Monitoring results

I just received the latest bloodwork results after 3 months of the new drug regime.

Testing hypotheses for high morning Blood Glucose

I've focused more on measurements of Hemoglobin A1c than blood sugar, as that gives a time average indication. Even so, my A1c numbers, while good for a diabetic at between 5% and 5.4%, are not great by Steve's targets. A value of 5.4% corresponds to an average BG of 108 mg/dL. Going the other way, having an average BG less than 100 mg/dL means an A1c of 5.1. The fact that my averages are much higher than my fasting levels during the day suggests that my post-meal levels, and early morning levels, are very much higher than I'd like. I've got a series of experiments underway looking at BG after eating, but what about overnight highs in Blood Glucose?

So I think this is a perfect setup for adaptive management.

There are two hypotheses for high morning blood glucose in a Type 2 diabetic. The "hormone hypothesis" posits that an increase in growth hormone production in the middle of the night triggers a cascade of events leading to the release of glucose from the liver. Presumably this is to set one up for a vigorous start to the day. In a healthy person this extra glucose triggers an insulin response, and blood glucose remains steady. In a person with insulin resistance (like me), the insulin response is not effective, and blood glucose rises. The second hypothesis is "The Somogyi effect", where low blood glucose values overnight trigger the the release of glucose from the liver to avoid too low blood sugar values. That also involves hormones, but they are activated for different reasons.

Which of these hypotheses is true does matter; under the 2nd hypothesis I should be able to avoid high morning sugar by making sure my overnight sugar levels don't drop too low. I could eat a snack before going to bed, for example. Under the hormone hypothesis there's not much I can do except to try harder to ameliorate the insulin resistance. I could start taking Metformin, which acts in two ways, first by suppressing liver production of glucose and second by reducing insulin resistance. Both of those actions should act directly on the dawn phenomenon.

I can also learn which of these hypotheses is true by monitoring my blood sugar. Ideally, I'd measure my blood sugar every hour or so all night ... hmm. That sounds like a sucky experiment. According to this article, I can get away with just measuring at bedtime, 3 a.m., and the morning.

So I tried it for 2 nights, and I think I've discovered a 3rd hypothesis, the "Drew Tyre Effect"! On the first night, 9pm 101, 3am 116, and 6am 122! OK, that's consistent with the hormone hypothesis, but maybe I missed a drop in BG between 9pm and 3am. So last night I took the middle measurement earlier, and 9:20pm 112, 12:25am 110, and 6:40am 116. My BG is high all night. 

So, a bit of reading on what stimulates Cortisol levels is warranted. As it happens, sleep deprivation stimulates cortisol production, so this could be an instance in which the observation of a process affects the process itself. I need one of those continuous glucose monitors!

Here's another hypothesis: I've been taking a fish oil supplement for the past 3 months after learning that I was low in tissue levels of EPA and DHA. However, it turns out that fish oil supplements can blunt insulin response and increase resting glucose levels. So looks like I have to wait 18 weeks and then try this experiment again.

And another hypothesis! I'm swimming in the damn things ... It turns out that taking a statin can interfere with blood glucose control. I starting taking 10 mg of Crestor every evening about the same time I started taking fish oil. Well, the results of my latest bloodwork will be very interesting. 

Famine & the evolution of diet

One cornerstone assumption of a lot of diet theories based on emulating hunter-gatherer diets is that famine was a constant companion. Predictable seasonal famines occur in all environments, the dry season in the tropics and winter in the high latitudes. The just-so story goes like this: in a predictable feast-famine environment "thrifty genes" that take advantage of feast times to store fat in preparation for the lean times would be selected for. And when you take a population that has these thrifty genes and put them in a constant "feast" environment, they get fat and develop diabetes. Like the Pima Indians in North America*, and the Tokelau people in the south pacific. To avoid triggering these "thrifty genes", paleo diet proponents suggest avoiding the products of agriculture, both modern and historical.

Colette Berbesque and co-authors [1] decided to test the frequent famine assumption by looking at anthropological evidence of famine in a fairly sophisticated way. In particular, they controlled for the effects of habitat richness by only comparing hunter gatherers from warm climates (Effective Temperature > 13C) with agriculturalists, and by using a metric of habitat productivity (a linear combination of Net Primary Productivity and Effective Temperature) as a covariate. From their abstract, they found:

... if we control for habitat quality, hunter–gatherers actually had significantly less—
not more—famine than other subsistence modes.

Digging into the details though, we find that this is true for some measures of famine but not others. For instance, on the variables Occurrence, Severity, Persistence, Recurrence, and Contingency of famine warm climate hunter gatherers do better (that is, less famine). In terms of short term or seasonal famine however, there is no difference between warm climate hunter gatherers and agricultural peoples. And that's important, because it is seasonal famine that would lead to the evolution of thrifty genes. And there was plenty of time for such genes to arise in paleolithic people before the advent of agriculture, so the fact that agricultural peoples are no better at avoiding seasonal famine simply means the selection for these genes wouldn't go away after agriculture.

Although Berbesque et al. paint their research as a critique of the "paleo diet" approach, it seems to me that it reinforces a key tenet: paleolithic peoples had it better than their agricultural neighbors. Hunter gatherers do significantly better on  "ordinary nutritional conditions and endemic starvation", have no difference in seasonal famine, and less long term and unpredictable famine. Huh.

[1] J. Colette Berbesque, Frank W. Marlowe, Peter Shaw and Peter Thompson. 2014. Hunter−gatherers have less famine than agriculturalists. Biology Letters 10, 20130853.

*The Pima aren't necessarily a good example, as they were agricultural before becoming "modern western". It is interesting to me that the National Institute of Diabetes and Digestive and Kidney disease article I linked to focuses on the increase in fat in the Pima diet, rather than the exchange of whole grains for refined carbohydrate.

Diabetes Management Experiment I

One of the key tenets of Adaptive Management is to "learn while doing", or using management actions to reduce uncertainties. In the management of Type 2 diabetes the real key is figuring out how your body responds to different meals. Meals are management actions, and the uncertainties are how one's body responds to each meal.

I just did one of these experiments. I've been trying to figure out why my A1c levels are representing an average blood glucose level higher than I measure before eating (pre-prandial). There are two possibilities. First, I know I have high blood glucose in the morning, which could be the "Dawn Phenomenon". The other possibility is that I experience high blood glucose levels after eating, a "post-prandial spike". The typical recommendation is to measure post-prandial blood glucose 2 hours after starting a meal. My post-prandial values then are typically pretty good, < 120 mg/dl.

So today I measured at 30, 60, and 90 minutes as well. The results are not good:

Minutes   BG
0              83
30            106
60            120
90            150
120          110

Check that out! A rise of more than 60 mg/dl! Now, if I'd eaten something loaded with carbohydrate... but no: stir fried vegies (cabbage, onions, carrots and green pepper. Not many carrots and green pepper either) along with 4 Lightlife Tofu dogs, and a couple tablespoons of Mayonnaise.

Now comes the difficulty with Adaptive Management. N=1. Even attempting to repeat the experiment I'll never replicate the conditions I had today exactly. However, the hypothesis that stirfried veg + tofu dogs isn't spiking my blood sugar just dropped in probability.

My low fat experience

From 1995 until around 2000 I followed a very low fat vegan diet. The target was to achieve < 10% calories from fat. Kris Gunnar over at Authority Nutrition has a typically well researched blog post summarizing the known effects of a low fat diet.

You should measure your fasting Blood Glucose.

And you should make a graph. Here's why.

The plot shows my fasting Blood Glucose values since I moved to Lincoln in early 2003. The dotted lines show the "normal range" reported on your lab results. So I was clearly outside the normal range most of the time since 2004. However, my physician didn't sound alarmed until mid 2009, the point marked with an asterisk. Why not? Well the value considered to be a sign of pre-diabetes is fasting glucose values above 120 mg/dL. We didn't do anything dramatic at that point; all that changed was that John* started ordering a new test, Hemoglobin A1c. My A1c value then was 5.8 % (more on what the means below).
In June of 2011 my A1c value was 6.2, and John decided it was time to declare me a Type II diabetic. I immediately changed my diet, and you can see the effects -- 3 months later my A1c value was back down to 5.0 %, which is the top of the normal range for that metric.
I was lucky; John is an enlightened physician. And he compared values backwards across my various labs, at least one step back. But I wish he'd started with the A1c alot sooner. A1c measures how much sugar is in your blood averaged over a few weeks. This is important because even if your fasting blood glucose levels are good, after a meal your blood sugar can spike way up if you are pre-diabetic. And those peaks damage things I value, like nerve cells! Even though my fasting BG value was lower in June 2011 than it had been, my A1c value was higher. After changing my diet my A1c values have been excellent (between 5 and 5.4), even though my fasting glucose values tend to still be high. So I could have been in trouble on the A1c scale much sooner, if I had been tested. I could swear that I've seen a graph of mortality as a function of A1c that suggested values above 5 were really, really bad, but I've been unable to find it. There are lots of studies that show increasing A1c increases all sorts of bad outcomes, but they mostly bin values together so you can't see the effect of a value of 5.5 vs. 5. Certainly, going above 6 % is bad. 

So, making a graph makes the data trend much clearer than simply seeing the lab results one value at a time. Your doctor won't do this. You can. 

I'm not going to describe the changes I made to my diet to get this result here; that's for another day. However, I want to connect this post to the Adaptive Management theme of the blog. In 1995 I made a major shift in my diet to very low fat Vegan following the recommendations of Dr. Dean Ornish, the McDougalls, and others. I had swallowed the dietary fat = death paradigm of the time. And it worked, at least for a while. I lost weight. My blood lipids improved. But 5 years later my blood lipids were worse than before, and 10 years after that I was diagnosed diabetic. I had to change the underlying hypotheses about the relationship between diet and health, at least for me.  

*Names may be changed to protect the innocent!