#10: Blood Glucose Variability and My Results

Normal humans and animals for that matter have elegant systems in place to maintain homeostasis. Blood glucose (BG) is one of many parameters a normal human maintains in a narrow range (on average about 83 mg/dl) while awake or asleep, while eating or fasting, while exercising or resting, under stress or relaxed, etc. This is accomplished by the beta-cells in the pancreas which monitor the blood on a moment to moment basis for any nutrients (mainly glucose and amino acids) appearing in the blood after a meal and then appropriately secrete insulin in small bursts. The insulin travels to the neighboring alpha-cells to control glucagon secretion and then directly to the liver which removes 1/3 of the insulin and then the remainder to the rest of the body to direct all of the nutrients (glucose, amino acids, and fat) to their appropriate destinations. In persons with diabetes, the ability to maintain BG homeostasis is impaired or completely disrupted. Persons with type 1 diabetes (T1D) have to try to mimic this elegant system with a BG meter and exogenously injected insulin. The insulin injected under the skin by a T1D goes to the whole body such that the concentration of insulin reaching the alpha-cells is one hundredth of normal and that reaching the liver is one third of normal. This in addition to variations in insulin and nutrient absorption mentioned in a prior post explain why BG can vary so much despite the best efforts of the person with T1D.

I’ll be writing a future blog post “Why Insulin Preparations Do Not Mimic Normal Insulin Secretion in T1D,” to explain in more detail why BG varies so much.

This inherent BG variability also explains why attempts to lower BG toward, and especially to normal, leads to more hypoglycemia.

Measures of Blood Glucose Variability

Now on to the various measures of BG variability and my results which are shown in the table below. Some of the measures of BG variability come directly from statistics, while others were derived specifically for persons with diabetes. I have listed as many as I could find that utilize self-monitored blood glucose (SMBG) values and devised a few myself. I’ll briefly explain them without getting into the actual mathematical formulas. Also, wikipedia and other websites have definitions for most of them. The column headings from left to right are listed below along with my assessment of my results.

  1. Median – the middle BG value, such that half of the values are greater and half are less than the median. This is not a measure of BG variability, just listed as a reference. My results – The median BG of 109 mg/dl in 2007 was exacerbated after starting triathlon training and consuming sugar during training to prevent hypoglycemia. Changing to The Paleo Diet in 2012 (for 5 weeks) did not improve my BG much, but did result in a decrease in insulin requirements (discussed in a previous post). Changing to the ketogenic low carbohydrate high fat (KLCHF) diet in Feb. 2012, did result in an improvement in BG along with continued exercise.
  2. Mean – the average of all the BG values. This also is not a measure of BG variability, just as a reference. My results – same as in 1. above.
  3. Standard deviation (SD)- the most common measure of BG variability, the lower the better. My results – actually very similar to that described above for median and mean. BG variability as measured by SD was made worse by exercising while eating sugar, and better by changing to the KLCHF diet + exercise.
  4. Coefficient of variation (COV) – the standard deviation the divided by the mean, expressed as a percent. This is useful because the different conditions (type of diet, exercise or not, time period) which have different mean BGs can be compared directly by using the COV. My results – very similar to that described above for SD.
  5. Mean BG change per hour – takes into account the time difference between each BG measurement and averages the absolute change per hour. My results – again, worse with exercise + sugar, better with KLCHF diet + exercise.
  6. Mean daily BG range – calculates the maximum BG minus the minimum BG each day and then averages them for the time period indicated. My results – again, worse with exercise + sugar, better with KLCHF diet + exercise.
  7. The next four columns include measures of BG variability that I devised and are simple to understand, but a bit complicated to calculate. It assumes a linear change in BG from one value to the next which is not necessarily the case primarily because I am only measuring my BG four times a day. In addition, it assumes that I take no action if my BG is elevated (like exercising sooner than later or taking extra insulin) or if my BG is low (like taking a glucose tablet or eating). Therefore the data in these next 4 columns could represent either an underestimation or an overestimation of time and degree of hyperglycemia and hypoglycemia, but the point is to see if I am making improvements over time.
    • Number of hours per day that my BG is greater than or equal to 120 mg/dl. My results – again, worse after starting triathlon training and consuming sugar, better with KLCHF diet + exercise.
    • Excess BG – shows how much higher the BG was during the number of hours in the previous column. For example, in 2015 my BG averaged 140 mg/dl (120 + 20) for 4.1 hours a day. For the remaining 19.9 hours a day, my BG was less than 120 mg/dl.
    • Number of hours per day that my BG is less than 51 mg/dl. My results – prior to the KLCHF diet, when I had hypoglycemic episodes I had symptoms and was usually in a panic to just treat them with glucose tablets and/or food. Thus, I did not measure my BG and none were recorded for these calculations. Since starting the KLCHF diet, I rarely have symptoms of hypoglycemia, thus I’m doing routine BG measurements and finding low BG values at times. These values are used in the calculations. Of course, I treat these low BGs with glucose tablets and/or food. After starting the KLCHF diet, the number of hours per day with BG < 51 mg/dl increased as the mean BG decreased. In 2015, this prompted me to back off on trying to achieve normal BG around the clock.
    • Deficit BG – shows how much lower the BG was during the number of hours in the previous column. For example, in 2015 my BG averaged 44 mg/dl (50 – 6) for 1.4 hours a day. For the remaining 22.6 hours a day, my BG was greater than 50 mg/dl.
  8. Interquartile range – The range of the middle 50% of BG values. Mathematically it is quintile 3 minus quintile 1 where quintile 3 is the BG value above which the top 25% of values occur and quintile 1 is the BG value below which the bottom 25% of values occur.


BG Variability

All of the above measures of BG variability decreased (improved) after starting the KLCHF diet except for the increased occurrence of asymptomatic hypoglycemia as the mean BG approached normal (83 mg/dl). This would be expected given the current technology of using a BG meter and insulin analog injections as a surrogate for a normally functioning human pancreas.

Does Blood Glucose Variability Matter in T1D?

The question is, does a greater glucose variability (GV) lead to more diabetic complications or hypoglycemic episodes independent of mean BG or HbA1c?

First, persons with T1D should be familiar with the results of The Diabetes Control and Complications Trial (DCCT) completed in 1993 because its finding that diabetic complications are reduced when the HbA1c is <7% is still used as the American Diabetes Association’s HbA1c target. The DCCT was the first large randomized controlled trial designed to test whether intensive insulin therapy could prevent or slow the progression of microvascular diabetic complications. After 6.5 years of follow-up, the authors concluded that “Intensive (insulin) therapy effectively delays the onset of and slows the progression of diabetic retinopathy, nephropathy, and neuropathy in patients with IDDM.” However, this reduction in diabetic complications from intensive insulin therapy was associated with a three-fold increase in severe hypoglycemia.

Figure 1A (shown below) from the DCCT shows the significant improvement in HbA1c with intensive insulin therapy.


Figure 5A (below on left) from the DCCT shows the relationship between HbA1c and rate of sustained progression of retinopathy and Figure 5B (below on right) shows the relationship between HbA1c and rate of severe hypoglycemia among those receiving intensive insulin therapy.


Those patients with lower HbA1c had less progression of retinopathy, but suffered more episodes of severe hypoglycemia (in which assistance was required or resulted in seizure or coma).

A subsequent analysis of the DCCT data failed to show a relationship between GV and diabetic complications independent of mean glucose or HbA1c. Note the DCCT was not designed to evaluate GV in the first place.

An extended follow-up of the DCCT cohort then reported that mean daily glucose as well as pre- and postprandial hyperglycemia (PPG) predicted cardiovascular disease in the same cohort.

I reviewed numerous other studies looking at GV, some showing that GV could independently predict diabetic complications and others unable to find an association. I also found several studies linking increased GV to an increased likelihood of subsequent hypoglycemia. So rather than review multiple studies with conflicting findings none of which looked at persons with T1D following a KLCHF diet, I’ll refer you to this article which reviews the topic in just three pages.


I think GV is as important as mean BG or HbA1c as a target of therapy in T1D. Published studies and my results shown in the first table above, both illustrate that lowering the mean BG toward normal increases the number of hypoglycemic episodes.

In my opinion, the KLCHF diet improved both my mean BG and GV. Another major contributor to my GV is exercise via its effect on insulin sensitivity. My efforts to adjust the type, duration, and consistency of exercise over the past few months has improved my BG coefficient of variation to an all time low of 35%. And finally, there is no published study that can show a difference in complications or lack thereof between normal versus near-normal BG control in persons with T1D. In my case, I think striving to achieve normal BG actually increased my GV as well as the frequency of hypoglycemia.

In persons with T1D, the glycemic target should be individualized to as close to normal as can be safely accomplished which means minimizing hypoglycemia given that up to 10% of those with T1D die from hypoglycemia.



  1. Pingback: #25 May 2016 Update on My T1D Management | Ketogenic Diabetic Athlete
  2. Pingback: #31 October 2016 Update on My T1D Management | Ketogenic Diabetic Athlete
  3. Pingback: #32 November 2016 Update on My T1D Management | Ketogenic Diabetic Athlete
  4. Pingback: #33 December 2016 Update on My T1D Management | Ketogenic Diabetic Athlete
  5. Pingback: #36 February 2017 Update on My T1D Management | Ketogenic Diabetic Athlete
  6. Pingback: #38 March 2017 Update on My T1D Management | Ketogenic Diabetic Athlete
  7. Pingback: #39 April 2017 Update on My T1D Management | Ketogenic Diabetic Athlete
  8. Pingback: #40 Metformin For Patients With Type 1 Diabetes & May 2017 Update on My T1D Management | Ketogenic Diabetic Athlete

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s