NZ case study; A citizen scientist controls autoimmune diabetes without insulin, with a low carb diet, a glucose meter, and metformin.

The case study is a very important type of medical publication that’s overlooked in this age of big data. Unlike large statistical studies, which tell you the probability of something happening, the case study tells you whether something CAN happen at all, and under exactly which circumstances it has happened.

Case studies answer questions like “Can autoimmune diabetes, with lower insulin production, be managed long-term without insulin?”

Yes, it can, and this is described in full detail and a clear and simple style in a new case-study from Christchurch.[1]
2017 Nelson Jacobs Case report Management of autoimmune diabetes without insulin

This is published on, an online data repostitory set up by people involved in CERN and other places. Warrick Nelson, the first author is the patient and is an operations manager at Plant and Food Research in NZ. The second author is his doctor. This is citizen science.  We love it.

So onto the topic of diabetes, and  management of the condition with low carb diets……

We reviewed the strong evidence for low carb diets in diabetes management in the New Zealand Medical Journal in a 2016 review cited in the current paper.[2]

Its case study is a great example of how the wisdom of Citizen Scientists, equipped with mass produced measuring devices and, in this case, a proven medicine, can discover the one way to treat a disease. It is written up by the patient and his doctor, who was wise enough to recognise this as the teachable moment it is.

The patient first presented with type 2 diabetes, non-alcoholic fatty liver disease, insulin resistance, and a high risk TG/HDL ratio. He was highly motivated and was able to lose weight and correct all of these issues with a better diet and metformin alone.

In early 2013 the patient, a white male 53 years of age, was diagnosed diabetic following a routine screening test, based on fasting glucose of 10.8 mmol/L and HbA1c of 58 mmol/mol (7.5%). Mild overweight, especially abdominal, mild hypertension, mild dyslipidaemia and elevated liver function panel (Table 1) indicated the onset of T2D. Metformin 500mg twice daily and 10kg weight loss were indicated. Advice on weight loss was primarily calorie control (particularly between-meal calories, as the patient self-reported a tendency to snack in the evenings on biscuits). Snacks were recommended to be reduced and cheese or almonds suggested as options. The patient was highly motivated and within nine months HbA1c test showed a pleasing result at 37 mmol/mol (5.5%) and BMI of 26 (Table 1).

A year later and the patient has signs of Type 1 Diabetes, this pattern is known as Latent Autoimmune Diabetes of the Adult (LADA).

However, a year later a routine HbA1c of 83 mmol/mol (9.7%) occurred, associated with unexplained weight loss and fatigue. By this time, the BMI target was very close to achievement. Anti-GAD tests were ordered on the assumption it was likely the patient had resolved to a Type 1 diabetic pattern. The result was unrecordably high autoantibodies and the patient was referred to a specialist diabetes clinic to begin insulin therapy.

But in the meantime the patient has begun testing post-meal blood sugars and eating accordingly – note that the doctors or dietitians had already recommended eating almonds or cheese instead of biscuits in the weight-loss phase, so he has a general idea of the alternatives.

In the interim, the patient had begun using a home glucose meter with nearly immediate resolution of blood glucose from the 15-23 mmol/L range to sub-10 postprandial tests (Figure 1). This was achieved by an immediate drastic reduction of bulk dietary carbohydrates, primarily experimenting with reducing carbohydrate intake to achieve acceptable postprandial glucose levels. A food diary indicated sub 100g carbohydrate per day, and more stringent dietary intervention from January 2015 suggests <75g/day is being achieved.
The patient attended two clinic visits, but expressed reluctance to begin insulin therapy while home blood glucose testing indicated dietary interventions were working. At this time, HbA1c had already reverted to 49 mmol/mol (6.6%) and by the second visit, 3 months later, was down to 38 mmol/mol (5.6%). Further autoantibody tests indicated both IA2 and ICA at the top of the measurable range. A fasting insulin test returned 66pmol/L.

But can it be healthy, eating such a restricted diet?

The patient reports the new diet is completely satisfying, tasty and easy to manage other than when faced with commercial food offerings eaten away from home. In particular, airline and hospital “diabetes” choices are completely incompatible with a low carbohydrate diet. The patient reports completely removing wheat flour products (such as breads, cake/biscuits, pasta, couscous), potato, rice, maize and other obvious high starch products (including gluten free options such as quinoa and buckwheat). The dietary bulk provided by these foods is largely replaced with salad and vegetable as appropriate. High carbohydrate vegetables (such as carrots, pumpkin, green peas) are not eliminated, but are eaten in moderation.

But surely the results are not sustainable?

Quarterly HbA1c tests have remained at ≤40 mmol/mol (≤5.8%) for two years on this diet. The patient has felt confident to reduce the frequency of home blood glucose testing to one day per week of pre/post prandial testing for one or two meals on that day plus occasional testing following introduction of new food items.

LADAThe authors discuss how failing islet beta-cells in late-onset diabetes tend to keep producing low levels of insulin (we think this might be more likely if they’re protected from lipotoxicity – excess fat – and glucotoxicity – excess sugar; the weight loss phase where NAFLD was reversed would have helped avoid the former, the low carb diet the latter).[3] We have theorised previously that the different gut hormone responses to a low carbohydrate meal, which include a lower release of glucagon and a higher release of somatostatin 28, can contribute to controlling post-prandial glycogenolysis, lipolysis, and proteolysis at a lower insulin level.

Reading this report, we are bystanders at a revolution in medicine. This patient has decompensated diabetes, and it’s unlikely any variation on a higher carbohydrate diet would be able to control their blood sugar without insulin. In fact, the hospital diabetes diet is completely unsuitable for this purpose. If the low carb diet is suitable, and the hospital diet isn’t, for this critical aspect of diabetes management, what then?

Hospital diabetes diets are generally the high-carb, low fat, dietary guidelines type of diet that was introduced into diabetes care without any RCT comparison with the low carbohydrate (<130g/day CHO) or very low carbohydrate (<50g/day CHO) diabetes diets that have outperformed it in RCTs ever since.[2] These diets are normally grain based, nutritionally poor, and high in glycemic load. Their only concession to diabetes management is that the carbs are counted and (supposed to be) spread throughout the day. They require higher insulin dosing than would be the case with low carb diets, and thus make it harder to maintain blood sugars safely within the normal, non-diabetic range, so higher cut-offs are accepted as “good management”.[2] This case, as well as the evidence we included in our review, and a more recent RCT of a low carb diet (70g/day CHO) for type 1 diabetes published by Jeremy Krebs’s team in New Zealand,[4] show that this approach needs to change.


[1] Nelson W, Jacobs P. Management of autoimmune diabetes for two years without insulin treatment: a case report. May 7 2017 doi:10.5281/zenodo.572338.svg
2017 Nelson Jacobs Case report Management of autoimmune diabetes without insulin

[2] Schofield G, Henderson G, Thornley S, Crofts C.
Very low-carbohydrate diets in the management of diabetes revisited.
NZMJ. 2016;129:1432.

[3] Schofield G, Henderson G, Crofts C, Thornley S.
What are we to think when results from mouse research contradict those from human experiments and clinical practice? Nutrition & Diabetes (2016) 6, e224; doi:10.1038/nutd.2016.31

[4] Krebs JD, Parry Strong A, Cresswell P, Reynolds AN, Hanna A, Haeusler S.
A randomised trial of the feasibility of a low carbohydrate diet vs standard carbohydrate counting in adults with type 1 diabetes taking body weight into account.
Asia Pac J Clin Nutr. 2016;25(1):78-84. doi: 10.6133/apjcn.2016.25.1.11.




13 Comments on “NZ case study; A citizen scientist controls autoimmune diabetes without insulin, with a low carb diet, a glucose meter, and metformin.

  1. This guy’s fasting insulin of 66 pmol/L translates to 9.5 microIU/mL for the units used in the United States. Very few doctors would call this guy type 1 with such a high fasting insulin. Maybe the puzzle here is that he was tested for type 1 antibodies despite not having a low fasting insulin. Maybe there are more patients like him out there.

    • Interesting – but he had the high HbA1c and weight loss. Maybe they thought the insulin was not compensating for any IR he still had because of those symptoms?

      • A further thought is that this is New Zealand, and an insulin reading probably takes just as long as the antibody test, so they did both at the same time, because of the urgency of the situation. So when they ordered the antibodies maybe they didn’t know about the insulin.

    • Yes, BUT… We don’t know this lab’s typical or normal RANGE of values for the C-peptide result. (And, I apologize for not having the conversion factor to ng/mL at hand to convert it to values I am accustomed to [0.78 – 5.19 ng/mL].)

      Also, a normal fasting level of C-peptide would be expected to reflect the lowest values in the normal range of values…so long as fasting BG was also low normal (maybe 75-80 mg/dL). Ultimately, the value is far less meaningful if we cannot compare it to a simultaneous measurement of blood glucose. If in this case his BG were, say, 70 mg/dL, that might be interpreted to mean that he was pumping out insulin at a high rate just to keep it “normal” (70 mg/dL), which for “normal” metabolism is not necessary (i.e., in fasting, the level of C-peptide is at the very lowest end of the lab’s range). Thus, it would be easier to suspect DM2, which in the early phases requires/produces a lot of insulin just to keep the BG normal.

      If the BG measurement were 300 mg/dL, we could know that even with substantial insulin production in this case, it just wasn’t enough to produce normal BG levels, or certainly for some reason wasn’t working as expected. This kind of insulin production together with this BG level would be rare for DM1 but not so rare for DM2.

      If his BG were 200 mg/dL, and his C-peptide level was near or below the bottom of the usual range, we could know that he was not producing much insulin (as the BG level indicates) and DM1 would be one logical conclusion.

      It is always possible to incorrectly interpret the C-peptide result if one does not know what BG level corresponds to the reported level of insulin production. They “normally” vary and work synchronously, not independently. It is the effect or result (that is, the BG level present) of the insulin produced that determines the latter’s functionality and sufficiency or normalcy.

      • Dr De Loach
        Thank you for your interest. The diagnosis was original T2DM based on HbA1c and fasting glucose, resolved for 18 months on metformin and minor diet change with weight loss, but thought to be T1D when a follow up HbA1c 18 months from original diagnosis returned 83 mmol/mol (about 9.7%) and A antibodies requested. Diagnosis changed to T1D based on these results.
        Reference range for insulin from this lab is 10-80pmol/ml (1.4-11.2uU/ml), so reading of 86 is 12uU/mL (not sure why different to Jeremy – lab notes say convert pmol/ml*0.14=uU/mL). Unfortunately no glucose done at the same time (problem of self requesting a test)!
        Last year (2017), fasting C-peptide + glucose done. C-pep=0.914 nmol/l (range 350-750) and glucose 9.3mmol/l (~170mg/dl). HOMA2 calculation gives %B=50.3, %S=40.9 and IR=2.44. As you note, strongly suggestive of T2D but with highly positive range of antibodies seems to automatically make it T1D. Suggests LADA – T1D with initial features of T2D.

  2. I ordered and paid for the insulin test on my own initiative. Clearly I’d survived without exogenous insulin so far and wondered just how much I was still producing. The insulin test is a simple blood sample and results within a few days, cost me about $50 or so.

    • Thanks for clearing that up Warwick – and thanks for such an informative paper.
      Our readers might be interested in how you ordered the insulin test – most people give up when the doctor says it’s not an option, but a fasting insulin test or a 2-hour OGTT result would be very useful.

      • Thanks for the positive comments.
        My local medical testing laboratory offers it. There is a very limited range available without medical request, but insulin is one. OGTT is relatively easy to do at home with a glucose meter – might not be as precise as a lab test, but we’re talking about broader trends rather than absolute numbers. I would be very wary of feeding myself 100g glucose now though! I did it when I first got my meter but drank equivalent of 50g as sucrose (I used ginger cordial but many options) so not quite the same as pure glucose. I got a nice trend line measuring every 15 minutes.

    • Thank YOU! There was no Reply button to your comment today, so I will post it here.

      I did understand the somewhat unusual, even perplexing evolution of the current diagnosis. From my point of view, what’s realistically important, whether one has DM1, DM1.5, DM2, or DMgestational, is the best chronic normalization of BG levels possible. And, as the article states and graphs, the fewer the carbs, the more normal the BG. The more normal the BG, the fewer the short-, medium-, and long-term complications of chronic hyperglycemia to be expected. The fewer the diabetic complications and the subsequent reduced anxiety over the arrival of diabetic complications, the higher and more fun the quality of life for those of us with DM of whatever type.

      Warm wishes to you! Perhaps we will run into each other during my upcoming visit to Auckland (30 October – 4 November, 2018).

  3. So if people are interested in lab tests outside the standard panel and have a bit of spare cash, they can ring their local laboratory and enquire? That’s worth knowing.
    2 hour insulin after OGTTs, at least in people eating the normal diet, is going to predict if someone is at risk of type 2 diabetes or cardiovascular disease etc ahead of the usual signs.
    I actually think that the fasting TG/HDL ratio is a very accurate proxy of 2 hour insulin based on my analysis of tables in papers that supply values for both (a doubled TG/HDL = a doubled 2 hour insulin close enough) but there’s no meta-analysis or direct study of the relationship I can cite to prove the point yet.

  4. Pingback: NZ case study; A citizen scientist controls autoimmune diabetes without insulin, with a low carb diet, a glucose meter, and metformin. | Karen'sLittleWonders

  5. Pingback: NZ case study; A citizen scientist controls autoimmune diabetes without insulin, with a low carb diet, a glucose meter, and metformin. - Low Carb Diet HQ

  6. Since it’s challenging to obtain the daily dose of
    plant sterols through food alone, many products are fortified with them.

    Exercise actually boosts the size of the protein particles that carry
    the bad and the good cholesterol. Simplify and conquer
    – With Omega 3 fish oil in a 1200 mg dosage, I can do more for my cardio vascular health in one shot than taking 3 to 4 supplements.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: