils from a complete blood count with differential also seems to be highly predictive of a lot about of bad things. So, if you’re missing some of those markers, I would probably try and get another blood test before you do the report. If you're dismissing lactate dehydrogenase, which for some reason is common on people that use Quest Labs like it doesn't seem to appear as part of their standard test and then also mean platelet volume doesn't seem to appear on a standard LabCorp CBC with differential for some reason, but it does appear on Quest, and it’s not that you can’t get mean platelet volume through LabCorp. You can. It’s just if you just asked for a CBC, they won’t give it to you. So, there’s definitely gonna be— Now, like nearly everybody listening to this will have a blood chemistry that is almost good enough. And if it’s almost good enough, you should just run it through the calculator and not worry about it.

Bryan:    Those were both great points. The withholding data is a critical piece. You're not just making this stuff up. You take some of it out and then you put that back in to see how accurate it was. That’s a really important point.

Chris:    You know what? I wanna say something. I finally feel like a scientist. You know, it says computer science in my undergraduate degree. I’m like I don't remember doing any science. You know, when I first started hanging out with Tommy, like I’m not a scientist. I’m an engineer. Why do they call it computer science? This is silly. And now, I’m finally a scientist. It’s amazing. So, you know, I make some predictions about the future, right, and then you’re able to test those predictions. And if they hold, then your hypothesis was correct, which is proven, but—

Bryan:    Welcome as a scientist, Chris. Nice to have you amongst us. Yeah. And another bit, I had— A friend of my wife, I just plugged in his stuff last night incidentally and it was a pretty lame panel. It didn’t have GGT. it totally lacked a lipid panel as well. So, it wasn’t really good at predicting some of the glucose stuff. But you know, I was surprised that some of things that came up for this guy, bisphenol A was the highest thing, which was something to certainly consider. Anyhow, any other final thoughts that you might have before we wrap this up?

Chris:    No. I think that’s about it.

Bryan:    Cool. All right.

Chris:    I mean, I'm really interested to— I mean, that was the final thing I wanted to say actually. That sort of thing. There’s this problem I see and I was guilty of this as well when I was a junior software engineering just after I graduated. My ego was wrapped up in the things that I produced. And when people said to me “oh, this is crap, Chris; it needs to do this, that, and the other” and you get this in finance a lot, right, like people who do not— who punches in finance. You know, they’ll bang the desk and say, “You fucking idiot, we just gone on all of this.” And so, you learn that your ego must not be wrapped up in your work. You have to embrace failure. Say “Thank you, Mr. Traitor. I'm so glad that you found a bug in my software. I'm gonna go away and fix this to make it better so that tomorrow this doesn't happen.” Right? That’s the right approach. And so, that is the approach that I’m taking with this software. If you’re listening to this and you’re thinking that’s bullshit or you’re thinking “oh, I can think of a way that this guy can make it better”, don’t think that you're going to insult me or damage my ego in some way by telling me because, you know, I wanna hear it and I wanna make it better. And if it turns out I'm doing something truly dreadful and I should just stop, I would totally stop and do something else. So, I'm really, really anxious and looking forward to hearing people's feedback, you know, when they listen to this.

Bryan:    Well, I can give you my feedback. I don't think there's much— As a clinical tool man, I can't even imagine practicing without this thing. No. And to that point, I will say that I can't think of— I mean, in terms of age might be the only differentiating factor with this, but I can't think of a single individual who couldn't benefit from getting that 65-panel panel or even if they have one that's missing a couple of the markers and just running this thing through. See what comes up. You can't test for toxicity very well today. You know, hormone testing is even questionable quite honestly depending on who you talk with. Nutrient testing is expensive. It's tedious. It's kind of a pain, you know. And to get these predictions and if it comes out, fine, good, you're doing well, but you might come up with things that you just had no idea that you have going on. So, I can't even imagine somebody that wouldn’t benefit from just— Even if they don't know how to interpret it, just taking a look at it, seeing it. If something's going on, finding a practitioner. If you are practitioner, like I said, I can't imagine using this. I've already used it. I can't even imagine having a practice without this. It's a game changer for how you do things. Absolutely. I wanna ask you this one last question.

[0:55:00]    

    If there is anything and this could be about health, or life, or love, or happiness that you would like to impart to your listeners who probably know you extremely well, but hopefully have enjoyed hearing you on the other side of the table, is there anything you wanna impart? Any wisdom? Any tidbits?

Chris:    Oh my goodness. That’s a 64,000-dollar question, isn’t it? You know, the thing I've been thinking about a lot recently is actually what Tommy's fiancée, Elizabeth, said in her TED Talk. I’ll link to in the show notes. And that’s specializing in not specializing. I think it applies to the interdisciplinary team that we've put together to work on this software, but it also applies in health. Right? So, one of the main problems that we see amongst the athletes that we work with is they start hyperfocusing on one particular thing, right? They start specializing. And usually, that thing is actually— Sometimes it’s important like the amount of carbohydrate you’re eating and then sometimes I don’t think is so important like MTHFR. You get someone. It’s like “Oh, can you treat my MTHFR?” No, I can’t. No, I can’t. I want you to focus out like a million miles and then just look at the big picture. And I think I’ve been guilty of this in the past as well. And now, I realize that it’s the big picture that counts. And if I was to show you my last gut test, it looks like a mess. And I definitely have had some GI symptoms and I’m working on it, but I’m nowhere near as bad as I was in 2011. Like you combine the gut infections with having no meaning in your life; being sad, and single, and lonely; you know overtraining; too much racing; and a very high carbohydrate, low micronutrient diet, then you've got a recipe for disaster.

Bryan:    Sure. Yeah.

Chris:    But maybe you can get away with one or two of those things and still feel pretty awesome. So yeah, I mean that's my imparting wisdom.

Bryan:    Love it.

Chris:    To focus on the big picture.

Bryan:    Yeah. And not just with health, but family and— I love that, man. That’s really, really good. Well, listen, Chris, it was an honor to be able to host this myself. I appreciate you doing that. I hope you enjoyed being the guest. And most importantly, I hope all the listeners got something out of this. So, thanks so much.

Chris:    Thank you. It’s been a pleasure. Thank you.

Bryan:    Thanks, Chris.

[0:57:00]    End of Audio

blog comments powered by Disqus

Register for instant access to your FREE 15-page book, What We Eat


© 2018 nourishbalancethrive