At last week’s Diabetes Technology Society Meeting in Bethesda, MD, Dexcom’s VP of Science & Technology Tom Peyser presented a brand new metric for glycemic variability based on CGM data that may be much more useful and easier to understand than Standard Deviation.

“Glycemic variability” is of course a fancy name for how out-of-control your blood sugar levels are. Swinging between severe ups and downs all the time (high variability) puts you at risk for severe lows, may contribute to complications long-term, and sure as heck has a negative impact on your mood and quality of life.

Up to now, the methods used by clinicians for measuring glycemic variability were deep-science complicated, with acronyms like MAGE, CONGA, MODD and SD (Standard Deviation).

“All are difficult to calculate. None are easy to understand. And all fail to differentiate between basic cases. A simple, easily understood, easily calculated glycemic variability metric is needed,” Peyser stated in his DTS presentation.

His new concept is simple: look at the CGM data graph below. Instead of concentrating on how high or low the line is fluctuating, think about the LENGTH of that line if it were a piece of string and you stretched it out; the more ups and downs the line shows, the longer it would stretch out, of course, because the line would actually be *longer*.

Based on this “length of line”(or distance traveled) concept, Peyser says Dexcom has developed a new, “intuitive topological measure” they’re calling the Glycemic Variability Index, or GVI. It can be easily calculated in Excel, using this “easy” trigonometry:

Then, calculated for a given length of time, the values are easily understood:

GVI = 1.0 to 1.2 means low variability (non-diabetic)

GVI = 1.2-1.5 means modest variability

GVI = >1.5 means high glycemic variability

Peyser says he’s spent the past months working on this concept and running it by glycemic variability experts like Dr. Irl Hirsch, “and they all felt it had merit,” he says.

To validate the concept, they’ve run comparisons to both the Standard Deviation and the commonly used MAGE measure, which he says is particularly complex to understand and calculate.

“Our results were comparable to both MAGE and Standard Deviation!” Peyser says.

But they didn’t stop there.

They wanted to develop something whereby a patient’s “full glycemic state can be characterized by a single number.”

What they came up with is something called the Patient Glycemic Status (PGS) measurement. It combines calculations of your GVI + mean glucose + percentage of time in range, using this formula:

They even did a study of diabetic and non-diabetic subjects to validate these values:

PGS ≤35 excellent glycemic status (non-diabetic)

PGS 35-100 good glycemic status (diabetic)

PGS 100-150 poor glycemic status (diabetic)

PGS >150 very poor glycemic status (diabetic)

“Both GVI and PGS can be used to ‘flag’ problems in glycemic control and help clinicians direct resources to patients who need further help,” Peyser asserted.

Don’t forget that Peyser works for Dexcom, a CGM company. And the successful use of CGM is defined as reduced glycemic variability. They’re all about ways to make better use of CGM data — and I think we can all get behind that as well!

I have no doubt we’ll be hearing more about the new GVI and PGS markers in the coming year.

In the meantime, next time you’re viewing your own CGM data, go get some string and scissors to measure your “line.” Then, if you’re not so hot at trigonometry, at least you can compare today’s string to the one you’ll cut out next time. Here’s to achieving a shorter line!

Now that’s out of the box thinking and two very useful metrics. For Dexcom, then can probably already calculate GVI and PGS with ease — I hope we see this come out soon.

One question I do have. If one end of the line is elevated (say from 120 to 240), then L0 would we bigger than L0 if the line was flat. Do they allow for that? Even if they don’t it probably doesn’t make a huge difference.

Amy – Thank you for writing on this important topic. As a long term (since 2009) CGM user, I’ve been using standard deviation (SD) and coefficient of variation (CV) to analyze the quality of my control. When my line is flatter I do feel better.

While I know just enough about math to be dangerous, I’d be grateful if you did a follow-up column to this one. I’m most interested in figuring out the actual steps to implement Excel to crunch the numbers needed to calculate a GVI and PGS. Discussion of a step by step example would be instructive.

Thank you for keeping on top of emerging diabetes topics like this one.

@Terry, the distance between two lines is Sqrt( (x1-x2)**2 + (y1-y1)**2).

In Excel the formula is =SQRT(POWER(A1-A2,2)+POWER(B1-B2,2)). If you assume the X value (time) is in minutes, and the Y value is readings, it’s not too hard to do.

Send me your email (bernard (dot) farrell (at) gmail.com) and I’ll email you a sample spreadsheet.

Possible correction: PTIR looks like should be the fraction of time in the target range [i.e. it can never be greater than 1] not the percentage of time in the target range. This implies that the PGS goes to zero when all values are within the target range.

One suspects with all the data crunching of CGM data over the last decade, this concept was already qualitatively understood by researchers. However, it would not be useful to introduce it to physicians and patients until the CGM manufacturers believed data was always of very high quality. Apparently we have now reached that point!!

@mcityrk is on to something, PTIR can never be 1. How does this work for a non-diabetic where PTIR could be equal to 1. Amy, can you point us to the original paper?

PTIR = 1 means 100% values in range = non diabetic

PTIR can never be > 1, ok, and for that PGS is always >= 0

Percentage is intended as 20% = 0,20

Just one question (yeah, right ….):

When researchers do “… a study of diabetic and non-diabetic subjects to validate these values … “. Does that mean they have achieved “valu-dation”?

Variability of blood sugar does correlate with diabetic complications as well as immediate problems from hypoglycemia. What may sound like theoretical math actually has a great impact on people’s lives. This was discussed a little bit in the meetting in Boston: sounds like they went into much greater detail at your meeting. Thanks for a great post.

I don’t understand the math (I didn’t pass algebra) but as a cgs pump User for 2 years – medtronic-I understand the WIDE VARIABILITY as a extremely brittle diabetic. Can you put this into EASY understandable language for an excel spreadsheet? Thanks!!

@c.a. costova.

Send me an email and I’ll forward you a working spreadsheet that does this calculation for you. bernard(dot)farrell(at)gmail.com.

Anyone knows how much is the constant L0 ? It’s needed to do the calcs….