I'm not a statistician (though my late husband was), but I'm finding myself trying to pick holes in your solid explanations. I'm one of those who believes that A1C is fairly useless without looking at SD alongside it. I *like* having my SD in the range of 20 - for me that's always shown that in addition to having had a decent average BG (if there are no major outliers), I've managed to keep the boat fairly steady. Maybe that's redundant, and it's enough to look for big swings without figuring a SD?
the idea that there, really, isn't a clear correlation between A1c and BG for any single diabetic
I think you're wrong going that far.
What matters is how much and how long our body stays in high BGs, because it ruins our body.
A1c is formed by a chemical reaction so it's a usefull measure of our BG history.
We are not all equals, but data has been collected to state that A1c correlate well with BG average.
For someone it measures a shorter period than others (4 weeks ? 8 weeks) but it's a usefull benchmark about our diabetes control.
Your discussion about SD regards another measure we need, a short number to express how much it costs us to accomplish our A1c.
SD is not the only one, and not the wider used.
I think the other question is "are there people who might benefit from using their SD in conjunction w/ A1C?" like running a 7.0 A1C w/ a, I dunno, 50 SD might be more hair raising and suggest a different approach than a 7.0 with a 20?
One could, after all, "cheat" an A1C by washing out highs w/ a proportionate number of lows, to get an ok "average" while perhaps having a more hair-raising time dealing with going hi-low-hi-low.
Yes, that's why we need A1C/CGM average plus another "indicator". Did you ever read about HBGI and LBGI (High Blood Glucose Index and Low Blood Glucose Index)?
They express your "risk" of ipo or hyper, they would show up if your A1C is "good" or hi-low-hi-low
You make perfect sense. Honestly, I'm not sure I would say "redundant" because that would imply that I thought SD as actually a good complement to knowing what your own big swings are from the raw data.
Like I said, if you find it useful as a complement to your A1c, I think that's great. You get the necessary information from it that helps you control your BG's, then my argument doesn't really apply to you and what you are doing personally.
Out of curiosity, you said "if there are no major outliers". What do you find outliers do to your SD? Does it screw up your relationship between SD of 20 = decent BG average?
Just to clarify, don't get me wrong. I do feel strongly that there is a between an individual's average BG and their A1c.
What I meant by my statement is that you can't simply go to the internet, look-up "Blood Glucose/A1c comparison Chart" and expect the relationship on the oft referenced chart to represent your Average BG/A1c relationship.
Here are the graphs showing the actual relationship between A1c and Avg BG I was talking about:
That's a lot of scatter around the regression line. Too much to think that you're relationship will be the one represented by one chart, I think the only way to determing your own relationship is to test, a lot, and have a robust average BG to compare to the A1c you get back from the lab,
Exactly why I assert that A1C isn't that meaningful without looking at SD beside it.
I just want to be crystal clear about this point jrtpup.
I think there is merit to looking at your BG variability along with your A1c. What I don't believe is that Standard Deviation is a robust, or even good, way to measure the variability in Blood Glucose data.
It seems like a lot of people automatically equate "Standard Deviation" to "Variability" when it's only warranted under specific assumptions and conditions. Remove those assumptions and Sd, literally, becomes meaningless or does not have it's intended meaning.
I think the other question is "are there people who might benefit from using their SD in conjunction w/ A1C?" like running a 7.0 A1C w/ a, I dunno, 50 SD might be more hair raising and suggest a different approach than a 7.0 with a 20?
Hopefully, I haven't argued that there aren't people who would benefit from the use of SD in this manner. If you understand the limitations, understand the inherent biases, undertand that an incremental increase or decrease in your SD doesn't necessarily equate to a change in the variability of your BG measurements, then why not?
One could, after all, "cheat" an A1C by washing out highs w/ a proportionate number of lows, to get an ok "average" while perhaps having a more hair-raising time dealing with going hi-low-hi-low.
I suppose that's true, and I've heard of people doing it, but i wonder how you could do that and expect to get an A1c in the "normal" range, lets say sub 5.8 for arguments sake.
Again, probably the subject of an entirely different blog but let's suppose this:
I'm shooting for an average BG of 90 and I find myself with a BG of 180. How do I "wash that out"? I would need a proportionate decrease in my BG, correct? So, ostensibly, the quickest way would be to shoot for a 0?
Or would you just shoot for a a whole slew of readings in the low normal range to offset the 180? Isn't that what we do for good BG control anyway? Shoot for a whole slew of readings in the low normal range?
Anyway, how about a new blog post on how you can "game" your SD reading?
Here's a preview...
1) Don't use all of your readings. 2) Use only some of your readings. 3) Don't log highs 4) Don't log lows.
Okay, maybe we don't need a whole blog post. That is a preview of what I'd start doing if my doctor started sweating me about my SD, other than get a new doc of course.
At the very least, the number that comes back from the lab for your A1c is not dependent on your use of any of your own equipment.
Interestingly, and something I didn't want to "appeal to authority" with is the mention from Mendosa's blog that Irl Hirsch, who I believe is one of the earliest, if not the most respected, advocate of using BG variability as a measure of BG control, doesn't even believe that SD is a good measure of variability for BG.
I've seen that important point get lost when I hear discussion about how Dr Hirsch talks about using SD in this manner.
I don't think I game any of my readings, as I only use the ones out of the computer? I don't think that anyone "sweats" about these numbers as I don't think I've ever seen anyone discuss SD which was > 20 or 30, which is probably "OK"? Or is *everyone's* SD 20-30 b/c of mathematical foibles and that's why the term is meaningless? Re "washing out", I think that most people who "encounter" a high bg, will look to get rid of it, meaning that their BG will drop suddenly.There may be other people who get an "OK" A1C by perhaps "hearing" that "150" is ok and sticking to it? Just about everybody in this thread is in the 5s so most of us aren't the outliers. It would be interesting to see if people with the 7s might have different approaches reflected in their numbers? A few years ago, when I was on R/NPH, if my numbers were off, I'd take IV correction shots which would get rid of the highs quickly but also led to ok A1Cs, I had some high 5s and low 6s. I think the first time I went to a doctor in probably 5-6 years I got a 5.8 which suprised him (new GP...) a lot but we discussed things and I think that he figured I was running up and down a lot? My numbers may also be anomalies because I've mostly tested more than a dozen times/ day for a variety of reasons. My doc rx'ed 17 and insurance rx'ed 14, I don't look at my meter reports so I am not sure how often I test but I put it all in there? The last time I googled National Average A1C, I think it was 9 something so it might be that SD would be meaningless for people who have figured it out enough to run in the 5s, even though that's > "straight" people have? The test would have to be in people who have the same A1C but different "reported" SD? I guess that you can make the mathematical arguments that SD is somewhat meaningless but I would think that if you had two people with the same A1C and signficantly (eeek, another math word...) different SD, one might find a difference in their diabetes experiences that could be interesting. Of course, neither "experiences" or "interesting" are math words. Maybe propensity for complications, or catastrophic hypos could be related somehow?
Sorry acid, I use the figurative "you". Not suggesting that you actually game any numbers.
I guess to belabor the point, beat the dead horse, put too fine a point on it...
I think my entire argument is mathematical in that, without meeting the underlying assumptions, the term "Standard Deviation" just becomes a mathematical calculation without the mathematical meaning ascribed to it. When you ask about "everyones" SD of 20-30, they become even more meaningless, if that is possible, without reference to what the average and data set that generate the SD of 20-30 are.
I hope your questions regarding when SD are more applicable to use are rhetorical because I don't see any point to using SD for anything other than your own personal edification, at all, unless you can figure out a situation where the required assumptions for your data set of BGs and the sample of BG readings can be met.
Sorry =/
Oh, and I think that, given what we know about the general attitude of people towards diabetes care is, we can say that anybody with a A1c in the 5s is probably an outlier and ought to be congratulated and celebrated for their efforts.
If you're, the literal and figurative you, are out there having success using your SD to keep your BGs in the 5s, or better, and under control, I think it's more than worthwhile to discuss "how" you are doing it. What does your raw data look like? How does your knowledge of your SD specifically drive you to change numbers that you find in your dataset?
Thanks again acidrock, I really do appreciate your interest and contributions to this post.
I just think our sample (people who pay attention to SD) is both very small and also skewed towards people who are more engaged with diabetes. I think there are likely to be some people who are not as engaged for whom the number, which I expect would not be in the 20s, would have some type of meaning however the problems are likely adequately “alarmed” by their A1C results, measuring avg bg over time. Similarly, the people who might benefit from making changes are not likely to be told to do what they’d need to do to kick more ■■■ because endocrinology, fascinating though it may be to those of us embroiled in its sturm und drang, isn’t about kicking ■■■?
Re what I do isn’t at all “strategic” A1C/ SD, it’s to “win” every time I test my BG which, in turn, gets the A1C/SD to be ok.
In reply to your (FHS) question Out of curiosity, you said "if there are no major outliers". What do you find outliers do to your SD? Does it screw up your relationship between SD of 20 = decent BG average? I suppose my post wasn't clear, I didn't mean to imply that SD has anything to do with decent BG average, they are 2 different animals. Outliers will make for a higher SD, but have no relation to my average BG over time. For me, at least, they need to be looked at together. An SD of 20 with an average of 180 would send me into a tailspin LOL. I usually hover around 100 or a bit lower as an average, though on my recent cruise my avg BG was 118 with an SD of 34. That's what happens when I SWAG everything for 9 days LOL
Boy, I wish I could comment right under what I'm commenting on!
FHS you said I think there is merit to looking at your BG variability along with your A1c. What I don't believe is that Standard Deviation is a robust, or even good, way to measure the variability in Blood Glucose data. etc...
I just re-read your original post, in particular the last section. I have readings every 5 minutes from my CGM. I know about how accurate they are because I test pretty frequently, and take that into consideration when looking at the data. I'm missing something in my mathematically disinclined brain - doesn't my SD give me a pretty good idea of variation around my average BG?
Yes, another post LOL ar, I think those of us who are OCDiabetic (those engaged here in particular) is a small/skewed group.
FHS, my A1C is in the 5s. I use my cgm/pump logs to see where I have bigger than acceptable (to me) spikes/dips, and use the data to tweak my regular everyday eating patterns; maybe lowering carbs somewhere, spreading them out more, yada yada. I've always seen a lower SD as an indicator of fewer spikes/drops, or less variability. When my SD is higher, I do see more variability - when lower, less.
Just about everybody in this thread is in the 5s so most of us aren't the outliers. It would be interesting to see if people with the 7s might have different approaches reflected in their numbers?
Well, just out of curiosity, I'm one of those people with an A1c in the 7s, and out of curiosity I checked. I test 10x most days (give or take 1-2x a day) and my average for the past 90 days was 9.0 (162 mg/dl) with an SD of 3.8 (68 mg/dl). So that's definitely outside of your 20-30 range. I think the way that I look at it in my head is that most of my readings fall within +/- 3.8 mmol/L of my average of 9.0, and actually to me (at the moment) this makes perfect sense in my head, but I'm also not a math person unless I make myself sit and think about it for a bit ... I think this is how most people think of the number even if it's not totally accurate (like I said, maybe a different term should be used?).
