Egregious - the Richard M. Fleming Story

An RCT paper claiming to show harmful effects of a low-carb diet passed briefly over the internet on the weekend before being shot down in flames.
The title is:

Long‐term health effects of the three major diets under self‐management with advice, yields high adherence and equal weight loss, but very different long‐term cardiovascular health effects as measured by myocardial perfusion imaging and specific markers of inflammatory coronary artery disease.
The paper is published in Clinical Cardiology [edit: not Preventive Cardiology], and is free to access.
https://onlinelibrary.wiley.com/doi/abs/10.1002/clc.23047

[update: the paper above has since been retracted. Its Pubpeer discussion is here
https://pubpeer.com/publications/AA81ABA153565C2A238B191DFE0FB6 ]
[Update: the paper has been quickly republished here, in what may well be a predatory journal https://biomedres.us/pdfs/BJSTR.MS.ID.002006.pdf
Retraction Watch story here: https://retractionwatch.com/2018/11/13/a-convicted-felon-writes-a-paper-on-hotly-debated-diets-what-could-go-wrong/ ]

Three of the authors work at Fleming's medical imaging company in California, one is a deceased psychologist from Iowa, another is a pediatric nutritionist from New York and one is a Kellogg's employee from Illinois.
How this group was able to run a 12-month diet trial in 120 subjects is something of a mystery.

The conclusions:

One‐year body mass changes did not differ by diet (P .999). Effect sizes (R, R2) were statistically significant for all indices. Coronary blood flow, R (CI 95%) = .48 to .69, improved with low‐to‐moderate‐fat and declined with lowered carbohydrate diets. Inflammatory factor Interleukin‐6 (R = .51 to .71) increased with lowered carbohydrate and decreased with low‐to‐moderate‐fat diets.
Conclusions
One‐year lowered‐carbohydrate diet significantly increases cardiovascular risks, while a low‐to‐moderate‐fat diet significantly reduces cardiovascular risk factors. Vegan diets were intermediate.

First I'll consider all the good reasons to reject this study, but after that I'll do something we should always do, even for the worst study - take it at face value.

1) Lead author Dr Richard M. Fleming is a self-confessed and convicted felony fraudster who has admitted falsifying data in another RCT.

Fleming admitted to knowingly executing and attempting to execute a scheme to defraud Medicare and Medicaid healthcare benefit programs in connection with the delivery of and payment for healthcare benefits, items, and services, namely by submitting payment claims for tomographic myocardial perfusion imaging studies that he did not actually perform. Fleming also pled guilty to one count of felony mail fraud in violation of 18 U.S.C. 1341 and 2 for conduct relating to money paid him to conduct a clinical study of a soy chip food product for the purpose of evaluating health benefits. As Fleming admitted during his guilty plea, he received approximately $35,000 for conducting a clinical trial, but he fabricated data for certain subjects.

https://www.federalregister.gov/documents/2018/09/28/2018-21210/richard-m-fleming-denial-of-hearing-final-debarment-order

2) Fleming obtained Robert Atkins' medical records by deception and shared them with Neal Barnard of the vegan activist group PCRM in 2004, another unethical behaviour and one demonstrating that Fleming has a long-standing animus against Atkins and his diet.
https://theskepticalcardiologist.com/2016/12/18/what-does-the-death-of-robert-atkins-tell-us-about-the-atkins-diet/

Now, vegans can do research into this exact question with a high standard of rigour, see Chris Gardner's studies - one can certainly dispute the interpretation of some results, but not the results themselves. And I have reviewed a vegan diet study favourably here (dealing with another Fleming paper in passing) - good results are good results; I don't doubt these diets can have also cardiovascular benefits over the short-to-medium term, but question their long-term effects on mental health, reproductive health, dental health, joint health etc.

3) The trial protocol number on the paper links to a study that was completed in 2002. This explains how a convicted felony fraudster was able to conduct a study. The protocol was posted in 2006, 4 years after the study concluded, which seems unusual on the Clinical Trials website.

https://clinicaltrials.gov/ct2/show/NCT00324545

4) The study has 35 citations - 15 of these are to Fleming's own papers. This self-spamming, which helps boost an author's citation rate, is frowned upon by reputable journals. One of these references has the word "quantum" in the title. Other references are to news articles and book chapters. The low carbohydrate diet references are more than 15 years old. Ref 16 is curious as an anonymous reviewer is given credit for the wording of a paragraph of interpretation, surely a run-of-the-mill interaction with a reviewer.
5) The novel aspect of this paper may lie in the reference to that novel vegan touchstone, Neu5Gc. Pro-tip - if the vegan diet had a magic mechanism, you ought to know it already; some major low-carb mechanisms have been understood for generations. At this rate, if there is a magic mechanism for vegan health benefit, it will be discovered by a low-carb scientist.

We  now  know  that  these  food  choices  and  their  impact  are  at  least partially precipitated by the inflammatory effect of our diets based given our inability to convert Neu5Ac to Neu5Gc and our bodies immune response to the Neu5Gc present in animal protein. 

At this point, let's take the study at face value. the vegan diet avoided animal protein and Neu5Gc, the low carb diet probably included twice as much protein as the other diets (based on reference 6), but the low fat diet included more animal protein and Neu5Gc than the vegan diet.

One‐year lowered‐carbohydrate diet significantly increases cardiovascular risks, while a low‐to‐moderate‐fat diet significantly reduces cardiovascular risk factors. Vegan diets were intermediate.

So Fleming's own study, taken at face value, doesn't support the Neu5Gc hypothesis. In fact, it's unusual for the vegan diet to be inferior to the low-fat diet in any vegan study, and it's unusual for the low carb diet to be inferior to the low-fat diet in any low-carb study.

6) adherence to diets over 12 months was 100%. Of course, this is unheard of and entirely implausible; if honestly reported, it seems to show considerable gullibility or self-deception in the study team.

 That 100% of participants continued on their respective diet plans through a full year of dieting contrasts  sharply  with  much  of  diet  research  experience  with  drop outs  and  with  common experience with difficulties of dieting and remaining on diets. This success can be attributed to attention to well-established psychological principles of habit acquisition and extinction and of behaviour modification through Bandura [17] counseling.

Bandura's ethos seems sensible enough and appropriate for such a project, except perhaps when the people using it for counselling already believe that one approach is preferable to another.

7) the original report of the 2002 study (ref 6), if it is the same study, reports diet groups differently.





8) Implausible randomisation was the red flag that saw the PREDIMED study and many others retracted. Here randomisation of n=120 into 6 groups produced this result:

The  58  female  and  62  male  participants  were  randomly  assigned  to  equal  dietary  groups  by casting  a  die.  There  were  no  statistical  demographic  differences  between  group  assignments. There  were  no  statistically  significant  differences,  or  even  trends,  between  diet  groups  at  the initiation of the study. Since the groups were unequivocally randomized for all fifteen-baseline indices, statistical inference to the initial population, described by Table 1, is appropriate. 

9) Fleming et al state "A  four-month  post-intervention  analysis was obtained to determine post-intervention treatment, which has not previously been reported in the literature."

Post intervention status was in fact reported at 4 years by the Shai et al DIRECT study group.



10) the sponsor is listed as the Camelot Foundation. A search turned up this mention - Dr Fleming is the editor of a predatory journal, and the Camelot Foundation has little other existence online, it seems to be a 501(c)(3) legal tax-avoidance scheme within Fleming's own business.
https://www.omicsonline.org/editor-profile/richard-maximus-fleming/

11) Cardiovascular improvement by Fleming's medical imaging method correlates with improvement in the TG/HDL ratio. Taken at face value, although TG/HDL doesn't improve in Fleming's "low carb" arm, it does in most of the people reading this who have tried a low carb approach, so if Fleming's diagnostics are accurate this is not bad news. Interleukin 6 also improves during fasting but not a ketogenic diet in a 6-day study, but improves in a low carb diet vs a low fat diet in a 6 month study here (as there was at least one previous study in the literature that came to different conclusions from Fleming et al with regard to an outcome they highlighted, this should really have been cited).

Both LFD and LCD led to similar reductions in body weight, while beneficial effects on glycaemic control were observed in the LCD group only. After 6 months, the levels of IL-1Ra and IL-6 were significantly lower in the LCD group than in the LFD group, 978 (664–1385) versus 1216 (974–1822) pg/mL and 2.15 (1.65–4.27) versus 3.39 (2.25–4.79) pg/mL, both P < 0.05.

Taken at face value, Fleming's possibly fraudulent paper predicts cardiovascular benefit from a low carb diet if people get different results from the ones he claims to have produced, which is usually the case in other studies and in real life...

The question is, how did this paper pass peer review with all the red flags above? [edit]



Credit to @MacroFour and Ivor Cummins @FatEmperor for the links regarding Dr Fleming's colourful past.

A Grand Unified Theory of Polyunsaturated Fatty Acid Misbehaviour in Inflammatory Disease

One of the great mysteries of nutrition is the behaviour of polyunsaturated fatty acids (PUFAs). They often look good in the kind of sloppy epidemiology used to drive or latterly protect dietary guidelines*, are more ambiguous in RCTs, and can easily be shown to have deleterious effects in a number of specific medical and experimental conditions that might be expected to have a "canary in the coalmine" validity as warnings when it comes to the longer-term effects of consuming more, sometimes much more, that the essential nutrient requirement for these functional molecules (which is, at a rough consensus, around 3% of energy, with 1% coming from omega-3 PUFAs).

However, higher intakes are sometimes tolerated well; any fairly liberal ketogenic diet including pork or olive oil or nuts or avocado will almost certainly exceed 3%, and even though PUFA over 3% is almost a requirement for the induction of NAFLD, Browning et al reversed NAFLD quickly with a ketogenic diet supplying 15%E as PUFA.[1]

So what gives? What is the nature of the interaction between PUFAs and other dietary components or metabolic states that produces inflammation?

In an earlier blog post I identified the enzyme systems upregulated in NAFLD as those of the microsomal ethanol oxidase system (MEOS) and also showed that the evolutionary function of the MEOS is to degrade PUFA, rather than alcohol which is a latecomer to our diets.
But what activates the MEOS when alcohol does not? How, for example, does fructose send PUFAs down this pathway, and how does this promote inflammation?

I found a clue in this hepatitis C editorial by Jenny Heathcote on a study in which weight loss improved liver function.[1] This is some quite brilliant speculation.

Here is the description of fatty liver due to insulin resistance (HCV causes IR by a pharmacological action of its core protein):

In peripheral tissues, insulin normally downregulates the hormone sensitive lipase (HSL) enzyme responsible for hydrolysis of stored triglycerides from free fatty acids within adipocytes. In patients who are insulin resistant, this enzyme is no longer suppressed. In addition, counterregulatory hormones such as catecholamines, glucagon, and growth hormone are increased in response to increased circulating insulin levels. These counterregulatory hormones stimulate HSL to hydrolyse more triglycerides into free fatty acids, the end result being an increased flux of dietary and stored free fatty acids away from the adipose tissues and towards the liver. Unfortunately, Hickman et al did not measure free fatty acid levels before or after the weight reduction programme. Within the liver, insulin upregulates esterification of free fatty acids to triglycerides. Once the triglycerides are formed, insulin downregulates the secretory pathways, thus favouring increased storage of triglycerides in the cytosolic pool. Furthermore, free fatty acids can themselves upregulate the esterification pathway. The net result is a positive feedback cycle contributing to an ever increasing amount of free fatty acids and triglycerides in the liver. Thus portal hyperinsulinaemia leads to hepatic steatosis.

And here is the description of how hepatic steatosis influences PUFA disposal:

 These studies have suggested that the presence of fat in patients with hepatitis C is associated with markers of progressive liver disease in that fat was associated with increased stellate cell activation, but the mechanism by which this takes place is uncertain. It is possible that this occurs secondary to saturation of beta oxidation pathways within mitochondria which then leads to free fatty acids becoming more available to intracellular microsomes where they undergo lipid peroxidation. There are three main products of microsomal lipid peroxidation: malondialdehyde, 4-hydroxynonenal, and hydrogen peroxide. Malondialdehyde has been shown to activate stellate cells to produce fibrin, and may be responsible at least in part for liver fibrosis in patients with non-alcoholic steatohepatitis.

Malondialdehyde (MDA) and 4-HNE are unsaturated products of PUFA, and H2O2 is also a step in the MEOS disposal of PUFA, requiring catalase for its reduction to H2O + O.



We can see how this relates to the "essential" role that PUFA plays in the development of alcoholic liver disease; not only can the liver become fatty from the conversion of alcohol to triglycerides, but also the disposal of excess ethanol through the MEOS has upregulated this enzyme system (hepatic CYP2E1 is upregulated 10-fold by ethanol); to add insult to injury, the liver's ability to dispose of excess fat via beta oxidation is impaired by the depletion of NAD+ during the conversion of ethanol to fat.
But another clue was supplied by Tucker Goodrich, the PUFA ninja, who found a rodent study showing that 4-HNE and 9-ONE could themselves be cleared if beta-oxidation pathways were upregulated enough, that is, by a ketogenic diet.[3]

Our results showed that livers from rats fed ketogenic diet or high fat mix diet had high ω-6 polyunsaturated fatty acid concentrations and markers of oxidative stress. However, high concentrations of HNE (1.6 ± 0.5 nmol/g) and ONE (0.9 ± 0.2 nmol/g) were only found in livers from rats fed the high fat mix diet. Livers from rats fed the ketogenic diet had low HNE (0.8 ± 0.1 nmol/g) and ONE (0.4 ± 0.07 nmol/g), similar to rats fed the standard diet. A possible explanation is that the predominant pathway of HNE catabolism (i.e. beta oxidation) is activated in the liver by the ketogenic diet. This is consistent with a 10 fold decrease in malonyl-CoA in livers from rats fed a ketogenic diet compared to a standard diet. The accelerated catabolism of HNE lowers HNE and HNE analog concentrations in livers from rats fed the ketogenic diet. On the other hand, rats fed the high fat mix diet had high rates of lipid synthesis and low rates of fatty acid oxidation, resulting in the slowing down of the catabolic disposal of HNE and HNE analogs. Thus, decreased HNE catabolism by a high fat mix diet induces high concentrations of HNE and HNE analogs. The results of the present work suggested a potential causal relationship to metabolic syndrome induced by western diets (i.e. high fat mix), as well as the effects of the ketogenic diet on the catabolism of lipid peroxidation products in liver.

So - any state in which beta-oxidation is inhibited, but fat is present, will see PUFA shunted into the microsome - essentially the MEOS - and a high production of damaging peroxides and aldehydes. This also happens when mice are fed a ketogenic diet, but the aldehydes can be disposed of by beta-oxidation.
Note that the high fat (non-keto) diet in the mouse study was the Surwit diet relatively low in PUFA and MUFA (coconut and soy oil), overloading beta-oxidation with a mixture of ~50% saturated fat and 22.5% sucrose. Don't try this at home, kids. 

For reasons of time I haven't gone into every possible ramification such as the role of peroxisomal oxidation in PUFA disposal, the proper function of the MEOS (making and disposing of eicosanoids), the hormetic effect on antioxidant systems of low level HNE production, and the difference between liver and other fat-burning tissues (i.e. is this relevant to heart disease if the same thing happens in muscle, macrophages, or endothelial cells? Magic 8 ball says very probably).

However, here's a model that allows us to predict and explain the likely role of PUFA in inflammatory diseases at a metabolic level. Especially, for now, liver diseases.




* FFQ epidemiology studies are notoriously inaccurate at capturing intakes of calories (and protein, which often looks wonky in epidemiology). They really can't tell you in what context PUFA is being consumed, and in any case it's hard to see how deep frying oil in food can really be measured - do you even know what your chips (fries) are cooked in and in what part of the FFQ would you put this information?

Edit 15/07/19
Structural and Functional Changes in Human Insulin Induced by the Lipid Peroxidation Byproducts 4-Hydroxy-2-nonenal and 4-Hydroxy-2-hexenal
https://pubs.acs.org/doi/abs/10.1021/tx200084d?fbclid=IwAR34wbn9HjlARatyfYwwzb0ZB4jBWGQkksCTDC5gY7ID3MLklz12qkrpFJY#

[1] Browning JD, Baker JA, Rogers T et al. Short-term weight loss and hepatic triglyceride reduction: evidence of a metabolic advantage with dietary carbohydrate restriction. Am J Clin Nutr. 2011 May; 93(5): 1048–1052.

[2] Heathcote J. Weighty issues in hepatitis C. Gut. 2002;51(1):7-8.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1773268/

[3] Li Q, Tomcik K, Zhang S, Puchowicz MA, Zhang G-F. Dietary-regulation of catabolic disposal of 4-hydroxynonenal analogs in rat liver. Free radical biology & medicine. 2012;52(6):1043-1053. doi:10.1016/j.freeradbiomed.2011.12.022
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3289253/