Please consider supporting our newly launched Citizen Science Foundation and its central endeavor, the LMHR Measurement Project. Your contributions are making citizen science a reality – thank you!

Dave’s presentation to Stanford University

BREAKING: Our Paper on Low Carb, LDL Cholesterol, and the LMHR Phenotype is Now Available

Note from Dave: This is a guest post from Nick Norwitz on our just released LMHR paper. This post also has a live “Corrections / Clarifications” section (at bottom) which we’ll be updating as needed. We welcome and look forward to all constructive corrections and comments to our paper.

Four of my colleagues and I just published a paper in Current Developments in Nutrition on the Lean Mass Hyper-Responders (LMHR) phenotype.

Now, if you’re familiar with the phrase “LMHR” – or if you are an LMHR yourself – then the findings of this paper will be akin to the statement “water is wet.” Nevertheless, the majority of people, including those in the medical community, are not aware of this fascinating metabolic phenomenon. That’s not a slight against anyone; it is simply the state of affairs.

So, let’s change that. What is a “Lean Mass Hyper Responder?”

Lean Mass Hyper Responders, a History

In 2017, a software engineer, Dave Feldman, made a curious observation: the people who adopted carbohydrate restricted diets who typically exhibited the most pronounced increases of LDL cholesterol (so-called “bad cholesterol”) were often very lean and/or athletic.

But the elevations in LDL exhibited by these lean persons on low-carb diets had two peculiar features that set it apart from other forms of high LDL:

Extreme LDL Increases

First, the LDL increases were much larger than those typically associated with living an unhealthy lifestyle. When most doctors think about high LDL related to being unhealthy and eating a poor diet, they think about levels in the high 100s. But lean people on carbohydrate restricted diets were anecdotally observing LDL levels of 200, 300, 400, and even 500 mg/dL or more.

In fact, some LMHRs exhibit LDL levels as high as persons with homozygous familial hypercholesterolemia, a rare and devastating genetic condition (1 in 1,000,000) that likewise associates with very early heart disease.

Very High HDL and Very Low Triglycerides

Second, when lean people do see increases in LDL on a carbohydrate restricted diets, they tend to be accompanied by very high HDL (so-called “good cholesterol”) and very low triglycerides (TG), fat in the blood. This pattern of high HDL and low TG is exactly opposite the profile of “atherogenic dyslipidemia,” which is defined by low HDL and high TG, and is, at present, the predominant risk factor for cardiovascular disease (Libby, 2021).

Simply put, when lean people on low-carbohydrate diets saw increases in LDL they were quite often in the context of otherwise excellent metabolic health markers. Therefore, Dave created a set of three cut points that combine to define what would become the LMHR phenotype:

  1. LDL cholesterol ≥ 200 mg/dL
  2. HDL cholesterol ≥ 80 mg/dL
  3. TG cholesterol ≤ 70 mg/dL

Now, before moving on I want to emphasize that the definition of LMHR is based on these three thresholds and NOT on any measure of leanness or BMI. The word “lean” is in the title because the people who were anecdotally reporting high LDL, high HDL, and low TG tended to be lean.

In other words, the name “LMHR,” posed in 2017 was itself a sort of prediction: maybe leaner and metabolically healthier people are more likely to exhibit increases in LDL on a low-carb diet? The name is the hypothesis.

Study Suggests that LMHR Exist!

Being Leaner & Having Lower TG/HDL Predicts Larger LDL Increases on a Low-Carb Diet

It was a long time coming, but we finally put that hypothesis to the test in a scientific study.

In this new study, published in Current Developments in Nutrition, we collected survey data from people who had been low-carb for >2 years, who were not on statins, and who had lipid data from before they started their low-carb diet as well as recent lipid data from on their lowcarb diet.

Then, rather than massaging the numbers to conform to our hypothesis, we engaged in a “hypothesis-naïve exploratory analysis” in which we took all the data we had on respondents — including age, sex, BMI, and current and pre-low carb LDL, HDL, and TG levels — and asked a computer to tell us which factors were most strongly and reliably associated with increases in LDL after starting a low-carb diet.

The results were clear. No matter how we approached the question (be it multivariate linear regressions or hypothesis-naïve computer-generated decision trees [Supplemental Figure 3]) we found that having lower BMI and a lower TG/HDL ratio associated with larger increases in LDL.

Hypothesis-naïve exploratory analysis finds that having lower BMI and a lower TG/HDL ratio associates with larger increases in LDL on a low-carbohydrate diet.

The relationship can be clearly seen in the bar graph below. The further you go to the left, the lower the BMI. The further you go to the back, the lower the pre-diet TG/HDL ratio. And the height of the bar is the median increase in LDL.

Picking out the LMHR

After establishing that those who are leaner with lower TG/HDL ratios exhibited larger increases in LDL with carbohydrate restriction, it made sense to try to separate the true LMHR (those who met all three cut-points) from the larger cohort and see how different they really were…

Of the 597 participants that met the inclusion criteria, 112 were bona fide LMHR (which is a lot, considering many people don’t believe LMHR exist). And, true to their name, they were Lean!

The average BMI of a LMHR was 21.9, as compared to 24.5 for the rest of the low-carb sample in this study (between group p = 1.04 x 10-11). Furthermore, LMHR exhibited higher LDL, higher HDL, and lower TG, with mean values of 316, 99, and 47 mg/dL respectively.

And, importantly, LMHR did not differ in terms of their pre-diet LDL when compared to the non-LMHR population. In fact, median pre-diet LDL was 135 mg/dL in non-LMHR and 135 mg/dL in LMHR. Exactly the same! This finding is consistent with the notion that, unlike familial hypercholesterolemia, the cause of LDL increases among LMHR is unlikely to be genetic (see more below).

In the Lean Mass Hyper-Responders (LMHR) paper, the average LDL, HDL, and TG of LMHR were 316, 99, and 47 mg/dL on low-carb, respectively. This was despite normal pre-diet LDL levels.

A LMHR Case Report Shows the Phenotype is Partially Reversible

Now, you’ve probably sensed a lot of enthusiasm from me, but don’t mistake intellectual excitement about a fascinating observation for a suggestion that high LDL levels in LMHR are benign.

Setting my own hypotheses aside, we do not yet know if the risk associated with high LDL is any different in the context of LMHR as compared to any other context. And most experts would agree that high LDL is dangerous, regardless of cause.

That said, this very question — is high LDL harmful in LMHR? — is currently being assessed in a prospective study (data from which are expected to drop in 2023). And I will be vocal about the data when they emerge, whatever they say.

Nevertheless, for the time being, many or most LMHR patients and their doctors are concerned about their high LDL. That said, many of those same people find a low-carb way of life to be tremendously beneficial for their various metabolic disorders. This begs the question, can you “fix” the LDL problem (perceived or true) through lifestyle? The answer, yes — at least partially.

As part of this study, we also wrote up a case series of five patients who were LMHR or nearLMHR. These patients all exhibited extraordinary increases in LDL upon starting a ketogenic diet. And, importantly, all were tested for genetic mutations associated with high LDL and all tested negative, supporting the notion that being a LMHR is not a genetic condition but a metabolic phenomenon.

One patient saw his LDL increase from 116 to 665 mg/dL (no surprise, he was the leanest.)

All of the patients refused, or were intolerant of, statins and instead opted to reintroduce a moderate amount of carbohydrate, ~50 – 100 grams, in order to transition from a very lowcarb ketogenic diet to a diet that was still low-carb (<130 grams net carbs per day).

Impressively, all participants saw their LDL drop by at least 100 mg/dL, with larger drops occurring on those with high levels. The patient who saw his LDL increase to 665 mg/dL exhibited a 480 mg/dL drop in LDL by doing nothing more than adding about a small, sweet potato’s worth of carbs per day.

Stop and think about that for a second. In this context, a sweet potato per day could drop LDL by almost 500 mg/dL!

SubjectLipidspre-VLCDVLCDLCDLDLc Decrease
IA*Total Chol214797294-480
MI*Total Chol209698497-223
ROTotal Chol197311180-124
NMTotal Chol179387272-122
ANTotal Chol218423318-100

Future Directions

This is only step one, putting the LMHR phenomenon on the map. This paper suggests that LMHR are real and, if I do say so myself, really interesting!

In my opinion, no true student of health and/or medicine can observe this phenomenon and not be intrigued.

But what this paper does not do is explain the “how.” It likewise can’t evaluate the risk. Those are the subjects of upcoming projects.

We are working on formulating an official “Lipid Energy Model” manuscript, explaining the potential mechanisms at work behind the findings of this paper. And, a LMHR responder study also recently launched out of ULCA, courtesy of Dave’s efforts, that will track plaque progression in the coronary arteries of 100 LMHRs. This paper was just the first domino…

Additional Notes

This paper describes a phenomenon. It does not explain the mechanism nor comment on risk.

Saturated fat intake was not measured; however, it seems highly unlikely variations in saturated fat intake can explain the findings as this would assume that, across the study sample of 597 people, lean people and those with good metabolic health preferentially and reliably consumed more saturated fat.

So, for those who want to protest that “LMHR just ate more saturated fat” (and we get this a lot), I’d want to ask: what do you think about the flipside of such an assertion. Is it reasonable to assume that eating saturated fat is strongly, and possibly causally, associated with being leaner and with better metabolic health markers? (For more on this, see Supplemental Figure 5 from the paper.)

The phenomenon is likely metabolic, not genetic. This is supported by at least three lines of evidence in this study:

  • LMHR have normal pre-low carb LDL levels. In fact, in the study, pre-diet LDL levels were exactly the same between LMHR and non-LMHR.
  • It’s been anecdotally observed that LMHR who gain weight exhibit drops in LDL, despite no change in their genetics.
  • Most importantly, genetic testing in the six patients in the case series were negative. Genetic testing performed on other subjects not reported in this study have also been negative.

Corrections / Clarifications

The current publicly available LMHR manuscript is an unedited version
To facilitate scientific transparency and collegial discourse, we made the full database and
statistical code publicly available, though a link in the manuscript.
We welcome feedback, including any problems in the statistical code, for incorporation and/or
correction in the final manuscript.
These issues have been identified for correction in the final manuscript, some of which relate to
discrepancies between versions of the R-code application:

  • Figure 1. “Potentially unreliable data” should be n=288 and “>130 carb intake” will be
    changed from n=23 instead of n=24.
  • Table 1. % M/F will be changed from 60/40 to 57/42. Data rows for Current TG and
    HDLc were reversed and will be corrected.
  • Table 3. Between group p-value for prior TG/HDLc will be changed from 3.7 x 10-16 to
    3.5 x 10-16. Frequency for male in LMHR group will be changed from 56% to 43%.
  • Supplemental Table 1. Current HDLc p-value will be changed from 4.26 x 10-8 to 4.09 x
    10-8. Current TG p-value will be changed from 0.079 to 0.080. Other minor rounding
    errors will be updated.
  • Supplemental Figure 1. The 2nd (left to right) lower bar will be changed from 76 to 61
  • Supplemental Table 2. AIC for the BMI will be changed from 7227 to 7214. AIC for the
    prior TG/HDLc + BMI will be changed from 7230 to 7213.
  • Supplemental Figure 5A should indicate baseline BMI as an effect modifier, not
    mediator, among other forthcoming clarifications.

Additional clarification: Cardiometabolic risk factors are often considered as those comprising
the metabolic syndrome, exclusive of LDLc, such as in this recent CDC study. To avoid
unintended interpretations, we will add the word “otherwise” to the abstract conclusion as

These data suggest that, in contrast to the typical pattern of dyslipidemia, greater LDLc
elevation on a CRD tends to occur in the context of otherwise low cardiometabolic risk.

Announcement: $20,000 Donation for Lipid Triad Analysis

BREAKING: Last night I posted a tweet (see below) curious if any data scientists would be interested in querying one of the major longitudinal datasets (such as UK Biobank) on the “Lipid Triad” (high LDL-C + high HDL-C + low triglycerides) vs All Cause Mortality.

To my great surprise, an anonymous donor has come forward who is interested in funding this analysis to publication for $20,000.

If you or anyone you know is a data scientist who would be interested in this endeavor, please reach out to me via twitter (@daveketo) or via comment below.

Discussing the LMHR Study on the Diet Doctor Podcast

I had a great chat with Dr. Bret Scher, Medical Director of Diet Doctor, on the newly launched Lean Mass Hyper-responder study.

0:00 Introduction

3:06 Welcome, Dave Feldman

4:06 A rise in cholesterol for people on keto

12:05 Dave’s pushback in his research

20:14 Thoughts on the recent Women’s Health Study paper

29:39 Exciting news about Dave’s upcoming research

37:10 The concern about the data from a one-year study

43:26 What Dave hopes to achieve with the study

47:48 Switching saturated to mono-unsaturated fat in lowering LDL?

53:33 About Dave’s other work: Own Your Labs

1:03:49 Blood glucose and insulin on hypo and hyper caloric intake of low carb

1:08:40 Where to find Dave Feldman


The Citizen Science Foundation is proud to announce the launch of the Lean Mass Hyper-responder study, effective immediately.

This research will be conducted in partnership with the Lundquist Institute for Biomedical Innovation at Harbor-UCLA.

For official details, eligibility criteria, and how to join, please visit:

One Last Step

One of the challenges in this IRB process is that we don’t have a lot of latitude to give regular updates on it as it unfolds. Under most normal circumstances, aspects of the protocol draft and review, legal and liability considerations, and even the stages of approval are kept quiet until formal announcement and start of the study.

With that said, I’m quite confident we’re on the last step of this process. There are quite a few things on our long checklist with the IRB, and right now we’re on the last one. Hopefully we’ll have some news very soon.

Note: this is a duplicate post from our sister site,