What's Your Standard Deviation

[Lloyd]

There is a large body of evidence that how much your glucose varies from reading to reading is important in preventing complications, independent of your average glucose.

Eating fairly low carbohydrate has lowered my SD significantly.

14 day SD 13 mg/dl


What's yours?

-Lloyd

[Injecto]

And how do you calculate this figure?

[Alice]

Lloyd, this is an interesting theory that I've heard nothing about. What study are you referring to? Also, how are you figuring this SD? Is it done automatically on your meter?

Could you give us more info on the research of this theory?

[statdeac]

Holy cow, Lloyd...that is LOW!! Congrats! (Mine varies.) ;)

[Lloyd]

Lloyd, this is an interesting theory that I've heard nothing about. What study are you referring to? Also, how are you figuring this SD? Is it done automatically on your meter?

Could you give us more info on the research of this theory?

I get this off my pump, or for time periods greater than 30 days off the cosmanager software that my pump uploads to. Some of the software on your computer that a pump would upload to will also give it. Also, some meters and/or the software you upload to will provide this data.

It's not a theory, it is the result of long term studies that have been done. If you think about it, it makes sense.
A 50 and a 250 average out to 150, which is not bad. But a 50 is a low, and a 250 is a high, neither of which is good.
So if your average is made up of highs and lows balancing each other out, that is not a good thing.

If you have "Pumping Insulin", by John Walsh, check out page 18 among other places.

"But glucose variability, measured by the standard deviation in some new meter or pumps or when data from a meter is downloaded to software in a PC, is drawing attention to an independent risk factor."

"The DCCT proved that exposure to high blood glucose was damaging. New emphasis suggests that that exposure to glucose variability is damaging."

If that is unconvincing, do your own research, I did.

-Lloyd

[shockme]

hey lloyd! i'm such a deviant that there is no standard for me......:D trish

[Lloyd]

hey lloyd! i'm such a deviant that there is no standard for me......:D trish

You never shock me Trish


-Lloyd

[Lloyd]

If you have "Pumping Insulin", by John Walsh, check out page 18 among other places.

-Lloyd

Sorry, typo. Page 19

[someone]

I really don't think using your meter for standard deviation or average for that matter cuts it. Both of these figures can be greatly influenced by the amount of time that is spent high or low. If you don't have enough data points, neither is going to give you much information. That said, the standard deviation I get on my CGMS is usually anywhere from 25-50.

Have a look at Diabetes Monitor - standard deviation buzz (http://www.diabetesmonitor.com/m57.htm). According to that article, if your BG average were 120 mg/dl, then a standard deviation of 40 mg/dl or less is acceptable.

Also, you have to realize that both the average and standard deviation are useful in determining control. Here is a good blog post I found. Average vs. Standard Deviation « Plastic Pancreas (http://plasticpancreas.wordpress.com/2007/08/06/average-vs-standard-deviation/)

[Lloyd]

I really don't think using your meter for standard deviation or average for that matter cuts it. Both of these figures can be greatly influenced by the amount of time that is spent high or low. If you don't have enough data points, neither is going to give you much information. That said, the standard deviation I get on my CGMS is usually anywhere from 25-50.

Have a look at Diabetes Monitor - standard deviation buzz (http://www.diabetesmonitor.com/m57.htm). According to that article, if your BG average were 120 mg/dl, then a standard deviation of 40 mg/dl or less is acceptable.

Also, you have to realize that both the average and standard deviation are useful in determining control. Here is a good blog post I found. Average vs. Standard Deviation « Plastic Pancreas (http://plasticpancreas.wordpress.com/2007/08/06/average-vs-standard-deviation/)

For sure, average is important, I never said it wasn't. But so is standard deviation.

I test 8 times a day, and am back to pre-meal levels 70 minutes after eating, usually. That gives me over 700 data points between A1c readings. An A1c will be affected by the speed of your drop to normal glucose levels. The SD will be affected by the height of your peak glucose, and the depth of your lows. Both the height of your peaks and the speed you return to normal are important.

Dropping my daily carb intake from about 125 to 80, dropped my standard deviation from 24 to 13. You can see it in the numbers, without quantifying them with standard deviation, nonetheless, SD helps tell you how much you have improved the spread of your numbers.

If you test once a day, SD will not tell you anything, that is part of why I posted here in the pumping section, pumpers test a lot more than that as a rule.

What is an acceptable SD is not the same as what is a good SD or a great SD, the same as a 7.0 A1c may be acceptable, but it is hard to argue it is as good as a 5 something.

It is all tradeoffs in lifestyle and risks of complications you are willing to take.

Just be aware that a good A1c is not sufficient to reduce the risks, if it is made up of high peaks and low valleys.

-Lloyd

[dgrilli]

I get it Lloyd. I am on board with you on this.

If one could keep his deviation as close to the same as possible this would be the optimum period.

The Swings in great deviation would be in theory worse than a constant high or low.

Claiming more damge done to the body through wild deviation is not hard to grasp.

I believe Doc Bernstien practices this and we all want to who pump or are on MDI by timing our bolus's and or basals to stop the Madness (Deviation through Swings in BG).

To have the optimum lifespan sans complications.

Did I understand?

[someone]

For sure, average is important, I never said it wasn't. But so is standard deviation.

I was agreeing with you, but also bringing up some additional valid points.

[HelenM]

And how do you calculate this figure?

If you really want to calculate it yourself see http://http://en.wikipedia.org/wiki/Standard_deviation ( http://en.wikipedia.org/wiki/Standard_deviation)

but its less hard work using a calculator in scientific mode.
You can do it with the microsoft calculator.

Click the View menu, and then click Scientific.

Type or click your first piece of data, and then click Sta to open the Statistics Box dialog box.

Click RET to return to Calculator, and then click Dat to save the value.

Type or click the rest of the data, clicking Dat after each entry.

Click s.

[Injecto]

Ugh!....

Let's say I have alll my BG readings posted in a spreadsheet....then what?

[statdeac]

I really don't think using your meter for standard deviation or average for that matter cuts it. Both of these figures can be greatly influenced by the amount of time that is spent high or low. If you don't have enough data points, neither is going to give you much information...True, depending on when and how often a person tests. I think that someone without CGMS could get a reasonable view with a decenly representative sample of tests throughout the day. That said, nothing beats CGMS for this.

[LoDeSp]

Based upon about 1.5 months of testing (87 points), my stddev is 28. So far this month (20 points), it's 24. Means of 119 and 108, respectively. Not Lloyd's levels, but I'm happy with them.

Injecto: if you have all of your numbers in a spreadsheet, enter in a new cell: "=stdev(range)", where "range" includes the values that you want to consider (e.g. B165:H213).

[Injecto]

Thanks LodeSp.

I suppose I'm not doing to hot then. My 14 day is 2.6 in mmol. What doesn't make sense to me is that in mg/dl it's 57, which would translate to a 3.1 mmol. Something doesn't seem right.

(to be accurage, my mmol is 2.617276525 and mg/dl is 57.18233)

Either way it's not good.:(

[shabbie6247]

okay so if ive done this right, yesterdays figures

5.8, 7.1, 6.7, 4.4, 4.9, 6.6, 4.8, 8.6, 9.6
totals = 58.5 divided by 9 readings = 6.5

so my standard deviation is 6.5 mmols? (117)

[shiftzor]

hey, standard deviation is 2.4mmol/L and average is 6.4mmol/L. No idea if thats good or not :)

Thats how to do standard deviation: Standard Deviation (http://www.gcseguide.co.uk/standard_deviation.htm) oh and mine is 171 tests over 30days.

[someone]

okay so if ive done this right, yesterdays figures

5.8, 7.1, 6.7, 4.4, 4.9, 6.6, 4.8, 8.6, 9.6
totals = 58.5 divided by 9 readings = 6.5

so my standard deviation is 6.5 mmols? (117)

That would be your average. Standard deviation is a little more complicated. ;) You can use this tool to calculate your std dev manually (as in you have to enter each glucose value). Standard Deviation Calculator (http://invsee.asu.edu/srinivas/stdev.html)

Carelink also calculates std dev and you don't need a minimed pump or CGMS to take advantage of this feature. All you need is a supported glucose meter. http://carelink.minimed.com

[LoDeSp]

Thanks LodeSp.
I suppose I'm not doing to hot then. My 14 day is 2.6 in mmol. What doesn't make sense to me is that in mg/dl it's 57, which would translate to a 3.1 mmol. Something doesn't seem right.


I suppose it could be rounding errors in the mmol figures, or a bad conversion. Standard deviation is scale-invariant, so it should have been ~47 in mg/dl.

[Lloyd]

All any of us can do, is our best.

I know being a T2, lower SDs are likely easier to obtain than for a T1.

But your body doesn't care which type of diabetes you have, or what method you use to treat it. All that matters, as far as complication risks go, is the results, what your glucose situation is. We all have the same ideal, get your glucose readings in the same range as a non-diabetic.

For most of us, at least, it can be done! It's a lot of work, and a pain in the !@#@#$@#$.

-Lloyd

[RobiJo]

This is something I've only recently heard about. It's interesting. My current take on my control is to limit the amount of time I spend out of range. Using CGMS, I can report the number of hours:minutes I spend below range, in range, and above range. I've been slowly lowering what sets off the "high" alarm on the pump/CGMS--started at 190 back in early August. Now I'm at 150. I also have the low set at 75. Realistically the more you stay in a tight range the lower your SD will be, so perhaps I'm hitting two birds with one stone by trying to stay in range for more time each day.

[Lloyd]

I get it Lloyd. I am on board with you on this.

If one could keep his deviation as close to the same as possible this would be the optimum period.

The Swings in great deviation would be in theory worse than a constant high or low.

Claiming more damge done to the body through wild deviation is not hard to grasp.

I believe Doc Bernstien practices this and we all want to who pump or are on MDI by timing our bolus's and or basals to stop the Madness (Deviation through Swings in BG).

To have the optimum lifespan sans complications.

Did I understand?

Yes!

A 6.0 A1c witout highs and lows is better than a 5.0 with wild swings and dangerous lows.

-Lloyd

[shiftzor]

Interesting or not, one of the articles i just read states that your standard deviation should be 33% (1/3) of your mean average. my mean average is 6.4mmol/L wich means i should have a standard deviation of 2.112mmol/L, my current standard deviation is 2.4mmol/L. Looks like i might have to do a little more work then, i have only just started to carb count so hopfuly it will normalise out eventually. I now know why i did all that maths :eek:

Article (dont know how trust worthy it is):
Diabetes Monitor - standard deviation buzz (http://www.diabetesmonitor.com/m57.htm)

[Lloyd]

This is something I've only recently heard about. It's interesting. My current take on my control is to limit the amount of time I spend out of range. Using CGMS, I can report the number of hours:minutes I spend below range, in range, and above range. I've been slowly lowering what sets off the "high" alarm on the pump/CGMS--started at 190 back in early August. Now I'm at 150. I also have the low set at 75. Realistically the more you stay in a tight range the lower your SD will be, so perhaps I'm hitting two birds with one stone by trying to stay in range for more time each day.
Staying within range, then tightening the range when you can do that almost all the time, will result in better and better Standard Deviations.

My personal range is now 80 to 115. Beyond that I figure I screwed up somewhere, and try to learn from it so I don't do it again. Higher than 115 usually means about a 120 to 125, which is what I get when I forget to take my metformin. This happens about twice a month, despite alarms to remind me.

For me, lowering my carb intake helped my control a lot, and has made weight loss possible.

-Lloyd

[someone]

Yes!

A 6.0 A1c witout highs and lows is better than a 5.0 with wild swings and dangerous lows.

-Lloyd

Lloyd, keep in mind that a standard deviation of 40 is considered normal for a non diabetic. Non diabetics can have their blood glucose anywhere in the range of 70-160 and even higher in some cases. If you take the low (70) and the high (160) then average them, that is a 115 average. In this case the standard deviation between the 2 data points is 45 mg/dl. Of course, this is an extreme because I am not taking any time into account, but it proves the point.

[Lloyd]

Lloyd, keep in mind that a standard deviation of 40 is considered normal for a non diabetic. Non diabetics can have their blood glucose anywhere in the range of 70-160 and even higher in some cases. If you take the low (70) and the high (160) then average them, that is a 115 average. In this case the standard deviation between the 2 data points is 45 mg/dl. Of course, this is an extreme because I am not taking any time into account, but it proves the point.

I don't agree.

From:

Diagnosing Diabetes: glucose tolerance test and blood glucose levels. (http://www.endocrineweb.com/diabetes/diagnosis.html)


The "gold standard" for diagnosing diabetes is an elevated blood sugar level after an overnight fast (not eating anything after midnight). A value above 140 mg/dl on at least two occasions typically means a person has diabetes. Normal people have fasting sugar levels that generally run between 70-110 mg/dl.


-Lloyd

[w5wjp]

Ummmm......bet that is what my GF is talking about when she tells me I am a deviant.

[someone]

I don't agree.

From:

Diagnosing Diabetes: glucose tolerance test and blood glucose levels. (http://www.endocrineweb.com/diabetes/diagnosis.html)


The "gold standard" for diagnosing diabetes is an elevated blood sugar level after an overnight fast (not eating anything after midnight). A value above 140 mg/dl on at least two occasions typically means a person has diabetes. Normal people have fasting sugar levels that generally run between 70-110 mg/dl.


-Lloyd

Have a look at this thread: http://www.diabetesforums.com/forum/diabetes/17046-normal-blood-glucose-graph.html

You're right, above 140 is not normal for a non diabetic but it can happen. I was giving the worst possible scenario for a non diabetic.

[LoDeSp]

Interesting or not, one of the articles i just read states that your standard deviation should be 33% (1/3) of your mean average.


I'd be interested to know what sampling frequency that standard deviation recommendation corresponds to. It seems to me that it would be much easier to get a low standard deviation with a CGMS-type sampling (e.g. testing every hour) than the before-and-after-meals approach.

I guess the bottom line is that standard deviation should be as low as possible and that standard deviations with different sampling frequencies aren't necessarily comparable.

[shabbie6247]

Based upon about 1.5 months of testing (87 points), my stddev is 28. So far this month (20 points), it's 24. Means of 119 and 108, respectively. Not Lloyd's levels, but I'm happy with them.

Injecto: if you have all of your numbers in a spreadsheet, enter in a new cell: "=stdev(range)", where "range" includes the values that you want to consider (e.g. B165:H213).

AHA! now thats more like it, thanks ;)

mine calculates as 1.9/34 (over 10 days testing)

[someone]

Here are my stats from the past 2 weeks. Standard deviation from the CGMS is consistently lower than that from my meter, but I guess it all depends when you test. Because I have a CGMS I tend to test when I am high or low.

[Lloyd]

I don't think results from CGMS are reliable enough to use as a reference, at this time.

I am basing this on published charts from funnygrl and others, sometimes fairly accurate results can be obtained, and sometimes results are way way off from fingerstick results. Very underwhelming.

I think CGMS will be a good source of info, but IMHO it is not yet ready for prime time.

I really look forward to seeing how the navigator does, if the FDA investigator ever gets off his duff.

-Lloyd

[Funnygrl]

Ugh!....

Let's say I have alll my BG readings posted in a spreadsheet....then what?
Make spreadsheet do math. =stdev

[HelenM]

I use SI diary to record levels. Its a useful tool especially if you fill in different activities, meals out etc.
It calculates SD over any period you want and for different testing times.

My SD for fasting levels over the last 3 months was15.7 but one weekly interval as low as 2.2 and another its 30. When I look at that week, its the week I was backpacking so its clear why my BS were less stable that week. Hopefully next time I do a similar walk the records will help me keep more consistent BS.

I'm not sure that comparing different peoples readings is very useful since it depends on testing intervals. I don't always test at 2 hours post prandial, finding 3 hours more useful but probably means I'm missing some of the higher figures whereas someone who tests at 1 hour might show a much higher variability.

All readings going back in 4 week intervals
av. 91 sd 20.2 147 tests (stress less exercise, less tests)
av 81 sd 18.5 163 tests
av 81 sd 20.8 165 tests
av 83 sd 19.8 174 tests



The software also 'calculates' HBA1c and has so far got it right every time.
(Sorry didn't mean to sound like an advert.)

[someone]

=I'm not sure that comparing different peoples readings is very useful since it depends on testing intervals. I don't always test at 2 hours post prandial, finding 3 hours more useful but probably means I'm missing some of the higher figures whereas someone who tests at 1 hour might show a much higher variability.

Right! If you are using fingersticks to measure std dev, your result is going to be dependent upon when you test. All you have to do to completely throw it off is test twice in a row.

[frige]

14 day std dev = 64 mg/dl
5.4 tests perday

I think I gotta superbolus.