Engineering a Cure for Type 2 Diabetes

Engineering Nutrition?

With the failure of forty years of dietary guidelines to arrest or improve the incidence of diabetes and obesity, new thinking and approaches are needed. Applying an engineering mindset to nutrition has attracted attention as some of the new thinking has emerged using root cause analysis and other engineering tools. This has resulted in new insights for the medical and nutrition communities.

This is not really new, I pay homage to doctors like Dr Bernstein who trained as an engineer first, then as a trained doctor realised how controlling diabetes was like an engineering control problem.

Recently, however, as a recovering type 2 diabetic, I plotted my HbA1c against the results of a long-term vegan ‘cure’ for diabetes study to see how it compared. I was astounded by the superior result and tweeted that it was a fifteen sigma improvement. While not really correct, it got me thinking of my recovery in terms of engineering control theory and quality management. 

Putting aside whether a cure is possible (for type 2 diabetes) and considering treatment, what if we view diabetes as an engineering control problem and applied control charting to understand the quality of different management options? Note that while I have type 2 diabetes, the glycaemic control problem is common to type 1 and so much of this analysis also is relevant to them too.

Broken Control System

Glucose comes from sugar and other carbohydrates (carbs) like starch from bread, rice and pasta. Your body uses about 130g of glucose a day (about 33 teaspoons). Normally, there is no more than about one teaspoon of glucose in your blood at any one time. Simply, if there is not enough glucose in your blood, you can black out or die as your vital organs cannot function. As your muscles, brain and other organs consume glucose as fuel, your liver, pancreas and digestive system release hormones including insulin to regulate glucose to a tightly controlled level. That magic number is normally about 5.6 mmol/L (or 100mg/dL depending upon the units you use).

You might wonder, what will happen if you don’t eat any carbohydrate? Fortunately, probably as a result of adaptation, the body is fine as it can make what you need from other sources. This happens mostly in your liver. It is called gluconeogenesis or GNG for short.

Essentially with diabetes, the control system that reduces blood glucose (BG) is broken. The homeostasis (self-regulation) of your BG is ineffective because your body’s response to insulin (which lowers BG) is diminished (called insulin resistance) and/ or your ability to produce insulin in response to carbs is insufficient to lower BG quickly enough. For type one diabetes, insulin production is at or near zero.

Consequently, glucose that your body gets from carbs (or makes in the liver through GNG) will raise your BG and it will only fall slowly because your body is unable to produce or respond to insulin properly.  So BG is easily raised but slowly and poorly lowered.

Conventional Diabetes Management

Let’s leave the medical theories about why the system is broken alone for the moment and assume we have to do the best with what we have got.

Conventional diabetes management seeks to lower your BG towards normal but not so that it drops too low. This is done by exercise (to consume glucose), diet and medications that replace insulin, reduce glucose production or eliminate glucose from the body.

In conventional diabetes management, juggling these factors on a daily basis is hard and is the focus for someone with diabetes. Every three months you go to see your doctor to see how you are doing overall and to see if your medication should be adjusted.

Unfortunately, it is hard to achieve and maintain this great juggling job. It is hard to replace a well working system in the body once broken. The typical person with diabetes has BG that, on average, is too high. It may also drop too low with too much medication leading to coma or death. High BG is associated with all of the ill effects that people with diabetes suffer including blindness, kidney disease and amputation. For most, eventually, doctor’s visits mean an inevitable adjustment upwards in medication and higher BG. High BG results in deterioration for a person with diabetes over time, more medication, more complications. Diabetes is therefore regarded as a chronic disease with an inevitable worsening progression.

With that prognosis, it makes little sense discussing getting back to normal BG. It makes little sense to see this as a control process that can be brought under near normal control.

THAT IS (FORTUNATELY) COMPLETELY WRONG!                                

HbA1c Measurement

I mentioned a three monthly visit to your doctor. Your BG changes throughout the day. In order to assess your overall BG control, a test measuring ‘haemoglobin A1c’ (HbA1c or just A1c for short) measures how ‘sticky and sugary’ (glycated) your blood is. As blood cells turn over every three months, A1c gives you about a three month average of your BG control.

I mentioned that if your systems were working properly, your BG would average about 5.6 mmol/L (or 100mg/dL). It turns out that this corresponds to an A1c of 5.1% (or 33 mmol/mol). This is an average for the healthy population or ‘population mean’. Statistically, the standard deviation from the mean is about 0.5% and it is deemed that you have prediabetes if you exceed the mean by one standard deviation or >5.6%. Similarly above about two standard deviations (>6.1%) you are diagnosed as having diabetes. The higher above one standard deviation you go, the worse becomes the health risks of diabetes.

Control Charts

Control charts are a tool used in engineering and management science to help us understand what is happening with a process. Essentially a control chart gives you a measure of how close a controlled system is performing to expected behaviour (the mean or average target for a parameter) when considering its deviation from the desired behaviour. Control charts give you a measure as to the quality of the outcome of a process and should help decide what you may need to do to bring a process back into control.

You can read about using control charts here.

If the aim is for a person with diabetes to approach the health of a ‘normal’ person, then we must restore the control as near as possible to the BG of a healthy person. A control chart type of methodology is used in some glucose monitoring programs to measure the quality of control of daily BG.

So when looking for long term control/ improvement, why not plot the mean of HbA1c and its standard deviations for healthy people? We can then use the control chart methodology as a yardstick to see how various treatments compare and to hopefully gain better BG control towards a cure.

Diabetes Control Chart using HbA1c

I have reproduced the results of a study on diabetes as a control chart. That study looked at about 49 vegans and another 50 people on a conventional diabetes diet. You can read this study here.  I have added to that a plot of my history on a low carbohydrate diet. I have added in the bands of standard deviations (s, 2s, 3s, etc) in bands of colour from green to red.

Some points about this control chart in general:

1. Excellent control would see points close to 5.1% and ideally in the light green zone within one standard deviation (±s).
2. In control chart theory, any data point more than three standard deviations (±3s) is deemed ‘out of control’. Something is really wrong with the system and control process itself for this to occur.
3. Not one of the measurements is below the population mean of 5.1%

Conventional Diabetes Diet

This diet was a low fat, calorie deficit diet designed for weight loss. This gave the worst outcome. At the end of the 74 week period, the average A1c results were nearly above where they started. No average A1c was better than 5s. By the end of the trial, only about half of the participants were adhering to the diet. This was despite cooking lessons, weekly meetings with a dietitian and other intensive assistance. This diet was high in carbs as they are 60-70% of total energy.

Vegan Diet

The vegan diet lacked meat, eggs and dairy but was not calorie restricted. This gave a slightly better outcome. No average A1c reading was better than 4s. By the end of the trial, only 44% were still adherent and the outcome was beyond 5s. This was despite similar intensive assistance to that given on the conventional diet. Probably, as a result, some of the gains in A1c made earlier in the trial were lost and the vegans also deteriorated again. Had the trial and the upward A1c trend continued, it appears that the vegans might also have ended up worse than they started. This diet was very high in carbs being 75% of total energy.

LCHF/Keto Diet

My diet lacked carbs. No sugar, rice, pasta, bread, sugary fruit and starchy vegetables. I also drank alcohol sparingly. Most people with diabetes are advised to eat between 200g and 300g of carbs per day spread out over the day. I aimed at first for less than 50g per day (<10% carbohydrate) and after about three months I was reliably lower than 25g (<5% carbohydrate) per day. This normally would be a ‘keto diet’ however it is hard for people with diabetes to stay in significant ketosis without extended fasting so I prefer to call it LCHF. I also did practise intermittent fasting simply because I was not as hungry as I was with a higher carbohydrate diet. Many people report this. Typically this involved not eating breakfast so that it was 16 hours after the previous night’s dinner before I ate the next meal.

There was no assistance from a dietitian or cooking lessons for me. I did read the free information on the dietdoctor.com website to get the bulk of my nutrition from real food sources (meat, eggs, fish, fruit, vegetables, nuts & dairy) that were low carb. Unlike the diets in the trial, adherence was easy for me, although I had to unlearn a lot of ‘advice’ that dietitians had previously told me on my way to developing diabetes. Unlike the study diets, I ceased three diabetes medications after three months but then began taking one-quarter of the dose of metformin again at that time.

I did no appreciable exercise like running, swimming, cycling but took an occasional walk. In the first six months I easily lost about 12KG of weight, moving from obese to overweight. My weight has been quite stable since then.

Unlike the other diets of the study and my previous diabetes history, all my readings (except baseline) were within 2s and went below s before the year on LCHF was finished. Clinically, below 2s is pre-diabetes and below s is non-diabetic so I have been very happy with that result. The downward trend was recently confirmed as still occurring with a recent estimate of A1c from my glucose meter readings.

Engineering Analysis

Straight away we can say that the study diets are ‘out of control’. With no points less than 3s there is little prospect of either ‘process’ (diet) bringing control to equal the population mean. Further with all points 4s or higher, the mean (goal A1c of 5.1%) will never be reached. Quite simply, something is causing the A1c to be unacceptably high that the process being used cannot overcome. From an engineering standpoint, these are defective processes that cannot achieve the target. The trends were initially towards but end up moving away from the target long term. Management theory would tell you that the individual in the process (person with diabetes) will be powerless to achieve control. It is ridiculous to blame the person with diabetes for this result yet many of us blame ourselves. The theory says that to continue to expect reasonable control to the target wanted is foolish. You must use a different process or make some other significant change to the system.

That is not the case with the LCHF diet. All points are within 2s, some s, and we have a trend that may eventually result in the target being achieved although none of the measurements so far have been below the target.

Engineering Solution

If I were presented this as a control system problem I would immediately conclude that there was an unaddressed control offset, especially in the study diets. The engineering solution would be to apply ‘Integral Control‘ to attack that offset so that the control range is eventually brought closer to the target. This means relatively slowly increasing or reducing the level of the controlling factor until control can be achieved.

Further, both diets represent a perturbation in the system that slowly corrects back to its original level. Like throwing a stone in a pond. The ripples eventually subside and things head back to what they were- in this case, a level that is too high.

We know that carbs, be they from the liver (GNG) or diet, raise BG and A1c in people with diabetes who do not have enough (or do not respond properly to) insulin. The amount of carbohydrate is the controlling parameter for BG and A1c. It is straightforward that a solution is to reduce carbs permanently- but by how much? For me, the LCHF result shows that even if we drop dietary intake to a minimum, the target would still not be reached quickly due to their production by the liver (from GNG).

So a very apt control analogy is a sink with a small inflow of water from the bottom (GNG), a drain draining away by a controllable flow (insulin action and exercise), and a tap with the ability to put in a variable inflow which by eating carbs could be continuous if spread into small meals, large and rapid if a lot of carbs (say sugar) is consumed or minimised if restricted.

Now if we want to keep the sink at a certain level (say half way) we can exercise to drop BG and eat fewer carbs to lower the level. If we leave the tap running at a rate that exceeds the draining rate or suddenly empty a large bucket of water into it, the sink fills and we will now be permanently above the level we want. This is what we see with conventional and vegan diabetes management in the study. In this situation, it is common sense to turn off the tap- carbohydrate restriction. 200 to 300g of carbohydrates per day is the problem in this control system.

Exercise Helps but Diet Rules

Exercise is a help but consider that the average person must run about 7 km to ‘burn the carbs off’ from a 500ml serve of coca cola. Even if you do run the 7km, in the time between drinking the drink and completing your run, those carbs are giving you high unhealthy BG. Better just not to eat or drink the carbs in the first place. You cannot outrun a bad diet.

All of the diets have too many carbs for the available and effective insulin to bring down BG to normal metabolic levels and that explains why the target was never reached by any of them.

Reaching the Target

Unlike the study diet, we should expect the LCHF diet might reach the target in the next nine months or so if the present trend continues.  The simplest course of action for the LCHF diet would be to keep going and see if the system settles to the desired target. If it does not or if a quicker result is wanted, other interventions could be tried to reduce carbohydrate including longer fasting, increasing exercise, upping metformin dosage or looking for another metabolic option. So now, as a vegan doctor (Dr Joel Kahn) commented to me upon looking at my results on Twitter, maybe slow and steady wins the race? That might well be the first of his advice I have ever taken.

Am I a Special Case?

At this point, you may be wondering if carbohydrate restriction might help your diabetes or am I a one off? Let us explore that. My results prior to carbohydrate restriction were consistent with the conventional diet people from the study. The best HbA1c I saw was 7.3% and as you can see below, carbohydrate restriction was the difference beginning around month 31.

The value of a case study is that it shows what CAN happen. There are no guarantees, but given similar circumstances to me, yes this can happen for you. Many other people report that it happens for them. In fact, we would expect it to happen from the biochemistry and control theory I have explained. This is even though everyone with diabetes is a little different. It means your mileage may vary.

Biochem is complex. Perhaps the major appeal of LCHF to an engineering mind is that, based upon engineering theory, it makes perfect sense. Dietitians are constrained by a myriad of epidemiological studies which show increased risk of this or that from doing that or the other thing. If you accept that A1c is a measurable proxy for the underlying health issues of diabetes, clarity to focus on the job of controlling A1c occurs and carbohydrate restriction is obvious. Once that is done, focussing on optimising diet within that constraint is the task. This fits nicely with the theory of constraints as a way to tackle complex systems.

LCHF, Vegan or Conventional Diets?

The vegan diet did perform better than the conventional diet in the study but both were a control chart fail. It is however theoretically possible that one of the 49 vegans achieved similar results to me. My result towards the end shows that my A1c was about fifteen standard deviations below the vegan mean. In other words if we assume a normal distribution and there were 100,000,000,000,000,000,000,000,000,000,000,000,000,000 vegans in the study, we could expect about one to have results as good as mine. Unfortunately, there were only 49 vegans in this study. This is a time when an n=1 (me) is statistically significant.

To be clear I am not saying that a vegan diet could not achieve the same result, but it would have to be low in carbohydrate and total energy so a vegan (or any) starvation or fasting diet would probably also work.

If common sense, the engineering theory, my simple Biochem explanation or my results do not explain why a carbohydrate restricted approach is best then read this paper. An excellent (and more complicated) comparison between the Keto (LCHF) and vegan approaches to managing diabetes is available from Marty Kendall’s website. You will also find a lot of other excellent information on nutrition there should you be concerned that restricting carbs may put you at risk of nutritional deficiency.

The Vegan propaganda machine is fond of saying that restricting carbs (the keto diet) masks the problem by addressing the symptoms whereas only the vegan diet ‘cures the disease’. Based upon the study we looked at, it appears to be an untrue claim. I don’t care whether you eat live chickens or just grass to avoid animal harm, the first thing that someone with diabetes should do is minimise their carbohydrate intake. If you must eat some, then not too many and make sure they are ‘complex’ and unrefined.

Dietitian Says ‘No’

So what if you see a dietitian and they try and dissuade you from a carbohydrate restricted approach. They may have the following objections to which I give you some answers:

1. You need carbs and your diet will lack fibre and vital nutrients from foods you will exclude like whole grains.
   Answer: Some fats and proteins are essential but carbs are not. Even if you could have zero carbs in your diet, your body makes them (via GNG). If fibre is of concern then eat more low carb vegetables. Vital nutrients? See Marty Kendall’s website. If a dietitian can’t give you a healthy carb restricted eating plan, time to walk!
2. It helps some people but people can’t stick to it in the long-term. We also don’t know how safe it is in the long term.
   Answer: Well what if a person it can help is me? Shouldn’t I try it? Looking at the conventional and vegan diets in the study, adherence was also less than 50%. Adherence is a matter for any way of eating and it is up to you. You don’t have to be a statistic. Finally, what does the long-term look like if your A1c stays at ~6, 7 or 8% and above? The risks of a high A1c are very well known. If LCHF is a devil, it is the devil you want to know.
3. Keto? Low Carb? Control charts? [Insert other doubt raised here]? Do they have any evidence of success from a study in a peer reviewed journal? My clients have excellent success on [insert a diet/ program here] instead.
   Answer:  Please give me evidence of a study showing [insert their diet/ program] can achieve an A1c approaching 5.1%. Please give me evidence of the success rate of your clients achieving a sub 5.6% A1c.
4. On LCHF/ keto you are limited. Studies show that eating [insert food of concern] or not eating [insert dietitian ‘superfood’] will make you die sooner.
   Answer: Have you ever seen someone on dialysis or with a diabetic foot? It is your job to give me a diet for normal blood glucose, then we can optimise it for other concerns. Do your friggin’ job and shelve your dogma.

The system is failing all of us. More of us are getting obese and diabetic following the standard way of doing things. I developed diabetes on a near exemplary low-fat diet. I can only encourage you to be a robust health consumer. You should not assume that in the face of the diabetes epidemic that has grown under national eating guidelines and dietetic advice, that the experts have it right. Diabetes takes no prisoners and you shouldn’t compromise your outcome just to be nice to a health professional.

Time for Dr Google?

Dietitian’s organisations lampoon ‘Dr Google’ just like clothing retailers said people would never buy clothing online. Honestly though, if you are seeing a dietitian who is not on board with carb restriction for diabetes, you are wasting your precious time and health.

If you can’t get proper help from a local professional then there are sites like dietdoctor.com, forums like the ketogenic forums and facebook groups like type 2 diabetes straight talk or type one grit. If you are in the US, Virta’s service could be a good choice. Any of these would be preferable to a low carb inexperienced dietitian!

If you DIY then be conscious that some medications that you may be on (notably sulphonylureas and insulin) can be very dangerous to take if you suddenly reduce your dietary carbohydrate. If trying this, you should consult your doctor to clear or adjust your medications appropriately.