Paladin joins the PREDICT 2 clinical trial on nutrition as a test subject. Find out why. What is there to learn?

“Medicine breaks down into two major strategies. You have prevention, which is highly successful, but does not pay well. It is tough to prove that something you prevented from happening at all was bound to occur, since symptoms of what you prevented never shows up. There is less glory in prevention. In fact, prevent long and well enough, people forget about the disease you were preventing in the first place, and will start to wonder if you just made it all up! But, with prevention, the disease is much less established or never establishes because of your prevention—so often faster and easier to treat. Fast, easy treatments tend to be cheap. On the other hand, you have treatment after the fact. This happens only after the disease shows up and establishes well enough to cause indisputable symptoms. More established disease, actively causing symptoms, is more difficult to treat. Disease, like atherosclerosis, that took years to develop may take years to treat. But there is glory in treatment, because the disease is established and causing huge, obvious problems. If you can make those problems go away, or reduce them, you’re a hero. To some extent, right fully so, because reversing that kind of damage is, indeed, difficult. So there is huge money in treatment, because it takes a lot of treatment, often long treatment, to chase off fully symptomatic, clear cut disease. It is always better for your patient to prevent them from ever having those problems to begin with. Your patients are human and human nature is to not to do much until something is an obvious problem though—it will always take more to convince your patients to do something for prevention than to treat a problem they already clearly have. Also, the system is kind of geared to reward treatment more than prevention, even if that’s kind of backwards. But you should still prevent whenever you can. It’s better medicine.”
—One of Paladin’s primary care professors back in medical school, paraphrased.

“Medicine is a terrible business model. If you make a great soda or a nice car, people want to buy more soda or another car. But as a doctor, when you have done the greatest good for the patient, the patient no longer needs you. To be the best doctor, you must eliminate demand for your ‘product’—and the world will be better off for it.”
—Paladin

Cover Art: Rue Mouffetard, photo by Virginia Birmingham, Paris Through My Lens Blog https://paristhroughmylens.blogspot.com/search?q=rue+mouffetard+night

We all know that nutrition is important for health. Proper nutrition can prevent a lot of disease. The challenge is that no one seems to have a real good idea what proper nutrition is. Just look at the multi-billion dollar industry of diets. Every year, it seems, at least two more pop into existence as major fads that a lot of people try. Some succeed—they become the testimonials in the ads. Many more succeed for a bit, but then just cannot sustain whatever the diet is. Or meet their goal, then go back to old patterns of eating and the cycle starts anew. This is good if you own companies that sell diets. Given the prevalence and incidence in obesity and diseases of an industrialized world (like type 2 diabetes, heart disease, some cancers, certain liver diseases etc.), we have not found the One Diet To Rule Them All which works for everyone. Despite all that money, and all those efforts.

Now, in part, that is because nutrition studies are tough to do. There is also clearly a lot about nutrition we just don’t know. But, we know that certain patterns of eating are much more likely to lead to the diseases of poor diet above. A better diet (and exercise) can significantly reduce your risk of a bunch of nasty things.

They don’t eliminate that risk. But they do reduce the chances bad things will happen to you. With your diet, you can either buy more tickets to the lottery that determines if you get some of these diseases, or less. You should really prefer fewer tickets for some of these “bad” lotteries…

Further, diet and exercise, as prevention, are far less expensive than some forms of medication to treat those diseases listed above should you actually get those diseases. So you can pay the cost to look and feel boss with diet and exercise—or you can wind up taking a bunch of pills. Which come with side effects. A diet you can eat safely, happily, and consistently, coupled with exercise you enjoy doing? The only side effects of that are things like “greater risk of contentment.”

“Alright doc, you’ve convinced me. How do I eat and exercise better?”

Good question.

I don’t have a very good idea. They didn’t cover it much in medical school (because good empirical data is lacking—nutrition studies are tough to do). Our professors told us that half of what they would teach us in medical school would turn out to be wrong over the course of our careers—but we just don’t know yet which half. Well, I can tell you some of the nutrition stuff I was taught looks increasingly like it was in the “wrong” half.

I have tried a bunch of stuff with my own diet to see if any of it works for me. Multiple “fad” diets. Some worked better than others. But all, uniformly, working only to a plateau. That said, I have been able to ride successively lower plateaus down to lose weight and body fat percentage. But, as I am sure many of you nod along knowingly, that wall was eventually hit or the diet restrictions (most break down into “good food” and “bad food” lists) left me bored and I left the diet plan for good.

Take the keto diet, for instance. Keto has worked very well for a couple people I know. They all lost weight with it—quite rapidly. They all felt fine after the famous “keto flu” at the beginning. But not all could not sustain the diet long term. Some of them came off successfully, staying pretty thin. Others rebounded some of the weight. I tried keto twice—it never even worked for the body fat loss!

What has worked best for me, as mentioned last Ramble, is from “Fat Loss Forever.” Their whole point is pretty much that “good food/bad food” lists don’t work in the long term, and they note studies suggesting some pretty broad individual variation in results. In fact, you can look at pretty much any given diet study. The headline in the newspaper will be “eating less of X may help you lose inches”! That will be true, if the study was properly powered around the question “does eating less of X lead to weight loss measured by waist circumference measurements?” This will be true for the aggregate, since you will have a group of patients eating more X and some eating less or none of X, and the study will report their average results and difference between the groups. But get into the actual data and you WILL find some individual patients who ate a lot of X, but STILL lost some inches on their waist. You will also find some who ate less X, but either did not lose or may have even gained inches around their waist.

What is true for a group in a nutrition study, which is what the headline is reporting, may or may not be true for you individually. If you are more like the group, you have a better chance of showing those same results. But only a chance. Not a guarantee.

So if best possible diet and way to eat is an individual thing, that’s actually a bit of a problem. A lot of guess work and carefully controlling what you eat and when, adding, subtracting, controlling as many variables as possible to determine what foods, meal times, and in what combinations work best for you. “Fat Loss Forever” advocates calculating a calorie deficit (for weight loss), surplus (for weight gain) or calories balanced to your activity level to maintain your healthy weight. This is your calorie budget. You then fill those calories by paying attention to the “macros” (protein, fat, and carb content, with a “shortcut” by figuring out your protein first and going from there, making sure to get enough fiber). Eating this way will tend to result in the “clean” eating of many other diets–with some candy bars you can sneak if you are willing to pay the macro/calorie toll for them out of your calorie budget. You quickly learn to identify and moderate calorie “dense” junk foods, since you can quickly blow out your calorie budget, but not ever feel full. This is a good place to start iterating around at least. For me, the method has been sustainable—but sustainable is not necessarily optimal.

I was curious to see if I could do better.

Last Ramble, I mentioned the PREDICT study from Harvard, Oxford, King’s College, Stanford and Zoe. PREDICT was accomplished with a strictly controlled diet and laboratory testing regimen, for probably the best look at individual response to different meal compositions over time. Inter-individual variation was much higher than intra-individual variation—meaning what “worked” in the blood tests for an individual would get predictable, but guessing what would “work” from person to person was less so.

Said another way, a keto diet might be really effective for some of my friends, but might not work for me. But you could tell quickly a keto diet was not working for me. Vice versa holds true. Just because one food looked great for my blood test results, it won’t necessarily predict what your results would be from eating that same exact food.

I doubt that surprises anyone reading this.

There were two major findings from PREDICT that caught my attention (and you can catch the interim data here). First, knowing just the macros in your food (the “Fat Loss Forever” starting point of protein, carbs, and fat), doesn’t get you the whole way. In fact, they estimated it accounts for only ~1/3rd of the prediction accuracy for what your blood sugar (glucose) and fat (triglycerides) will look like after a meal. Put another way, the macros can be the same between two different kinds of food you are eating. But one might spike your glucose and triglycerides in a big, long, unhealthy spike, while the other food would have a healthier quick up, quick down and lower total spike in glucose and triglycerides. I did not think the macros would be perfect, based on personal experience. I’m sure you can think of similar foods that make you -feel- different after you eat them! But I expected better prediction success than just 1/3rd. Second, the twins they included on this study had mind blowing results. The control data for their twins was their height. If you knew the height of one twin, you unsurprisingly pretty much knew the height of the other. We all know twins are nearly always about the same height. However, despite identical genetics, twins’ lab responses to the same meal, at the same time hardly correlated at all! The key takeaway is that individual variation (in their environment, exercise habits, gut microbes etc.) vastly exceeds the influence of even genetics in how you respond to food. Genetics matters (also about 1/3rd)–but individual level differences matter more than we may suspect. We don’t have a good way to predict those individual level differences. Yet. That’s why they ran PREDICT.

At the end of PREDICT, they had an algorithm which, if it knew a few data points about you, could predict with ~73% accuracy what your post-eating labs would look like for a given meal. That’s not bad. Better than a coin toss. But they might have room to improve.

So they opened PREDICT 2, which lets you participate from wherever you are if you meet the inclusion criteria. I believe enough in solving nutrition as a way to improve health outcomes for everyone that I volunteered to participate in PREDICT 2 as a study subject. Before the many, many accolades for my courageous altruism come pouring in, I also get some data (for free!) on my own reactions to set meals as trial and error to help me make better choices in what I am eating.

For free.

Can’t stress that enough.

So, I am now done with the 10 day PREDICT 2 study and awaiting their final report on my data. All it took was a couple questionnaires for eligibility, and a call to go over the informed consent. Then they ship a big box of all the stuff you will need for the 10 days you are on the study. This is a bunch of standardized meals (as muffins, glucose drinks, and milkshakes) that you had to eat for breakfast or lunch on specific days. They collect a DNA sample of you and your gut microbiome. I also had to wear an activity meter 24/7 (except showers) and a Freestyle Libre Pro continuous glucose monitor for 24/7 through the full 10 days. While the Freestyle Libre was remarkably easy and intuitive to use, it did cost me a week of jiu jitsu (which was the biggest drawback to participation) as contact sports can dislodge it. So plan appropriately for your 10 day window if you do any contact sports regularly. I had to fast for 2-4 hours after eating their set meals (nothing but plain tea, coffee or still water), and answer some questions on how full I was feeling and how alert I was during the fasting period. They also had some questions on how I was feeling throughout the entire day that pinged me before bed. All of this was really easy because you will download their app, and it will tell you what to do, and when, and keep a team of nutritionists available pretty much around the clock if you have questions on any of it.

That app was a really great way to run and manage participation in a clinical study, and the app runs really well. If you do MyFitnessPal or something similar, it’s pretty close to those.

The first two days you also had to collect blood samples via finger stick on some cards to get lab data on some very specific study breakfast/lunch combos. After that, it was smooth sailing. When you don’t have a study meal to eat, you could eat whatever you wanted/normally ate. You just had to measure the ingredients and log everything. Since I already use MyFitness Pal to track calories and macros religiously, this was no big deal, but did add a bit of time to every meal I ate. The last two days of the study are just whatever and whenever you want to eat— so you’ll get a good measure of what your “usual” diet is doing to your blood glucose! A few days after the 10 day study period, I had to go to a Quest Diagnostics lab center to get the last set of blood draws done for study labs. An appointment may be a good idea, but the one down the street from me is walk in only. I was in and out in less than 10 minutes once they had the paperwork from the study sorted out.

Overall, a really easy study to participate in.

And I learned a few things already doing it. The LibrePro they give you comes with a reader, so you can check your own blood glucose whenever you get bored. Now, the graph on the reader is not terribly detailed (the computer download the PREDICT 2 study team will get off it will be), but you can tell when, how high, and how long your blood glucose was going after anything you ate or drank. That said, the continuous glucose monitor was designed for diabetics to monitor their blood sugar and calculate how much insulin they might need, not normal healthy people like me. So I view it more as -directionally- accurate than necessarily that my glucose was EXACTLY x mg/dL at time y. But I could tell when my glucose was higher than my baseline and how long it was staying high after a meal.

A normal glucose response after a meal should see a spike in about 30 minutes to an hour after you are done eating. How high it goes is influenced by what’s in it. Pure sugar will lead to a bigger, earlier spike. More complex carbohydrates, like pasta, will spike glucose, but perhaps not as high and usually with a little more delay than just pure sugar. A meal that is mostly fat or protein should result in a much lower spike that develops a little later, as your body has to digest these a little more.

A normal, healthy response to glucose is a quick spike up, and then quick down, like the blue/grey line at bottom here:

A less healthy response is a larger spike and/or a longer taper down, like red/orange line above.

Basically, the area under the spike is the amount of glucose in your bloodstream over that time period and should correlate to how much insulin is floating around in your blood stream trying to get that glucose level back down to baseline. A high spike with a long taper is less healthy, because more insulin is around for longer, and insulin will tell your liver (and muscles to an extent) to store that glucose floating around as glycogen, and fat cells to store it as newly formed fat. Some of the thinking about how type 2 diabetes starts is this higher, prolonged insulin becomes like the boy who cried wolf—the body stops listening to the insulin, requiring more and more of it to have an effect, which we call insulin resistance. Eventually, you can’t make enough insulin to get the body to pay attention to it properly and clear the high blood glucose, and diabetes (high, unregulated blood glucose) results. So a high spike with long taper, and more area under the glucose curve, is bad.

I was shocked by my first couple “own choice” meals to see a moderate elevation in my blood glucose that lasted for hours after the meal! There were only a couple items in common between those meals, and I immediately started a “study within the study.” I isolated out those items to find out which one was causing that pattern when my “own meals” and snacks would allow me within the PREDICT 2 study. I quickly found the culprit:

The whey protein shake I had been using to up my protein macros. I had been having a couple of those per day for a few months now. The chocolate version with some almond milk (low cal option) or 2% milk (higher cal option) — it’s so good. Meals where I had a protein shake had the “bad pattern” glucose tracing after them. Meals where I did not have a protein shake had the “good pattern” glucose tracing.

But–the whey itself was not to blame. Lots of that in milk, and when I drank low fat milk to ~equivalent protein content with the whey shake, I did not see the “bad” pattern in the blood glucose like I did with the shake. Likewise, it wasn’t just general protein—using egg whites (mostly albumin) to equivalent protein content did not create the bad pattern.

So I looked at the ingredients list on the whey shake, and was surprised. I knew it had a few carbs in it from MyFitnessPal and was assuming they snuck in some sugar for the chocolate flavor. Nope. Most of the sweetness was coming from sucralose and acesulfame potassium—two artificial sweeteners.

Having isolated out and eliminating everything but those, I had a lead candidate now for that prolonged, almost insulin-resistant like pattern I was seeing in my blood glucose readings. There are rat/mouse studies that show that acesulfame, for instance, in associated with higher insulin. But I am not aware of studies in humans showing that yet.

I don’t have just pure sucralose or acesulfame potassium though. But I can approximate them. So to isolate the artificial sweeteners in the protein shake, I ate the same meal again (a lunch)–only this time, I took a dance with my old archnemesis:

Cherry Coke Zero uses acesulfame and aspartame as its artificial sweeteners.

Sure enough, adding that Cherry Coke Zero brought the “bad” pattern back, just like having the whey protein shake with acesulfame.

For me, at least, these artificial sweeteners are catastrophically bad, and have likely been sabotaging me for a long, long time. They don’t cause a huge spike in glucose. But the glucose goes high and -stays- moderately elevated over my baseline for 4-6 hours after a meal. No bueno.

But again, that may just be me. May not hold true for you. There are some studies out there associating artificial sweeteners with weight gain though, and this potentially could be a mechanism. Hopefully, whatever they spot in my lab work, from DNA to microbiome, to other factors, sticks out and helps identify others like me who may have this bad response to these artificial sweeteners.

Interestingly, it may be ALL artificial sweeteners. When I told the study team what I had done with my “own choice” meals and drinks to isolate this long, moderate glucose pattern I had seen with acesulfame/aspartame/sucralose, the nutritionist asked if there were any other artificial sweeteners I could test. So I found a drink with stevia and erythritol. Stevia has been controversial in the literature, with some rat/mouse studies suggesting higher glucose and/or insulin levels after eating stevia. Erythritol was tested in humans in a Japanese study, but only by itself—not with any food or other glucose taken along with it. In that study, erythritol taken by itself does not elevate blood glucose or insulin. Despite that, erythritol was recently associated with weight gain in a study out of Cornell.

Unlike the Japanese study, I took it real world style—with food.

When I did (along with stevia), I got the exact same long, long, long, moderate glucose elevation as with the other artificial sweeteners. Several of these have very different chemical structures, but they all hit sweet taste receptors. That must be the biological start of this effect, but will take more science to tease apart.

Regardless, it looks like artificial sweeteners of all sorts are a bad idea for me. Specifically, they look like a mild-moderate insulin resistance effect. The duration of that effect —4-6 hours—is equally surprising. Again, my results with artificial sweeteners may or may not apply to you. You may be genetically or environmentally blessed where some or all of them do not have this effect on you. Hopefully, what my data (and others like me) in the PREDICT 2 final analysis shows is something that the algorithm that results can use to predict if YOU are likely to get this result too.

For my own part, looks like I am done with all artificial sweeteners for the foreseeable future.

One other interesting finding. I had a piece of birthday cake during the study period (one of my kids’ birthdays, so couldn’t exactly say no—it was glorious cake though). As you can imagine, that was a huge spike that took a little longer to come down all the way. Too much birthday cake is not healthy—surprising no one. One other meal looked just like that cake on a post meal blood glucose tracing though. That did surprise me. See if you can guess which one:

Is it:
A) Spaghetti (3 oz) with ground beef in the marinara, onion, and roasted zucchini and yellow squash

B) Homebaked rye and barley bread with 1.5 tablespoons of honey (in the entire loaf—maybe a quarter table spoon in the portion I had)

C) A quarter cup of steel cut oats with a half cup of 2% milk and a half cup of blueberries

or D) Turkey chili with about a quarter cup each of black beans, kidney beans, great northern beans, diced tomato, and corn (1 bowl)

The other three, despite being carb heavy too, showed only a moderate to high rise in blood glucose (as you would expect), which went down in the normal, expected window. The “as bad as cake meal” up there was my single highest glucose spike on the study. In fact, higher than the cake. Higher even than the 75 gram glucose drink I had as an official PREDICT 2 breakfast. The spike was around as long, or a little longer, than the cake’s. A surprisingly unhealthy, “worse than birthday cake” pattern overall!

So what’s the surprise “bad for me” (not sure if bad for you yet meal)?

D). That turkey chili.

But wait–there’s more. I had the turkey chili again two nights later. One of the other interesting PREDICT 1 study findings was that order of meals may matter, and activity levels. You might respond slightly differently to the same food with changes in what you ate prior to it, or what you were doing.

Sure enough, the second time around, the turkey chili behaved itself. A high spike, but nowhere near what it was the first night, and down within the expected normal window. That night, it seems, the turkey chili was much safer! There were differences in my breakfast and lunch those days, and maybe that’s the culprit. Unfortunately, I was out of days on PREDICT 2 (and chili) to isolate and find out which combination was blowing the chili up.

I would have never guessed the “Turkey Chili Anomaly” prior to this study, and now I’m really curious what the more detailed PREDICT 2 data for me might suggest.

So stay tuned! I’ll follow up on the larger study report when they release that. It will be a little while, and I will plan on seeing how their recommendations go for me after the report as well, just as a little test for accuracy. I’ll report back here in a separate post what I find out…

Tagged Clinical Study, Diet, Keto, Low Carb, Low Fat, Macros, Optimal Nutrition, Other Posts, Performance, PREDICT 2, Prevention