Testing the effect of a high fat diet in severe type 2 diabetes - Garg and Unger metabolic ward study from 1988

It's all very well to test the metabolic effects of high-fat diets in RCTs. There are usually beneficial results in type 2 diabetes, but compliance is limited. The trial isn't showing what the diet does, but what effect the advice has on people who may be more or less indifferent to it. In fact, it's amazing these trials produce the positive results they do.
A metabolic ward study involves subjects who follow the diet because they have nothing else to eat; all variables such as exercise are kept constant. Because you don't get huge numbers volunteering for these studies, and the cost is high because of the round-the-clock supervision and testing, the crossover method is normally used. Half the subjects eat the test diet, the other half the control, then they switch over. Results from the end of each period in both groups are averaged.
This is a 1988 study authored by Abhimanyu Garg, Roger Unger and 3 colleagues.

N Engl J Med. 1988 Sep 29;319(13):829-34.

Comparison of a high-carbohydrate diet with a high-monounsaturated-fat diet in patients with non-insulin-dependent diabetes mellitus.

Garg A1, Bonanome A, Grundy SM, Zhang ZJ, Unger RH

Abstract

We compared a high-carbohydrate diet with a high-fat diet (specifically, a diet high in monounsaturated fatty acids) for effects on glycemic control and plasma lipoproteins in 10 patients with non-insulin-dependent diabetes mellitus (NIDDM) receiving insulin therapy. The patients were randomly assigned to receive first one diet and then the other, each for 28 days, in a metabolic ward. In the high-carbohydrate diet, 25 percent of the energy was in the form of fat and 60 percent in the form of carbohydrates (47 percent of the total energy was in the form of complex carbohydrates); the high-monounsaturated-fat diet was 50 percent fat (33 percent of the total energy in the form of monounsaturated fatty acids) and 35 percent carbohydrates. The two diets had the same amounts of simple carbohydrates and fiber. As compared with the high-carbohydrate diet, the high-monounsaturated-fat diet resulted in lower mean plasma glucose levels and reduced insulin requirements, lower levels of plasma triglycerides and very-low-density lipoprotein cholesterol (lower by 25 and 35 percent, respectively; P less than 0.01), and higher levels of high-density lipoprotein (HDL) cholesterol (higher by 13 percent; P less than 0.005). Levels of total cholesterol and low-density lipoprotein (LDL) cholesterol did not differ significantly in patients on the two diets. These preliminary results suggest that partial replacement of complex carbohydrates with monounsaturated fatty acids in the diets of patients with NIDDM does not increase the level of LDL cholesterol and may improve glycemic control and the levels of plasma triglycerides and HDL cholesterol.

Thanks to Ivor Cummings, I have the full-text pdf, and it's very interesting.
The other dietary variables are well controlled for.

The types of fatty acids, if that makes any difference, are also well-matched between diets (low fat diet used corn and palm oils, high fat diet used olive oil, so neither was high omega-3).
The results are fascinating (this is the average from the last week of each period, days 21-28).


Who knew that a urinary glucose output of 142 mg/day was normal on a high-carbohydrate diet in subjects with "non-insulin dependent diabetes mellitus treated with insulin" - to disappear completely on a diet with 50% of calories from olive oil?
Oh, and the base line? That was after a week on the diet recommended by the ADA in 1988, which was the lead-in diet.
What about lipids? They improved too:



What's especially interesting aboout these lipid results is the comparison between this study (second phase T2D) and Garg and Unger's 1992 study of the same diets in mild (first phase) T2D. In mild T2D, a high MUFA diet improved lipids but did not influence insulin sensitivity. This seems consistent with high-carb/high-calorie diets and hyperinsulinaemia in those prone to diabetes driving lipotoxicity, when then produces the phase 2 phenomenon of hyperglycaemia plus hyperlipidaemia by altering the ratio of alpha- to beta- cell sensitivity and activity. Dietary carbohydrate drives fat which drives endogenous glucose.

The authors of the 1988 paper sum up thus:

Abhimanyu Garg has authored this convenient review of all the studies using a high-MUFA diet for Type 2 NIDDM.
It includes this classic line:

 "The improvement in the glycemic profile with high-monounsaturated-fat diets may not be related to changes in insulin sensitivity but to a reduction in the carbohydrate load, which patients with type 2 diabetes may not be able to handle readily because of severe insulin resistance and b cell defects."
Amen.

Of course what we lack is a comparative series of studies with high SFA diets, or indeed diets in the normal range of mixed SFA, MUFA and PUFA. Does the type of fat matter if carbohydrate is low enough? Quite possibly not, at least for the majority. Is 35% carbohydrate low enough to see the full benefit of a high-fat diet? Maybe not, but the results, after only 28 days, were impressive enough.

Some Answers to that Question

Let's back the truck right up to 1923. Insulin was introduced the year before and is now being mass produced by Eli Lilly & Co. Diet research into diabetes has crystallized into a proven diabetic diet (the diet that will best extend life in the absence of insulin). This is defined by Ladd and Palmer in the American Journal of Medical Science of August 1923 with this formula.



Thus we have the classic 1:4 ratio of carbohydrate to fat that defines the limits of ketosis, with the refinement that protein straddles the separation, with 58% of protein counting as carbohydrate (gluconeogenic) and the remainder as fat (ketogenic).
Note the epilogue; "we feel that adequate dietary control will remain the basis of treatment for many cases, especially those of the milder type".
As in this example from David Unwin's 2014 case series. This is the sort of paper I like best today; what it lacks in randomised, controlled rigour (ffs people, the point has already been proven) it makes up for in realism - these types of papers show what it takes to treat real people in clinical practice, it's hard work sometimes but the results are clearly worthwhile.



And here's a picture of Dr Unwin, getting the point across. 

How does this fit with the glucagonocentric restructuring of diabetes by Roger Unger (the "father of glucagon" and 1975 Banting Medal recipient)?

When someone with deficient insulin response (like the woman in Dr Unwin's example - polydipsia is a sign of beta-cell failure) eats carbohydrate, these things happen:

1) Glucose enters the blood and glucagon is elevated as a result.



High glucagon concentrations stimulate the liver to break down glycogen, releasing glucose, and to make more glucose from protein (58%) and triglycerides (10%), as well as ketone bodies.

2) If the blood sugar concentration rises far enough - more likely after a carbohydrate meal - glucose enters the liver cells at an excessive rate. This glucotoxicity increases gluconeogenesis further. In fact I suspect this is the tipping point, and that this high blood glucose and its effect on the liver (and kidneys) is what separates the ketoacidosis of starvation from the lethal ketoacidosis of diabetes.

3) because extra carbohydrate has been consumed in the meal, the clearance by cellular oxidation of hepatic glucose and ketone body output is proportionately delayed, maintaining a higher level of glycaemia than would otherwise be the case.

In 1923 ketogenic diets (then known for the treatment of epilepsy and tested alongside diabetic diets) were not considered suitable for diabetics.
[Edit: this is incorrect as the diabetic diet of Newburgh and Marsh, 1923 is a true ketogenic diet at 35g carbohydrate per day, 0.67g protein per Kg per day, and the remainder of 30-40 calories per Kg per day from fat. Protein is restricted to 35g day in the induction period with 85-95 g fat per day. Newburgh and Marsh report the complete absence of acidosis in 180 patients, including juvenile cases, maintained on their diet, which would be considered a ketogenic diet in non-diabetics. "Not only does acidosis not develop in patients who are living on this diet, but it is a fact that all our patients showing at admission an acidosis short of coma rather promptly lost their acidosis while taking the high fat diet."]
At some point in the 70's or 80's they became acceptable (that the brain can run on ketones was discovered in 1967). The added benefit of a ketogenic diet is a further restriction of the glucagon curve, less competition from dietary glucose, less glucotoxicity, and, in animal experiments, a diminishing of pancreatic alpha cell to beta cell ratio.
In normal starvation metabolism a high level of ketone bodies stimulates the release of insulin from beta cells, just as a high level of glucose does, so that the action of glucagon is kept optimal, resources are not wasted, and toxicity is minimised.

A quick search shows that insulin is not universally available today, and not always affordable by everyone. In communities where this is the case the research of 1923 is still relevant for type 1 diabetics, in terms of survival until medicine is available and affordability (making a limited supply of insulin last longer). Safety is also an issue - the lower the dose of insulin required, the lower the risk of hypoglycaemia. It is certainly still relevant for type 2 diabetics. Low carbohydrate, high fat diets will probably become the norm again sooner than we think.
Even the Daily Mail has dispensed with the usual "experts warning about all that fat" add on when discussing LCHF therapeutic diets.

The $64,000 Question

I've been obsessed with this question. It all started after reading the literature of pre-insulin treatment of diabetes and insulin-free animal models of the disease. Feeding fats and restricted protein, with no or minimal carbohydrate, gives the best prognosis without insulin; also fasting, which tends towards a similar mix of substrates. Then I read the literature showing inferior prognosis of type 2 diabetes with higher-carbohydrate, post-1977 diets; the longer carbohydrate is fed, the more fasting glucose climbs. Most of the glucose in the blood of diabetics comes from hepatic GNG, not dietary carbohydrate (I'm grateful to Carbsane for pointing this out to me originally - it's an important point).
I read the Richard K. Bernstein paper which describes tight control of glucose using low doses of insulin and a low carb diet, and then I watched the Robert Unger lecture I linked to in the previous post, and saw a slide of blood glucose levels in a mouse with no beta cells, given insulin normally (wide fluctuations ranging into hypo- and hyperglycaemia) or given insulin plus a glucagon antagonist.


Also compare the recent case study "Type 1 diabetes mellitus successfully managed with the paleolithic ketogenic diet" by  Tóth and Clemens.

"He was put on insulin replacement therapy (38 IU of insulin) and standard conventional diabetes diet with six meals containing 240 grams carbohydrate daily. He followed this regime for 20 days. While on this regime his glucose levels fluctuated between 68–267 mg/dL.
Average blood glucose level while on the standard diabetes diet with insulin was 119 mg/dL while 85 mg/dL on the paleolithic-ketogenic diet without insulin. Fluctuations in glucose levels decreased 
as indicated by a reduction of standard deviation values from 47 mg/dL on the standard diabetes diet to 9 mg/dL on the paleolithic-ketogenic diet. Average postprandial glucose elevation on the standard diabetes diet was 23 mg/dL while only 5.4 mg/dL on the paleolithic-ketogenic diet." 

Again the question - why does LCHF (or fasting) act like the glucagon receptor antagonist? Why does feeding glucose worsen hyperglycaemia and ketoacidosis, and fat improve them, when fat, and protein, not glucose, are the gluconeogenic and ketogenic substrates?
Could it be that in diabetes - when there is no insulin present, or when the cells of the liver are highly insulin-resistant, or when subcutaneous insulin fails to give attain an adequate concentration in the portal vein feeding the liver - glucose itself in some way promotes gluconeogenesis and ketogenesis?
Consider first that in uncontrolled diabetes blood glucose is very high and becomes even higher after carbohydrate feeding. This is especially so in the portal vein feeding the liver. Hepatocytes without insulin are not resistant to glucose, especially at high concentrations. The Glut2 receptor is not wholly controlled by insulin, though the metabolism of glucose within the cell is.

Concentrations of glucose approaching 10 mM are pre-diabetic levels. Concentrations of glucose above 10 mM are analogous to a diabetic condition within the cell culture system. This is important because the same processes that can affect cells and molecules in vivo can occur in vitro. The consequence to growing cells under conditions that are essentially diabetic is that cells and cell products are modified by the processes of glycation and glyoxidation. These processes cause post-translational secondary modifications of therapeutic proteins produced in cell cultures. [Sigma cell culture guide]
This excess glucose is getting into the cell, and is modifying its metabolism in ways that promote and increase the hormonal action of glucagon.

For example, in this mouse study.

Glucotoxicity Induces Glucose-6-Phosphatase Catalytic Unit Expression by Acting on the Interaction of HIF-1α With CREB-Binding Protein, A. Gautier-Stein et al. 2012.

We deciphered a new regulatory mechanism induced by glucotoxicity. This mechanism leading to the induction of HIF-1 transcriptional activity may contribute to the increase of hepatic glucose production during type 2 diabetes.

If that's true in humans (and I have to say it's very unlikely that glucotoxicity will do anything good for you) then minimising post-prandial glucose spikes is going to help keep a lid on fasting glucose levels as well.

There's also the concept of reductive stress; the metabolism of excess glucose will result in a buildup of NADH and a relative deficiency of NAD+. The cell copes with this by a number of mechanisms. Ketogenesis itself helps, because the conversion of acetoacetate to Beta- hydroxybutyrate generates NAD+. Under conditions of high glucose, glyceraldehyde-3-phosphate will build up in the cell unless cytoplasmic NADH is continuously re-oxidized. Cells oxidize cytoplasmic NADH by a combination of three pathways, the aspartate:malate shuttle, the glycerol:phosphate shuttle and during the conversion of pyruvate to lactate.Pyruvate may not enter the mitochondria. It may be reduced to lactic acid by lactic acid dehydrogenase. This reaction is driven when the cell’s need to oxidize NADH to NAD for use as a substrate to keep glycolysis working. Pyruvate reacts with hydrogen peroxide and forms water, carbon dioxide and acetic acid. This non-enzymatic reaction helps the cell defend itself from oxidative intermediates.

Now, in our model, pyruvate will not enter the mitochondria, because that step is controlled by insulin. This means that lactate will either be recycled to glucose or exported. So what is the link between lactate and diabetes?
Plasma lactate predicts type 2 diabetes here.
And lactic acidosis is a common finding in cases of diabetic ketoacidosis, here.
In starvation (very good account here, thanks to Ash Simmonds for the link), pyruvate, lactate, and alanine are exported to the liver for conversion into glucose. So, glucose is a gluconeogenic substrate. Meanwhile the poor hepatocyte is trying to oxidise fatty acids, making some ketone bodies in the process, but also struggling with the need to fend off, by metabolizing, devastatingly high glucose concentrations.I speculate that the liver's ATP needs are not being met under these conditions (of futile cycling), and that this is a trigger that increases sensitivity to the lipolytic effect of glucagon in adipocytes (as it is supposed to increase appetite in the liver homeostasis model of appetite regulation, how no-one knows).
And that ketogenesis is also increased by glucotoxicity. But the mechanism of all this is beyond me at present, I'm just sayin' that these are possibilities.

I don't feel that I've answered the $64,000 question yet. But I do think that idea of a glucose -> gluconeogenesis vicious cycle has merit in the type of imbalanced systems we've been looking at, where adding glucose has been a bad idea, and removing it a good one, since history began.

Now it may be that the answer is very obvious and doesn't need any of these baroque explanations.
In which case, please feel free to tell me. All I want is a formula that's consistent with every fact. Is that too much to ask?

Further Notes on Glucagon Dominant Hepatic Metabolism

The idea of the various forms of diabetes and pre-diabetes as a too ready susceptibility to glucagon dominance of hepatic metabolism, as described in the previous post, is not intended to diagnose or treat any disease (which I am not qualified to do, but I could not afford to buy the iamnotadoctor.com web address). It's a heuristic, a rule of thumb which puts what I've been reading into context, and which hopefully encapsulates a useful way of thinking about these diseases.
It produces many other thoughts, questions, and findings that seem consistent with the facts.

Firstly, a question; if the liver is producing higher than normal blood glucose levels, even between meals, why is there any need for glucose in the diet at all?
Secondly, a fact and a question; not all glucose usage depends on insulin. The brain uptake doesn't, and the liver uptake doesn't either, but in the case of the liver important pathways of glucose usage (glycogenisis and lipogenesis) do require insulin. So what happens to glucose taken up by the liver when these pathways are blocked? I am guessing that glycolysis generates excess NADH - the reductive stress that high glucose concentrations generate in cell cultures - which slows down fatty acid oxidation, which requires NAD+. This may be why feeding a little glucose decreased ketonemia even in pre-insulin diabetes. However, then we have to explain Petren's 1924 finding that a high fat, restricted protein diet suppressed ketonaemia. I wish I had a translation of the original paper, which was in German - all I have is this tantalizing hint. Was he talking about the patients in his practice (Karl Petren was a famous diabetes clinician), or some one-off short-term experiment?Edit: Ketogenesis consumes NADH and relieves reductive stress. This is more in line with the Petren finding.

If you had no, or low, insulin production from beta cells, or severe insulin resistance, it seems likely to me that carbohydrate itself would contribute to elevated glucagon. This seems wrong, but in fact a rise in glucagon is the first step in the insulin production cascade. Glucagon stimulates its counter-hormone insulin, which then suppresses glucagon.





 So, imagine what happens in column one without insulin to bring down the glucagon, or if the phase 1 insulin response is delayed (as in pre-diabetes) or if the alpha cells of the pancreas have become insulin resistant. Glucose in this context is contributing to an elevation of glucagon, yet it's not a substrate that glucagon-dominated metabolism can act on. Instead, the rise in glucagon is going to result in even more glucose being produced from hepatic metabolism.

What is the requirement for carbohydrate? I see it as a requirement for those useful and essential nutrients that are highly associated with carbohydrate. Magnesium, ascorbate, folate, potassium (which meat also supplies), carotenoids, fibre, and a different variety of trace elements to supplement those found in meat.
From this perspective it makes sense to include non-starchy vegetables and fruit in the diet if possible. Fruits and sweet root veges, in a low carb diabetic diet, may have advantages over starches (and are certainly not inferior to them, unless sweetness is an appetite trigger). Pre-insulin diets for diabetics were woefully lacking in micronutrients and this may well have produced inferior results to those seen today.
There are other reasons why we see better results today than was generally the case historically. The general diet is higher in carbohydrate and sugar and lower in fat, so there is more room for improvement. Diabetes is usually diagnosed sooner, pre-diabetes is diagnosed more often, there are multiple noninvasive biofeedback devices to check sugar and ketones with, and there is the safety net of insulin. It is hard to avoid the conclusion that insulin, if needed, is the ideal drug treatment for diabetes. Drugs which stimulate beta cell insulin secretion give inferior results and seem liable to exacerbate the loss of beta cell function, especially if they simultaneously upregulate amylin secretion; whereas insulin, like a ketogenic diet, is giving the beta cells a rest.
And what effect does this have? Hat tip to the astute Melchior Meijer for pointing out the relevance of this study on the Hyperlipid blog.

Long-term ketogenic diet causes glucose intolerance and reduced β and α-cell mass but no weight loss in mice.
(Sounds bad! What happened????) Long-term KD resulted in glucose intolerance that was associated with insufficient insulin secretion from β-cells. After 22 wk, insulin-stimulated glucose uptake was reduced. A reduction in β-cell mass was observed in KD-fed mice together with an increased number of smaller islets. Also α-cell mass was markedly decreased, resulting in a lower α- to β-cell ratio.
That sounds like an effect that might rein in glucagon a little, if it translates to humans.

And here we have Calorie's Proper's 2012 take on this; Gluca-Gone Wild!

And Peter D.'s post on insulin that got me started on this road.

Also, a video lecture by Robert Unger - "A New Biology for Diabetes".

And an AHSC2012 presentation by Maelán Fontes about antinutrients which, among other things, dysregulate glucagon.


These antinutrients are opioids, and, curiously, opiates and other psychoactive drugs were once used in attempts to control diabetes.

(excerpt from Fatal Thirst - Diabetes in Britain until Insulin, 2010 by Elizabeth Lane Furdell)

Diabetes - an Evolutionary Hypothesis

        Hyperlipid wrote an interesting post a while back about the Inuit, and how a genetic mutation means that many of them are never in ketosis even eating a very low carb diet. Basically, the mutation ensures that low carb brain metabolism runs predominantly on glucose, not ketone bodies. Because similar mutations are common in other populations dependent on low carbohydrate seafood diets, it must be a valid alternative to ketone metabolism in these populations.
        But what really interested me is that not everyone has the mutation. Natural selection has kept the Inuit’s options open. Should the population meet conditions under which the mutation is a handicap, the race will continue. And there is something very human about this. I doubt you will see a comparable variation in bowerbirds, for example. Humans are the acme of long-term survivalists, retaining a diversity of metabolic variations rather than developing a single, fiendishly clever, niche specialisation. There are humans that make vitamin D from very little sunshine (but are highly vulnerable to burning solar radiation), and humans that make D slowly (but are largely immune to sunburn). Our ancient and modern population movements and interbreeding have increased this diversity in any given place, but the Inuit demonstrate that there is something innate about it. It gives us what Nassim Nicholas Taleb memorably calls “antifragility”.
          Now consider the genetic propensity to insulin resistance and type 2 diabetes. At the genome level this is expressed in more than one way (as are the carnitine mutations of the Inuit type). The result of becoming insulin resistant is the glucagon dominance of hepatic metabolism. Glucagon wants to rip fat and protein into ketone bodies and glucose. In type 1 diabetes this gives you diabetic ketoacidosis (systemic acidosis triggered by a toxic brew of hyperglycaemia and hyperketonaemia) and the wasting of fat and protein reserves, but when dietary carbohydrate, or perhaps any food, is unavailable this glucagon dominance is the means of survival. Those individuals who become keto-adapted easily – those who become insulin resistant quickly – have an advantage. These are the strong ones in lean times (and as humans are social animals, they can help the others survive). But in the interests of anti-fragility, human evolution also favours some individuals with extra amylase gene copies (just a copy, the low-hanging fruit of evolution) and insulin sensitivity. These individuals can become strong when starchy foods are plentiful, and will survive longer when animal foods are unavailable.
         Let’s say you’ve descended from the insulin-resistant line (perhaps also exposed to accidental insults I haven’t mentioned, such as toxins or pathogens colliding with your metabolism, or unlucky micronutrient scarcity). The diabetician tells you that your high FPG and HbA1c are down to your genes. This has two meanings – the good news is that your personal behaviour didn’t result in your diagnosis (although in fact the odds are it has had some bearing on it), the bad news is that there is little you can do to prevent what is a progressively deteriorating condition (although we will prescribe a high-carb, high-fibre, low-fat diet that will make it worse, and drugs that won’t cure it).
         Whereas what this diagnosis should mean is this; you have glucagon dominance. You have a genetic adaptation to a diet low in carbohydrate but high in fat and protein, with periods of fasting. If you go with the flow of glucagon dominance, if you feed yourself on the foods that are glucagon metabolism substrates and avoid the insulin metabolism substrates, and if you go hungry some of the time, you will likely get better. At any rate, you won’t have a deteriorating condition that will eventually take a ton of drugs to control poorly.

Links - Hyperlipid on Inuit
Unger and Cherrington on Glucagon
UKPDS. Failure of low-fat diet and drug treatment of T2D
Westman and Vernon. Success of low-carbohydrate treatment of T2D
Lim and Taylor. Success of fasting treatment of T2D