While there is still more analysis to collect and unpack, here are the preliminary results from the Ketofest Cholesterol Experiment.
Did it meet the hypothesis? You be the judge:
I can’t possibly express enough appreciation for PTS Diagnostics, the great 2 Keto Dudes and the truly generous volunteers who took this journey through the experiment with me. Thank you so much for your participation!
Thank you for sharing the data
Lots of learing from your studies
Thank you, Shashikant — I feel as though I’m barely cracking the book open myself.
Is there statistical significance??
Yes…
The p-value is .01426. The result is significant at p < .05.
… what Duane said. 😀
Did the folks in the test group track their macros and will that data be available too?
More than macros, the test group did 100% photo logging of all their food and drink over the 6 days! I’m looking forward to Dave’s analysis of this data.
I strongly requested both photos and logs. But to be sure, I don’t think there was 100% compliance — and don’t be hard on my subjects, they don’t have as much practice as I do.
For instance, both Carl and Richard (of the 2 Keto Dudes) likewise joined the experiment but conceded they didn’t get many photos, so I’ll be relying more on the logs with them.
This doesn’t bother me as much given both Carl and Richard as well as most of the subjects did a 3 day fast for the Low-Calorie phase. So in a sense, *any* amount of food over the High-Calorie phase would have been larger.
As an aside, this is why I want my first *formal* experiment to be a ward study. I’d like people to eat from meal plans with food prepared for them so as to be 100% certain of quantities, nutrients, etc.
Congratulations on this incredible experiment. Well done everyone! Really excited to see what the future holds for our understanding of cholesterol.
Thank you, Joanie — I likewise share your enthusiasm. 🙂
Is there data to show why a few people didn’t go down? Does obesity play a role? Thanks.
That’s probably a very difficult question to answer. While I believe the Inversion Pattern accounts for 30-50% of the cholesterol markers, that leaves the other 50-70% as multifactorial. Obviously, a large amount is long term energy use, exercise, and likely glycogen store status given my own experiments from this blog — but there’s still much more to discover.
Regardless, those that did go up only went up 1-2%. Whereas those that went down 5-38%. So it’s very possible it is a matter of standard deviation as well.
What about eating windows?
Time restricted feeding seems do able and gets results quick!
Eric
Eating once a day keto
I haven’t done a lot of experimentation on this as of yet. For myself, I try to always hit as close to 14 hours fasting from the last meal of the night before to the blood test of the morning after. With regard to participants in this experiment, I just required 12 hours specifically (or more, of course).
Hi Dave.
Further thoughts on your results.
I think there are two things that can explain what you are seeing.
1. Metabolic channeling.
2. The Randle Cycle.
There are a number of questions that come up in looking at your results, ie:
1. Why do hyperresponders exist.
2. What explains your lipoprotein dynamics.
First your “Feldman Protocol”.
Metabolic channeling is a real phenomenon and indicates it may be useful to think of two
pools of LDL, pool a consisting primarily of LDL I/II gestated through, primarily,
VLDL2 and pool b consisting primarily of LDL III/IV/V gestated through, primarily, VLDL1.
Pool a has a turnover of 2 days, pool b has a turnover of 5 days, therefore
your preferred period, approximately 3 days represents, possibly, the mean
turnover period of a mix of these two pools and explains the agility of the
cholesterol response to a change in demand.
Second the tight correlation between dietary fat and LDL-C cholesterol test results.
One way of expressing this correlation is to say that changes in dietary fat are
reflected in changes in LDL-C but opposite, ie delta fat + delta LDL-C = 0. This
is expressing a conservation relation and what is conserved is lipid energy (LE) so
what the relation may be saying is that lipid energy is represented by two systems,
variably partitioned, but the total is relatively well conserved over the relatively short period.
So we might say Fat + LDL-C = LE, since you are hard keto mostly this might make sense. There is
obviously an issue when saying conserved as your are continuously consuming lipid energy
irrespective of its partitioning. However it might make sense in your representation as the period
is short and you are looking at two different measures ie fat weight is not the same as LDL-C weight,
and the characteristics of supply and demand are different (supply is more pulsatile with demand
possibly steadier) so you may have the illusion of conservation, but perhaps not far off. However
the correlation is obtained it seems to be robust.
Alternatively when the correlation is lessened it might be that Fat + LDL-C = FE’
where FE’ < FE, the difference being represented by glucose energy that is coming
into the system via the switch in the Randle Cycle. This makes sense as glucose oxidation
means less demand for dietary fat and less demand for LDL-C since there is less overall energy
being shared between the two lipid partitions. At this low level of glucose oxidation following
relatively hard keto you may not see a large rise in trigs as your glycogen stores may be being
rebuilt (primarily in liver and muscle) or liver is not putting a large net amount into circulation as
triglycerides for other metabolic reasons. This may change with time, but I suspect it may be robust,
I hope so.
The big question, however, is why do hyperresponders exist. I think this is also explained by 3 things.
1. Metabolic channeling.
2. Triglyceride levels.
3. LDL pool dynamics, particularly pool size, size variability within a given population and catabolic rates.
As I explained before at very low triglyceride levels, generally, LDL pool size is low, catabolic rates are
very high and variability is low. I would therefore expect to see few hyperresponders here. They obviously
can exist but I would expect they would have some other explanation (high LDL can be explained by low catabolic rates caused by LDLR genetic mutations as per familial hypercholesterolemics for example).
As triglycerides increase LDL pool size increases, variability increases and catabolic rates decline.
The combination could explain hyperresponders, if so they might exist within a relatively narrow range of triglycerides where LDL pools size is highest, catabolic rates lowest and variability may be very large. If in addition LDL is coming out body wide then this could be supercharged. This could explain why hyperresponders exist and why they are relatively rare (5-20% of the LCHF population).
As triglycerides increase further catabolic rates increase again and LDL pool size starts to drop. It also
changes character to atherogenic. Therefore if this discussion is reasonable it seems hyperresponders are
less likely at higher trig levels and so are likely to be relatively healthy.
Anyway Dave I want to thank you very much for your efforts they have allowed me to frame my thoughts regarding my own situation. I will be very interested in your ultimate conclusions.
Thanks again.
References:
Metabolic channeling.
http://atvb.ahajournals.org/content/17/12/3542.long
Randle Cycle
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2739696/
Tim–
Sorry for the late reply. This is one of those awesome epic comments that I need to read through while also following up on the two links you shared with Metabolic Channeling and the Randle Cycle. I’m extremely interested in not only working through this different paradigm but also seeing if I can devise a new test for it.
Are you collecting genetic markers from ppl also ?
It occurs to me that most biohackers will have their dna and it might be more than just apoe that might explain variances.
I’m apoe3/4 like you but I also have
the butylcholinestrinase K variant
rs1803274(A;G) ( assn with high ldl)
rs2075650(A;G) (ldl-chol / alz)
cyp3a5 non expressors (aka gs 155)
rs2075650(A;G) 2 2x higher Alzheimer’s risk
rs4420638(A;G) 2 ~3x increased Alzheimer’s risk; 1.4x increased heart disease risk ; increased LDL
rs17228212(C;T) 1.26x increased risk for heart disease
rs1260326(T;T) slightly higher risk for gout
rs4939883(C;T) associated with higher HDL cholesterol
rs3843763(C;T) Slightly lower HDL (‘Good’) Cholesterol.
rs1883025(A;A)
rs2650000(T;T)
rs471364(A;G)
rs2839619(A;A)
rs4253772(C;T)
rs693(T;T) Â Bad elevated lipids
Not saying its all genetics, could also be microbiome but only one way to find out
I certainly agree that genetics plays a role. How much, that’s still up in the air.
Here are my corresponding SNPs. (We have a few in common…)
rs1803274(A;G)
rs2075650(A;G)
cyp3a5 gs 155
rs2075650(A;G)
rs4420638(C;C)
rs17228212(T;T)
rs1260326(C;C)
rs4939883(C;C)
rs3843763(C;T)
rs1883025(A;G)
rs2650000(G;T)
rs471364???
rs2839619(A;A)
rs4253772(C;C)
rs693(C;T)
Thanks Dave. Great information. I am a recovered T2D through ketogenic diet and intermittent fasting. Had an unexpected high LDL-c an my last check and Doc recommended a station. I demanded a CAC test – and refused the statin with A1C of 5.1 and CAC score of 0!
I made a mistake on the last blood draw. I fasted for three days immediately prior to giving the sample! After reviewing your results I will eat LCHF prior to the next test. Want to bet on the over-under on the next LDL-C?
Kudos for being such a strong advocate for your health, and congrats on the CAC 0 and great A1C!
As a basis for the over-under guess, what was your 3-day fast LDL-C from last time? (And TC, TG, HDL-C, for context)
was weight track during the experiment ? any weight loss even it is insignificant ?
maybe a dxa scan could be included in the data ?
Weight was not tracked, nor was there a dexa scan, but that might have been interesting.