I Risked Death by Cholesterol in the Name of Science and Productivity!
Or how I doubled my cholesterol with Keto and how I lowered it back with the Feldman protocol
What a journey.
I went keto two years ago. I have tested so many productivity techniques before and since then. Yet the ketogenic diet is still #1.
After a year, I faced a dilemma. I got a blood test and saw that I had lethal levels of cholesterol. It was a choice between keto and a normal lipid profile.
Then I remembered the greatest words of modern film history…
So, in the face of overwhelming odds, I’m left with only one option…
I’m gonna have to science the shit out of this.
Okay, okay — I let the internet do the science part. And it turns out my fear of cholesterol was overblown. But I still produced some nice data, with graphs and colorful tables.
I now talk about “death by cholesterol” tongue-in-cheek, but I truly feared for my life the first time I saw my lipid profile after going keto, before nerding out on this topic:
Long story short, science wins.
After tons of reading, video lectures, and more than a year of diet experiments, I got my lipid profile to look normal.
During my months of exploration, I’ve encountered two solutions to high cholesterol in keto. The first is to reduce saturated fat. I compiled the best resources I’ve encountered for this solution:
It turns out my problem was not the gene polymorphism present in 20%-30% of people that makes them respond to keto high in saturated fat with high LDL. This is shown by my 9 months of lower and lower saturated fat, with no decrease in LDL: the violet line, from tests 3–5 (5 is the one with the lowest saturated fat and the highest fiber, but with the highest LDL! See my hypothesis in the next section).
The other solution is the Feldman protocol (test 6). The simplest version requires eating 2x to 3x your usual calories for 3 days preceding the blood test, maintaining ketogenic macros (~70% fat). What a great solution. I get to eat a lot — in the name of science!
The protocol was created by Dave Feldman (@DaveKeto). Dave is the Tony Stark of the keto-cholesterol world. After selling his software company, he has applied his engineering prowess and startup hustle to this problem that more and more people are encountering, with the growing popularity of Keto.
Watch this video if you want to know both the Why and the How of the Feldman protocol:
If 30 minutes is too much for you, and you just want the How, here’s what you should read:
If you are lean and athletic, and your cholesterol shot up after doing keto, you should also read this:
To understand the mechanism behind the protocol, I found it essential to understand how the body distributes energy through the following channels:
- Triglycerides via chylomicrons
- Triglycerides via low-density lipoproteins
- Glucose
The best resource I’ve encountered to understand these is this comic series, also by Dave:
I’m eagerly awaiting the 3rd part of the trilogy.
One Tip and Lots of Pics
My bias is to seek the most efficient version of a solution: the one which requires the least time, least effort, least complexity and least cost. Like Feldman, I was trained as an engineer, but I’m a 3rd world engineer.
For instance, my solution for doing keto in the Philippines is to simply replace each cup of rice, pasta or bread with 2 tbsp of coconut oil, and avoid all sugar:
Although it was fun eating lots of fatty food to lower my cholesterol (in the name of science!), it costs time and money. If I were to do it again, I’d simply do this:
- Buy a tub of peanut butter (select the one with the lowest carbs)
- Eat the usual keto meals but add 3–4 tbsp of peanut butter 4x a day. Scoop a mound for each tablespoon, not a flat pool.
I recorded all of my meals with the MyFitnessPal app during the 3 days prior to my blood test. And took pictures of almost all of them. The biggest bang for the buck in terms of high-fat calories — and the most delicious — is undoubtedly peanut butter.
Here’s what I ate in those three days (in the name of science!), with total calories and macro % breakdown:
4 parting thoughts for @DaveKeto, other keto hyper-responders, and N=1 experimenters
1 The Feldman protocol is effective for getting your doctor not to prescribe statins, as shown by the 100+ experimenters as of the time of writing. Yet, the protocol is not really a solution, is it? It is deeper than that. It questions the problem. It is a demonstration that the rules currently used (eg, if high-risk profile and LDL-C>190, then prescribe statins) may be too simplistic. What do you think?
2Is it correct that LDL-C is not correlated with heart disease but LDL-P is? Hyper-responders have elevated LDL-P in a ketogenic diet, yet have great scores in more advanced indicators like Remnant Cholesterol and Atherogenic Index of Plasma. Is this simply a case of the old analytics adage that “all data in aggregate is essentially crap”?
3 If using multiple signals (eg, 3 for Remnant Cholesterol) is a better predictor than using just one signal (eg, LDL-P), why not just give the entire Framingham data to data scientists for them to create a model that uses all the variables to predict risk of heart disease? I imagine that the reason for using low-dimensional algorithms (eg, if high-risk profile and LDL-C>190, then prescribe statins) is for ease of use. But if all doctors have powerful computers in their pockets, as is the case now, why still use clunky rules?
4 I’m curious why the LDL-C of my test 5 is so high. This is the test after my lowest saturated fat and my highest fiber diet.
My hypotheses:
- Prior to test 5, I have been lifting for 4 months (5x5 squats, 5x1 deadlift, etc, 3x a week, increasing every session), in keto and sometimes in the middle of a 24hr fast. Presumably my chylomicrons, blood glucose and liver glycogen are low, so my liver has to transport more triglycerides, thus shipping more LDLs.
- I doubled by Vitamin D3 supplementation starting a few weeks prior to test 5. I found this paper later on: Influence of vitamin D supplementation on plasma lipid profiles: A meta-analysis of randomized controlled trials “In our meta-analysis of randomized controlled trials, vitamin D supplementation provided a statistically significant increase in LDL-C (3.23 mg/dl).” If LDL is a carrier of fat-soluble vitamins, aside from triglycerides and cholesterol, then the liver would produce more LDLs to carry those vitamins.
How do these hypotheses sound? How would you design experiments to to check their validity? How would you design experiments to measure their influence on LDL compared to the Feldman protocol? How would you design experiments to figure out how many days prior to the blood test they would have still have influence (the same way Dave discovered that fat intake has influence just 3 days prior to the blood test)?
Anyhow, what a relief!
P.S. — A little disclaimer… hahaha
