Start Here (Pinned)

Welcome to This site serves as an information and research hub for emerging data on cholesterol. particularly in the context of a low carbohydrate lifestyle.

If you know little to nothing about cholesterol ->

If you’re wanting to learn more about why cholesterol could be higher, particularly on a low carb diet, we present the Lipid Energy Model (LEM) ->

  • Our recently released full paper on the LEM published in Metabolites
  • You can watch our video abstract for this paper here (5 min):

If you’re looking to better understand the risk associated with high cholesterol on a low carb diet->

  • Note we are still recruiting for our LMHR study through the Lundquist Institute. Check out our official recruitment page at to find out if YOU qualify!
  • While several articles on this site present a more “cautiously optimistic” perspective on cholesterol in the context of fat adaptation, we strongly encourage everyone to consider the conventional view as well. Consider reading The Case for Lower LDL on Low Carb by our colleague and co-investigator, Spencer Nadolsky.

If looking to understand the “Lean Mass Hyper-responder” profile ->

If you’d like to understand possible relevance of cholesterol and the immune system, you can read Siobhan’s overview article on the topic here or watch her presentation here

If you’d like to learn more about lipoprotein(a), you can watch Siobhan’s presentation on it here

Lastly — you can always just ask us anything our Questions Page. (Just be aware our site does not constitute medical advice and we always recommend consulting with your doctor.)

Heading to Florida with Some New Pins in Hand (Get Yours Here)

I’m excited to announce I’ll be heading to the Keto Orlando Summit this weekend.

I’m also excited to announce we made some commemorative magnetic pins for this occasion to help fundraise for the Citizen Science Foundation. If you donate $25 or more (see below), we’ll ship you the pin directly, or I’ll actually hand it to you if you’re attending the conference since I’ll have them on hand there.

For more on the details, visit the official page for this drop at CSF

Lean Mass Hyper-Responder: Is it the saturated fat?

It’s a question that comes up again and again: Could Lean Mass Hyper-Responders (LMHRs) just be eating more saturated fat?

Rather than repeatedly answering this question in cumbersome Twitter threads, I thought I’d save both Dave and myself some time by consolidating 5 points that challenge the saturated fat hypothesis of LMHR:

1. Magnitude of effect. Many LMHRs exhibit LDL-C levels of well over 300 mg/dl. Mean levels from the LMHR participants in our cohort study were 320 mg/dl (Table 3), and we certainly have a handful with LDL-C above 500 mg/dl. I’m not aware of any data suggesting saturated fat intake, even at extreme levels, can cause increases in LDL-C to such an extent. If you – the reader – are aware of such data, please do share.

2. LDL-C change has an inverse relationship to BMI. In the same study, we observed that LDL-C change has an inverse relationship to BMI on a carbohydrate restricted diet. If this is the case and saturated fat was the primary driver of the increase in LDL-C observed, then the logical implications are either that there is a dose-response effect of saturated fat on body weight loss (i.e. saturated fat makes subjects leaner) OR that persons who are lean and adopt low-carb diets preferentially eat higher proportion saturated fat diets. For example, if we had three subjects, Jamie, Nicky, Leslie, and Amir with BMIs of 24, 26, 21, and 30, respectively, and they were all to adopt low-carb diets, Leslie would eat more saturated fat than Jamie who would eat more saturated fat than Nicky who would eat more saturated fat than Amir. 

3. The triad. A critical point is that LMHR are defined, not by LDL-C alone, but the triad of high LDL-C, high HDL-C and low triglycerides. Therefore, any complete model explaining the LMHR phenomenon must account for this triad. Thus, we must ask, could eating more saturated fat cause to the triad of markers seen in LMHR? (And that’s assuming LMHR eat primarily high-saturated fat low-carbohydrate diets.) In our study, mean levels were 320 mg/dl LDL-C, 99 mg/dl HDL-C, and 47 mg/dl triglycerides.

4. Adding back moderate carbs attenuates the phenotype. In the case series associated with the aforementioned study, six LMHR or near-LMHR subjects were instructed to introduce 50-100g/d carbs to help replenish hepatic glycogen and, via the postulates of the Lipid Energy Model, lower their LDL-C. No further instructions were given, i.e. subjects were not told to lower saturated fat intake. Nevertheless, all subjects exhibited substantial reductions in LDL-C, including one who exhibited a drop from 665 mg/dl to 185 mg/dl. Even if saturated fat intake occurred spontaneously (without instruction), it seems unlikely it could explain the magnitude of the effect. 

5. Case series. Although it is an n = 1, we published a case series of an individual consuming a ketogenic diet relatively low in saturated fat (~82% unsaturated, ~18% saturated). His pre-low carb LDL-C was 95 mg/dl. But following carb reduction, LDL-C rose to a peak of 545 mg/dl, despite relatively low saturated fat intake. Furthermore, LDL-C trended inversely with BMI and, when saturated fat intake was altered for experimental purposes, LDL-C remained concordant with the BMI trend and, thus, inconsistent with the prediction that saturated fat drives LDL-C change in this subject. Even if this is only a case report, recall that these data represent a real LMHR patient and are consistent with both the LMHR cohort study and the Lipid Energy Model.

Certainly, none of these 5 points dispute that saturated fat could contribute to the increase in LDL-C seen in LMHR; however, they do suggest that saturated fat is not the primary driver of change. Disregarding point 5, could high saturated fat intake be permissive or even required for the LMHR phenotype? Possibly. However, unless one can explain how saturated fat intake could account for these data, we should assume there is more at play worth investigating. 

For more reading and viewing:


Lipid Energy Model:

Lipid Energy Model video abstract:

Lean Mass Hyper Responder study:

Lean Mass Hyper Responder video abstract:

Lean Mass Hyper Responder Case report:

Lean Mass Hyper Responder Case report video abstract:

Lipid Energy Model Published

We’re excited to announce the publication of the Lipid Energy Model in Metabolites.

The Lipid Energy Model (LEM) examines the phenomenon of rising cholesterol levels for those on low carbohydrate diets and how this may provide powerful insights into lipid metabolism overall.

The LEM provides a mechanistic explanation for:

  • The Lean Mass Hyper-Responder (LMHR) phenotype, characterized by the triad of high LDL (at or above 200 mg/dL), high HDL (at or above 80 mg/dL or above), and low triglycerides (at or below 70 mg/dL), as well as for
  • The phenomenon that LDL-C change on low-carbohydrate diets tends to associate inversely with BMI.

The LMHR phenotype and observation that leaner people with better metabolic health markers are possibly at greater likelihood for increases in LDL-C were described in our prior observational cohort study:

We have also recently provided a particularly comprehensive clinical vignette of an LMHR:

As a personal aside from the first and last authors (Nick Norwitz and Dave Feldman), while we’ve longed looked forward to this milestone, we wish to note that this is just the first of many. Interest in LMHR continues to grow, not only within the low-carb community, but within academic medicine. The publication of the LEM hypothesis paper is a landmark, not because it provides a comprehensive theory with rigorous support from human trials assessing the model, but because it presents a concrete hypothesis with direct and testable predictions.

It is our hope that the publication of the LEM paper (version 1.0) will encourage fellow researcher to help us test these ideas in interventional trials and, thereby, advance scientific knowledge regarding LMHR and, perhaps, human lipid metabolism more broadly.

LMHR Case Study – Genetics

In our recent Lean Mass Hyper-Responder (LMHR) case report, we performed extensive genetic testing on the subject, LM, and activity sought input from clinicians and geneticists outside the research team. Because of the broad nature of our testing and absence of notable findings, as determined by expert consultants, and manuscript length limitations, we stated in our publication: “Whole exome sequencing performed by Veritas Genetics, and independent dyslipidemia and ASCVD genetic risk testing by GB Healthwatch, revealed no pathogenic or likely pathogenic variants that could account for LM’s phenotype.”

As part of our continued commitment to open and transparent science, we are happy to disclose what results we can. But before doing so below, it is important to highlight two caveats:

1. While a whole exome sequence was ordered, it would obviously be inappropriate to release the entire Variant Call Format file. This represents the subject’s actual genetic code and to disclose the code would be a serious infringement on the subject’s privacy, and one that could potentially be used to his disadvantage in future. Thus, calls for access to the complete raw exome sequence by individuals not part of the patient’s care team will not be entertained.

2. Given the scope of the genetic testing performed, it is inevitable variants will be found. Every person carries risk variants. The important question is whether variant can explain the clinical and metabolic (LMHR) phenotype at hand. By way of example, a risk variant for elevated triglycerides is not clinically relevant if the subject’s triglycerides are 40 mg/dl.

Moving onto the genetic findings: In addition to the exome sequence with professional geneticist interpretation, we also ordered a Dyslipidemia and ASCVD comprehensive risk panel — the same one that is being ordered for the LMHR prospective trial (recruitment underway) and that includes the following targets: LDLR, APOB, PCSK9, LDLRAP1, LPL, CETP, LCAT, LIPC, LIPE, LIPG, LPA, PPARG, STAP1, ABCA5, ABCA6, ABCG5, APOC2, LMF1, GBIHBP1, CREB3L3, GCKR, SCARB1, ABCA1, APOA1, APOA5, BHMT, CBS, MTHR, MTR, MTRR, PEMT, SHMT1, GUCY1A1, ITGB3, MEF2A, NOS3, PLA2G7.

As shown below, and consistent with the published text, testing by GB Healthwatch, “revealed no [0] pathogenic or likely pathogenic variants that could account for LM’s phenotype.”

In terms of variants not classed as pathogenic/likely pathogenic, there were no variants in genes classically associated with familial hypercholesterolemia at levels seen in this patient: LDLR, APOB, LDLRAP1, or PCSK9. Overall, six potentially notable variants were identified, as follows:

Pathogenic and Likely Pathogenic Variants (0) and Variants of Uncertain Significance (VUS) and High-Risk Variants, provided by GB Healthwatch

  • ABCG8 521G>A, heterozygous VUS for sitosterolemia. Sitosterolemia is a condition in which plant sterols and other sterols can be hyperabsorbed. Given that this is a hyperabsorption disorder, the mainstay of therapy is dietary restriction of both cholesterol and plant sterols. Of note: (i) testing performed on LM on October 20, 2021 confirmed normal campesterol levels at 8.2 mg/L. (ii) Most importantly, a hyperabsorption phenotype is inconsistent with the clinical presentation. As reported in the manuscript, when hyperabsorption was considered in the case of LM,LM was recommended to reduce dietary cholesterol intake, eliminating liver, shellfish, and egg yolks from his diet (in substitution for lean chicken, fish, and egg whites). One month later, in September 2020, his LDL-C was remeasured at 545 mg/dl (HDL-C 94 mg/dl, TG 58 mg/dl).” Thus, LM’s dietary cholesterol intake was lowest when his LDL-C was at its peak, and hyperabsorption cannot account for his LDL-C phenotype on a carbohydrate restricted diet.
  • APOA5 3’UTR, heterozygous variant and risk factor for hypertriglyceridemia. The patient’s triglycerides ranged from 39 – 58 mg/dL. Again, the risk variant is inconsistent with the metabolic phenotype.
  • CBS 133C>T missense heterozygous VUS for homocystinuria. LM has no clinical signs of this disorder and homocysteine last measured January 3rd 2020 normal at 9.7 umol/L.
  • CDKN2B-AS1 22124478A>G homozygous variant that codes for an altered form of a long non-coding RNA that is associated with increased risk of myocardial infarction in some studies. While this variant does associate with increased risk of myocardial infarction, the increased risk does not appear to be driven by alternations in blood lipids, including LDL-C and there is no reason to believe it contributed to the patient’s phenotype on a carbohydrate restricted diet.
  • LPL 953A>G heterozygous variant that has been linked to increased risk for hypertriglyceridemia. As noted, the patient’s triglycerides ranged from 39 – 58 mg/dL. Again, the risk variant is inconsistent with the metabolic phenotype.
  • SLC22A1 1022C>T heterozygous variant and risk factor for elevated Lp(a). As stated in the report, Lp(a) was high in the subject prior to adopting a ketogenic diet and is elevated in the patient’s father. This allele is presumably paternally inherited and there is no reason to believe it contributed to the patient’s LMHR phenotype.

In addition to the targeted genetic risk panel, a 144-page report corresponding to the exome sequence also revealed no known pathogenic or likely pathogenic variants for heart disease. Under “important” risk variants, the patient was noted to be a carrier for HFE 845G>A for hereditary hemochromatosis, MEFV 442G>C for Familial Mediterranean Fever, and MMP2 524G>A for multicentric osteolysis, nodulosis, and anthropopathy. These are all autosomal recessive conditions, and the patient does not present with signs or symptoms of any of these disorders. Homozygosity for lactose intolerance was also noted.

Under “noteworthy” variants, 7 variants were identified for cancer risk, 3 for clotting disorders, 4 for neurological disorders, 4 for other organ health, and only one for cardiovascular disease. This variant was in KCNE1 with classification of “no known risk” (VUS), for long QT syndrome, as detailed below. Taken together, the exome sequence revealed no notable findings that could explain the patient’s LMHR phenotype.

Thus, these genetic data provide no means by which to explain the patients presentation, defined by a shift from normal LDL-C of 95 mg/dl while on a mixed macronutrient diet to an LMHR triad of LDL-C 393 – 545 mg/dl, HDL-C ~115 mg/dl, triglycerides ~40 mg/dl. In fact, several of the identified variants were associated with increased risk for hypertriglyceridemia, which is in obvious contrast to the patient’s presentation.

In summary, our interpretation is that, while one cannot rule out genetic contribution or modification, the evidence at hand is most consistent with the hypothesis that the LMHR phenotype is driven by non-genetic factors including leanness and dietary macronutrient composition. More details on the Lipid Energy Model will be forthcoming shortly.