There seems to be a lot of buzz about the fact that it’s not just about A1C’s anymore; rather, it’s about using the Standard Deviation to evaluate the success of your diabetes management. If you think about it, it’s common sense really: a simple average of your blood glucose over the past three months doesn’t tell you how far you’ve strayed from the ideal range. You may get an excellent “average” number that is nothing but a middle point between the highs and lows you’ve been experiencing.

At an informational meeting a few days ago, I learned where all the buzz started. The idea of concentrating on Standard Deviation is not entirely new, but has found an avid proponent in Dr. Irl Hirsch, a Type 1 diabetic himself, who is medical director of the University of Washington Diabetes Care Center.

He apparently presented his initial paper on the Power of the SD at the Clinical Diabetes Technology Meeting last April, and his definitive article on the subject is slated to appear in next month’s Diabetes Technology and Therapeutics Journal. This publication is interesting in itself, as it is driven by the Diabetes Technology Society, a group which I’ve discovered is headquartered here in the California Bay Area very near me! This group organizes two annual conferences, the clinically focused April meeting mentioned, and also the Annual Diabetes Technology Meeting in November. Both are intimate events (less than 100 attendees) for the movers and shakers in D-technology.

**SD formula for diabetes: ****A1c x SD/100**. Find that confusing? Try the online Standard Deviation Calculator here. Also, if you can manage to export your downloaded BG montior data into Excel, that program will calculate the SD for you.

The theory is that the greater the deviation in your blood sugars, the more likely you are to experience mircovasular damage in eyes, kidneys, etc. Dr. Hirsch suggests that diabetics should aim for an SD of one-third of their mean blood sugar. So, if your mean blood sugar were 120 mg/dl, you would want your standard deviation to be no more than 40 mg/dl, or one-third of the mean.

Anyone out there using the SD formula successfully yet? Has it helped you improve your control just *knowing* how much you are “all over the map”? Do share!

I’m ashamed to admit that I don’t download my glucose meter results — maybe I’m too lazy, but heck, the rest of my life is already tracked by a half-dozen other spreadsheets (kids’ appointments, nanny schedule, work deadlines, dinner menus, and so on). So just haven’t had the impetus to fuss with D-Math. (You might say I prefer telling D-stories

I haven’t done anything with i>formulas, because trying to get a single value that expresses BG control is useless. “control” of diabetes is gauged by the statistics of blood glucose readings; and any statistician will tell you that you can’t make a statistical judgement with only one value. Most statistical analyses result in two numbers: An average, and an average deviation from that average. These are usually called different names, and each of those names uses a slightly different method to calculate, but they all wind up being pretty much the same thing.

But still–nobody’s going to tell you that one number can get everything done. Basically, you need the average, and you need a measure of how confident you are in that average.

Really, of the two, the standard deviation is actually more important, because it has more to do with what people think of when they talk about “control”. The average is important (and that’s what A1c tells you) but the “swings” are just as important–because damage accumulates faster when BG is high, and there are obvious problems when BG is low. If your BG scores have a high standard deviation, then you are more likely to hit both high and low numbers–even if your average is nice and steady.

The math here is not hard; you could do it yourself with a calculator, if you had the time to tote up all the numbers. But with a computer it’s a snap; the only annoying part is remembering to get all the numbers into your spreadsheet.

Any diabetic who wants to control their diabetes, instead of just being dependent on the medical industry, needs to do the math. You need two points to draw a line; you need two numbers to control your body.

Good luck educating people on SDs. Our lives are ruled by SDs. The normal range for all tests is defined as plus or minus 2 SDs, which generally equates to 95% of the population. By definition 2.5% of normal people are too low and 2.5% are too high!!

This guy is trying to push for tighter and tighter control; the real question is how tight is tight enough? Too tight means too many lows.

This is from a guy that has had least one low a day forever. The A1C shows all aspects of blood glucose it is biased chemical average versus a machine average; it is still the best we have.

People with poor control are the ones who say “too tight means too low”. The big secret of diabetic control is that your average BG doesn’t have much to do with how well controlled you are…you can get your average to be wherever you want it by adjusting your treatment. The real trick is keeping away the lows and the highs…and that’s what people mean (or should mean) when they talk about control.

I posted this elsewhere on the site, but someone with an average of 130 and a standard deviation of 20 is far better controlled than someone with an average of 110 and a standard deviation of 60.

Fun with Math! Hmmmm……this is interesting. I started putting my blood sugar readings in a spreadsheet in December. I calculated daily average, running 7-day average, and average error (absolute value of difference between reading and 90 mg/dL).

So far for the month, my average is 103 mg/dL. I just calculated the SD–33.

So that’s just about right–a third of average.

The average error is 9 mg/dL. Thanks for sharing the idea. I will use this new metric to assess my control.

Nick

A clinical study was performed in 2002 which enlisted wireless devices for automating the collection of glucose meter data and automatically charted each patient’s data in the study population.

One of the web-based reports enlisted the mean blood sugar levels over 14 days plotted against the standard deviation and showed all of the patients on a single graph. You can imagine how insightful this report was especially as the relative patient control changed over the course of the 4 month study.

We learned some interesting things through this research including the ability to efficiently identify those patients in most need of intervention. A1c clearly falls short when it comes to understanding which patients to focus on in a proactive healthcare system. Another number which we identified as a valuable 3-Dimensional statistic in addition to A1c and SD is the percentage of hypoglycemic bg tests. With those 3 markers we should be able to create a significant improvement in the “gold standard” for diabetes.

The study was conducted by Stephen Ponder, MD CDE and presented at the Diabetes Technology Society Meeting in November 2003.

Hmm….Standard deviation (mg/dL) vs. one-third of average blood sugar (mg/dL)?

I started logging all my blood sugars in December 2004. I started Dr. Bernstein’s diet/exercise program on January 12, 2005. In December, my actual SD exceeded avg/3 by 25 mg/dL. Since then, it has consistently fallen every month, to +7 mg/dL in June. NOW, in July,for the first time since starting the Bernstein regimen, my SD has dropped below avg/3 by 2 mg/dL. WOW!!! I knew my glycemic control had dramatically improved, and that for the first time in thirteen years I felt in complete control of my type 1 diabetes. But now, for the first time, you’ve given me an objective way of quantifying my control. THANKS!!!

This is important. Really important. It is the first positive thing I’ve had in the last couple of months. And they’ve been a really, really rough couple of weeks. Last week I actually lost the will to live. I am still very, very depressed. I am just taking it half a day at a time right now…

re: your article “Standard Deviation Buzz.” You asked “Anyone out there using the SD formula successfully yet?” Well, I am; however, I’m Type-II not Type-I.

Initially, my A1c was 7.2 in Aug’03 when I was first diagnosed as diabetic type-II. It’s now been down at 6.2 for last 3 consecutive 6-month interval tests.

This year my accumulative weekly average (mean) blood glucose values are:

BFAST = 109 mg/dl, SD = 14

LUNCH = 110 mg/dl, SD = 14

DINER = 117 mg/dl, SD = 20

And, thus, my MEAN/SD ratios are:

BFAST = 109/14 = 7.8

LUNCH = 110/14 = 8.0

DINER = 117/20 = 5.7

To “keep” my diabetes “under control,” my physician tells me to “keep” my MEAN/SD-ratio at 4:1, or higher (preferrable). So, my “goals” are MEAN = 100 and SD = 5.

So, “Yes,” I am using SD!

I happened on your Blog tonight and read with great interest this SD formula story and Dr.Hirsch’s involvement in it. Dr. Hirsch was my doctor for eight years and he he checked my SD and A1c. Though he didn’t have the formula worked out yet he used these numbers to provide me with the best care.

I’d like to say that we diabetics are lucky to have Dr. Hirsch working for us. He is a diabetic himself and the most compassionate and caring doctor I have ever had in my fourty-one years as a diabetic.

I have since moved to Sacramento and have yet to find medical care anywhere close to what Dr Hirsch provided me.

I agree with this SD priciple. My A1c has been about the same before and after starting pump therapy. Before the pump I had lows everyday and up and down like a yoyo. After the pump I rarely have lows.My A1c says I havent improved but in my heart I know I have.

I want you send me every new information about updated monitoring of diabets

You might want to head over to the nice chaps as http://sweetspot.dm and upload your meter data there.

It’ll give you the std dev, average, hirsch ratio and a bunch of graphs. Quite nice really.

I have been tracking my average and mean blood sugar readings as well as the standard deviations since August, 2004. I’ve been using insulin since 1958. However, I’ve not found any one with any interest it it. That first month, my average was 141.5, mean 128 and SD 76.1. I knew that was terrible but just tried to lower it by myself. Last month, February 2008, my average was 93.1, mean 86 and SD 41.5. I also track the percentage of readings that fall between 79 and 121. The first month was 20%, last month 33% That’s progress but I guess I’ve got a ways to go!