Diabetes Talking

Question:

Thad…. your explaination sounds like a practical application….

I don’t know what you were listening to but it’s obvious that it wasn’t my post. I don’t see any practical application at all in it. I only see a description of an experiment. Charly was posting the model to *prove*

No models don’t prove anything. To prove something you run an experiment. What Charly posted was an experiment (or rather a model of an experiment ) to measure the length of A1C. the that the A1c reflects a 4? week average, which is not yet *proven*

That A1C reflects an ( exponential )average of bg is already a given, and I doubt that any health care proffesional even disputes it. The question is what the period of time it is an average over. That is not something that is proven, it is something that is measured. Both Charly and I have proposed experiments to do that. I just think that mine is significantly simpler to preform.   You use some pretty sloppy language in a post that is scientific in nature. Science is about precision and saying things acurately. You need to bone up on your science. Sadly, this was stuff that I learned in 7th and 8th grade.

Response:

- Hide quoted text — Show quoted text – You’re experiment seems to be flawed in this way. The question is the period of time over which A1C averages. In essence this is equivalent to asking what the decay constant/half life/time constant is. So say that you have a group of people that have their bg well under control. Arrange to have their control broken for a day or two ( give them a small measured amount of insulin to drop their average 10 points, or remove some insulin, or attach a glucose solution to non diabetics ). Then restore control. measure their A1C to see how it evolves over time. Plot the A1C vs time on good old semi-log paper. Viola you have the decay time. Of course I’m oversimplifying a little ( *OK* a lot ), but that seems doable. The hard part of this is that you need to modify peoples bg for a short while and then restore control. Also you need to measure A1C ( and bg ) with three-four digit accuracy. In order to get a good  set of numbers you would need at most one or two weeks  of measurements. Not several months.

Well, I’m not convinced.  This is, however, a technical issue we can persue via e-mail if you wish.  I suspect we can agree that whether or not a definitive experiment is possible, it has not been done and that those of us who are convinced of one averaging interval or another do so in the face of some conflicting evidence. — Charly Coughran

Response:

Thad…. your explaination sounds like a practical application…. Charly was posting the model to *prove* the that the A1c reflects a 4? week average, which is not yet *proven* — k t1 12 yrs now sending as Tiger Lily It would be very interesting to hear why definitive clinical data [that would show the averaging interval of HbA1c] is impossible to obtain.

– Hide quoted text — Show quoted text – Long explaination clipped. You’re experiment seems to be flawed in this way. The question is the period of time over which A1C averages. In essence this is equivalent to asking what the decay constant/half life/time constant is. So say that you have a group of people that have their bg well under control. Arrange to have their control broken for a day or two ( give them a small measured amount of insulin to drop their average 10 points, or remove some insulin, or attach a glucose solution to non diabetics ). Then restore control. measure their A1C to see how it evolves over time. Plot the A1C vs time on good old semi-log paper. Viola you have the decay time. Of course I’m oversimplifying a little ( *OK* a lot ), but that seems doable. The hard part of this is that you need to modify peoples bg for a short while and then restore control. Also you need to measure A1C ( and bg ) with three-four digit accuracy. In order to get a good  set of numbers you would need at most one or two weeks  of measurements. Not several months.

Response:

It would be very interesting to hear why definitive clinical data [that would show the averaging interval of HbA1c] is impossible to obtain.

Long explaination clipped. You’re experiment seems to be flawed in this way. The question is the period of time over which A1C averages. In essence this is equivalent to asking what the decay constant/half life/time constant is. So say that you have a group of people that have their bg well under control. Arrange to have their control broken for a day or two ( give them a small measured amount of insulin to drop their average 10 points, or remove some insulin, or attach a glucose solution to non diabetics ). Then restore control. measure their A1C to see how it evolves over time. Plot the A1C vs time on good old semi-log paper. Viola you have the decay time. Of course I’m oversimplifying a little ( *OK* a lot ), but that seems doable. The hard part of this is that you need to modify peoples bg for a short while and then restore control. Also you need to measure A1C ( and bg ) with three-four digit accuracy. In order to get a good  set of numbers you would need at most one or two weeks  of measurements. Not several months.

Response:

To all you folks that convinced us all that HbA1c tests show and average for a period of one month (not 3 like we all thought)… I need to get some hard copy evidence of this.  I looked on the ADA site to find something about Dr, Cerami’s speech but could not find anything.  I would like to have something that I could print out from some *official* source rather than just an email or post from someone as these are not too convincing. Something more than just the studies from the 80’s.   Thanks (in advance)   T1 for 33 years, pumping for the last 11

Response:

To all you folks that convinced us all that HbA1c tests show and average for a period of one month (not 3 like we all thought)… I need to get some hard copy evidence of this.  I looked on the ADA site to find something about Dr, Cerami’s speech but could not find anything.  I would like to have something that I could print out from some *official* source rather than just an email or post from someone as these are not too convincing. Something more than just the studies from the 80’s.

Ed Reid’s discussion of the issue in the FAQ has most of the relevant citations for the 4 week averaging interval. Note that while both Ed and I are convinced, there are articles in the primary literature arguing for other effective averaging intervals.  This is an instance where definitive clinical data is impossible to obtain, for reasons I’d be happy to try to explain if anyone cares. The result is that you have to choose which model you believe on some other grounds than its supporting clinical data. The FAQ is posted regularly and is available at http://www.landfield.com/faqs/diabetes/ among other places. — Charly Coughran

Response:

anecdotally…… at first diagnosis i tested my A1c every month, and it went down each month….. almost concurrently with my monthly average bg tests…. after the first 5 or 6 months (once i was "in control")….. i went back to quarterly testing……. this does not scientifically support the 1 month average theory, but my own experience concurs k

– Hide quoted text — Show quoted text – To all you folks that convinced us all that HbA1c tests show and average for a period of one month (not 3 like we all thought)… I need to get some hard copy evidence of this.  I looked on the ADA site to find something about Dr, Cerami’s speech but could not find anything.  I would like to have something that I could print out from some *official* source rather than just an email or post from someone as these are not too convincing. Something more than just the studies from the 80’s. Ed Reid’s discussion of the issue in the FAQ has most of the relevant citations for the 4 week averaging interval. Note that while both Ed and I are convinced, there are articles in the primary literature arguing for other effective averaging intervals.  This is an instance where definitive clinical data is impossible to obtain, for reasons I’d be happy to try to explain if anyone cares. The result is that you have to choose which model you believe on some other grounds than its supporting clinical data. The FAQ is posted regularly and is available at http://www.landfield.com/faqs/diabetes/ among other places. — Charly Coughran

Response:

It would be very interesting to hear why definitive clinical data [that

would show the averaging interval of HbA1c] is impossible to obtain.

Ok, here goes.  I should probably have said impractical rather than impossible, but they wind up meaning the same thing in the real world of research.  I should also have said "up till now". With the new technology, it should be possible to do it in a few years.  Whether anyone will pay for it is another question. I’m going to wave my hands a bit in the following to avoid long digressions into technical issues and still have it make sense. The task is to compare HbA1c values to a history of daily blood glucose averages, and determine to what averaging interval the HbA1c values most closely correspond.  We believe due to the length of red blood cell life, the maximum possible averaging interval is 4 months. The first question we encounter, is how do we get an accurate daily average bg measurement?  (For reasons that will become clear a little later, we have to have a good estimate of daily bg averages in order to produce a good estimate of the HbA1c averaging interval.) Bg levels fluctuate during the day.  They certainly respond to meals, and there are events during sleep that can affect them, e.g. Somogyi.  So our average bg measure must capture the effects of these events.  This leads to a picture of bg level rising and falling in 4 cycles over a 24 hr day, three meals and some overnight changes. (Note that this is a minimum number of cycles/day.  Exercise, snacks, and other events raise the number, but we’ll ignore them because that makes the problem even worse.) Elementary sampling theory says in this situation a minimum of 8 samples must be taken in order to capture the variability and produce an average that reflects the effects of that variability. Of course, the more samples you take, the better your estimate of the daily average glucose.  No data analyst I know would be happy with less that twice the theoretical minimum sampling frequency, but we’ll stick to 8/day. Our experiment requires, then, some number of volunteers to take a blood glucose test every 3 hours for 4 months. Unfortunately, even this rather daunting prospect isn’t really sufficient. If the volunteers are very stable, there will be no changes in average daily bg or HbA1c to measure.  Ideally, we would like to have a step function change in the signal which makes it easiest to track the effects of the change.  This means the volunteers should be stable at one daily average bg level, then switch to another stable bg level.  Since the maximum HbA1c averaging interval is 4 months, that means the experiment must now last 8 months.  4 months to assure the stability of the HbA1c, then another 4 months to watch the full effect of the change in bg to work out in the HbA1c. We’re now up to 8 months of testing every 3 hours 24×7. The larger the step function change, i.e. the bigger the change in daily average bg, the better the signal to noise ratio.  This means your volunteers are going to have to be stable at significantly higher than desirable bg levels for 4 months. This gives you an idea of some of the problems with getting a really good data set to answer the question.  There are, of course, a number of ways around these problems.  However, each short cut you take degrades the quality of the data set. As one example, you can substitute a fasting bg for the 8 daily bg tests as a metric for average bg value.  First order, you are interested in how the change in average daily bg relates to the change in HbA1c, rather than in the absolute correlation.  The use of the fasting bg metric introduces noise into the data since fasting bg is not perfectly related to daily average bg and any single measurement value is going to be noisier than the average.  The latter being one of the reasons we use averages in the first place. The noisier the data is, the harder it is to pick out the exact averaging interval of the HbA1c.  I will gladly confess that I haven’t done a sensitivity (error) analysis of any of these papers.  That would take getting the author’s raw data and, probably, a couple weeks of my time.  My educated guess is, however, that none of the data sets I’ve seen would allow you to distinguish between an averaging interval of 4 weeks and 3-4 months, much less peg the averaging interval with any level of confidence. I should point out, in all fairness, that none of the papers have tried to do the data driven analysis I have outlined here. Typically, they propose a model of the HbA1c reaction, which they analyze, then present some limited clinical data which they show to be consistent with the model.  The degree to which they are persuasive, then, has more to do with the model than with the clinical data. The length of this answer is the reason I didn’t insert it into my previous comments.  I hope it wasn’t too boring. — Charly Coughran

Response:

It’s funny you brought that up again.  An instructor at the Joslin Center mentioned, in passing, that the test covers two to three months.  I considered asking about it, but since I had walked in half an hour late, I figured I would wait for a later date to ask if there was any information challenging that. — ~ Mookie ~ T1 since 10/80

– Hide quoted text — Show quoted text – To all you folks that convinced us all that HbA1c tests show and average for a period of one month (not 3 like we all thought)… I need to get some hard copy evidence of this.  I looked on the ADA site to find something about Dr, Cerami’s speech but could not find anything.  I would like to have something that I could print out from some *official* source rather than just an email or post from someone as these are not too convincing. Something more than just the studies from the 80’s. Thanks (in advance) T1 for 33 years, pumping for the last 11

Response:

– Hide quoted text — Show quoted text – To all you folks that convinced us all that HbA1c tests show and average for a period of one month (not 3 like we all thought)… I need to get some hard copy evidence of this.  I looked on the ADA site to find something about Dr, Cerami’s speech but could not find anything.  I would like to have something that I could print out from some *official* source rather than just an email or post from someone as these are not too convincing. Something more than just the studies from the 80’s. Ed Reid’s discussion of the issue in the FAQ has most of the relevant citations for the 4 week averaging interval. Note that while both Ed and I are convinced, there are articles in the primary literature arguing for other effective averaging intervals.  This is an instance where definitive clinical data is impossible to obtain, for reasons I’d be happy to try to explain if anyone cares.

I care. I see that there are difficulties, but I don’t see any obstacles that can’t be overcome aand haven’t been overcome in other studies. But then I may not see all the problems.

Response:

Ed Reid’s discussion of the issue in the FAQ has most of the relevant citations for the 4 week averaging interval. Note that while both Ed and I are convinced, there are articles in the primary literature arguing for other effective averaging intervals.  This is an instance where definitive clinical data is impossible to obtain, for reasons I’d be happy to try to explain if anyone cares. The result is that you have to choose which model you believe on some other grounds than its supporting clinical data.

There was a thread about this not too long ago.  Most people are still under the impression that the test covers 3 months but the faq and other posters say it covers one month.  Dr. Biggs posted that at the recent ADA conference that the doctor who discovered the test stated that he has changed his thinking on this which seamed to put the topic to rest. .   It would be very interesting to hear why definitive clinical data is impossible to obtain. TIA, Marilyn T1 for 33 years, pumping for the last 11

Response: