• Guest - w'd love to know what you think about the forum! Take the 2025 Survey »

Blood sugar readings and Standard Deviation.

J

jpg

Active Member
Messages
36
I am electrical engineer and occasionally have need to use statistical tools in the course of my work. Apart from the generally well understood stats concepts such as the mean (average) of a set of readings, there is also the standard deviation (SD). I don't want to bore people with a load of maths but I will just say that the SD of a set of readings is a measure of the variability of individual readings from the mean of all the readings.

Imagine a graph of your blood sugar readings over time. The graph will show a degree of ups and downs as your blood sugars vary over the day. Typically just before eating you will have a relatively low reading; your reading 2 hours later will (usually) be somewhat more, dependant on your carbohydrate intake and/or medications. If your control is good you should see quite small variations. If your control is not so good you are quite likely to see a very spikey graph. The spikey graph will have a much higher SD figure than the smoother graph.

My interest in SD and blood sugar readings was spiked (excuse the pun) beause I noticed that the Glucofacts software had a column showing SD, so there must be people about who see some significance in the SD value of blood glucose readings.

The important thing, to me at least, is that the average reading, and to some extent the HbA1c, can only tell you part of the truth. It is possible, indeed likely, that you may have a good average and good HbA1c, but your readings spike all over the place. Take these hypothetical examples of 6 daily readings:

Good control 5.8 7.5 5.2 7.1 5.2 7.9 - Average 6.45 SD 1.2
Not so good control 5.0 9.2 3.8 6.9 4.0 9.8 - Average 6.45 SD 2.6

Note that the poorer control is shown by a higher SD. Even though the average (and by extension, the HbA1c) is reasonably good, the second set of readings suggests that an individual with these readings willl spend some period of time in the "danger" zones (<4.0, >8.0) whereas the person with readings in the first set will not spend any time in the danger zone (although we can never be sure what spikes/dips may occur between blood tests).

In all my 13 years since diagnosis, I have not seen much discussion of SD, but I think it could be quite an important indicator. A quick google shows some American learned and informal dicussion on it, and it has been mentioned on this forum from time to time, but no-one seems to have given it any serious consideration.

A fairly non-technical website explaining SD:

http://www.robertniles.com/stats/stdev.shtml

jpg
 
We have had this basic argument many times on here. A HbA1c is only an average. Therefore it hides a multitude of sins, highs and lows. That is why I never cease to tell people frequent daily readings are far more relevant to any Diabetic. I am not too bothered about SD although I can see with your technical background it has relevance.

I don't plan to get too technical, we try to keep it plain and simple as much as possible.....otherwise people just glaze over and we have lost them. I just know what I think is more relevant to BG control.......it 'aint just relying on a 3 monthly HbA1c.......that only gives you the trend......upwards or downwards.

Ken
 
I think that SD is a useful tool, if one has enough readings to make it meaningful. However for most people a graph will show the same thing in a much more immediate and easily understood way.It has been suggested that the SD should ideally be less than a third of the mean.

Having said that, though common sense makes it sensible to aim for fewer spikes there isn't actually much evidence from long term trials in type 1 that day to day variability is as important as sometimes stated. ( actually there isn't much good evidence at all )

Technical bit warning!
: evidence from an analysis of the DCCT by E. Kilpatrick suggests that HbA1c determines the risk of microvascular complications and mean glucose determines those of macrovascular complications. People with the similar HbA1cs or mean glucose levels had similar risks.The day to day variability or how the HbA1c or mean glucose was 'achieved' was not significant. HbA1c and mean glucose were not perfectly correlated in the DCCT. Variability between HbA1cs, in other words widely fluctuating periods of control was significantly correlated with adverse risk
 
Does tissue damage occur with sustained high values, or or can accumulated shorter periods of high blood glucose be just as hazardous? If the latter, the duration of accumulated high BG could be mathematically calculated from the average and SD.

I agree with Ken that frequent testing is important. I'm an engineer and in some ways I treat this disease as an engineering problem but that's just me. Because the Glucofacts software has an SD column I'm going to plug in some values before and after Byetta treatment and see how it compares, although I could probably do that easier using Excel.

I'm pretty sure that there will have been a marked drop in SD when I started Byetta.

jpg
 
I don't see a response to the question posed by jpg, which is in my view a highly significant question, viz. "Does tissue damage [only] occur with sustained high values, or or can accumulated shorter periods of high blood glucose be just as hazardous? "

I was seeking guidance on this site as to when I should be carrying out tests and found the page "Blood Glucose Testing Times", and the question,"When should I test my blood glucose?". Just what I wanted to know. The answer offered there however is sadly quite useless...
"How often you test will depend partly on your experience, your diabetes type,
and your diabetes treatment regime. Some people with diabetes do not test their blood glucose regularly, and some do not need to. However, blood glucose testing is a very useful tool for people with diabetes." So when do I test ???

My doctor has not told me when I should test other than "vary the times of your tests", so I take random tests usually once in every 2 days, guessing beforehand the expected value (and I'm usually quite close) and enter the values in an Excel spreadsheet. I understand the argument for SD analysis, but I want to know if this analysis has any medical foundation to support.

In case someone wishes to offer me advice I should provide some basic data.
Male, type 2, 57yrs, BMI 25(static), Metformin 850mg + Gliclazide 80mg twice daily. HbA1c 6.7
My range of BG tests is from 4.6 to 12 mmol/L, and occasionally up to 14 and 15. Having attained the goal for HbA1c - after 2 years - I am now concerned that the peak values for BG may be dangerous for me in the long term.

Any answers? Thankyou in anticipation.

Roger
 
I don't claim to be an expert is statistics...far from it, but agree that the sd is, indeed an important consideration. My meter, (Accuchek combo displays the sd of my results over a 7, 14 and up to 90 day period. My 7 day results are ave 6.1, with an SD of 3.2.....showing exactly why looking at the sd is important...as for me,n ave of 6.1 looks pretty good...but the sd is awful! I have to admit though, I dod skew my own results however.....I tend to test before meal, driving and also when I "feel" low or high. If I am low...I then do multiple tests to make sure my BG is returning to "normal"...so a high proportion of my results are not taking through teh course of an ordinary day. Interesting to see that teh sd should be less than a 1/3 of the average...something else to bear in mind.
 
Personally, I can't say that SD is of any interest to me as I test frequently and do not just rely on HbA1c to see what my levels are doing throughout the day.

I can see looking at the graphs which my meter shows (OneTouch UltraSmart) exactly what is happening and when. I really don't need to further complicate things with another statistic.........however if you are interested in the subject here is a link to an article by an Endocrinologist which does mention standard deviation in the text.

http://www.bd.com/us/diabetes/page.aspx ... 1&id=14131

Check your blood glucose more often. Imagine if you checked your blood glucose only once a day, in the morning. If you drew a line connecting those morning readings you might have a nice flat line making it appear that there is little variation. You know better! If you start checking before each meal and even throw in a few blood glucose readings after meals, you will get a much truer picture of your blood glucose range.
Click to expand...

I have always maintained that people who have the same HbA1c can have significantly different Bg control........it really is only a trend indicator and should not be the only reading relied on. That's one of the reasons why I always advocate frequent daily testing. Checking only one a day or maybe once or twice misses out so much detail.
My opinion.
 
I have a very big excel sheet with all my logs and averages, sd's etc (also an engineer!).
I find the sd to be an excellent measurement of control.

If you manage to have a low average, and low sd then you are doing pretty well.

Using the sd allows you to track your progress over a period of a weekly or a month - rather than waiting the 6 months usually between hba1c's......
 
As an engineer (retired) and a T2, I too initially treated my newly (in 2002) diagnosed T2 as an engineering problem and applied the techniques and mathematical approach that I am / was well versed in.

I produced some wonderful graphs and Excel spreadsheets, and kept accurate records of everything.

Such an approach can give a sense of being in control and certainly helped me understand what was going on in my body.

The trouble is, T2 is a shifting target, it's slippery, it sneakily progresses, it cannot be quantified and dealt with like an engineering problem, it is a day to day thing, keep your BG down to the lowest you can, try to iron out the spikes, what works for me might not work for you, but I do find that a lot of wheat bran and fats help keep my spikes down.

I cannot help but feel, in the light of my 8 years experience, that T2 is much less of a science and a good deal more of an art.

SD statistics could be very interesting, but could we really do anything with them? Could we do anything different or better to improve them? I don't know, but I do know my Doctor has no interest at all in any test results I produce, only being interested in the latest A1c.

H
 
hallii said:
SD statistics could be very interesting, but could we really do anything with them? Could we do anything different or better to improve them? I don't know, but I do know my Doctor has no interest at all in any test results I produce, only being interested in the latest A1c.

H
Click to expand...


I totally agree H. My GP, who has little or nothing to do with my Diabetes care looks blank when I show him my carefully printed off, graphs, pie charts, averages and daily Bg level readings. He is only concerned with HbA1c too. :(

However when I forward the same information to my Endo and/or SDSN........a World of difference. The results are analysed, dissected and links made to events which have had an effect on my Bg levels. This allows them to help ME gain better control and avoid those peaks and troughs in my readings.

Now I have no doubt that SD may assist some, but if my Team were confronted with such figures.....I think they would run a mile. Much the same way many of us who may have no background in engineering.......it just complicates the thinking, the message that we propound to let people get a grip on their Diabetes control. It's bad enough as it is without factoring in yet another item. So as said before, SD is fine for some........I don't think it is something that us poor muddled Diabetic's really need to spend too much time on though ....... :)

I too have an engineering background, many many years ago, and 14 yrs as a Diabetic. I think what I have learned in the last 2 yrs has been of more assistance when controlling and analysing Bg levels than my engineering skills and knowledge. Being analytical and methodical is all in my view. The figures.......well, they tend to look after themselves without too much head scratching. :wink:

Each to their own I suppose.
 
It is most defintely an art rather than a science!

I have to say my doctor (also) doesn't care for figures and graphs, however, the DAFNE nurses were very interested and, as you say, took some time to dissect and offer their advice on various trends - which I found very helpful...

I personally think my docter see's my hba1c and thinks - tick. He fits the 'acceptable 'stats and moves on. I suppose there are probably others at the clinic who needs the doctors advice more than I do... However, I do wish my doctor would set me some more challenging targets, rather than treating me like another statistic.
 
jameshallam said:
I personally think my docter see's my hba1c and thinks - tick. He fits the 'acceptable 'stats and moves on. I suppose there are probably others at the clinic who needs the doctors advice more than I do... However, I do wish my doctor would set me some more challenging targets, rather than treating me like another statistic.
Click to expand...

I would say your gp will be more than happy with your a1c been in the lower 6's James, what challenge would you like him to set you?

Nigel
 
Bernstein started out as an engineer too and most of his ideas on control come, fundementally, from those engineering concepts.
Lois Jovanovic also believes in minimising "excusions". In otherwords less SD.
Both of these are people who I admire for their knowledge and sense.
Hana
 
Back
Top