• Please Remember: Members are only permitted to share their own experiences. Members are not qualified to give medical advice. Additionally, everyone manages their health differently. Please be respectful of other people's opinions about their own diabetes management.
  • We seem to be having technical difficulties with new user accounts. If you are trying to register please check your Spam or Junk folder for your confirmation email. If you still haven't received a confirmation email, please reach out to our support inbox: support.forum@diabetes.org.uk

Exactly same HBA1C test result over 4 months

I used a polynomial line or the power one. I deleted some rogue high readings first. Occasionally i’ve had high readings that when I re-took them a few minutes later were closer to expected.
Never really been tempted to try and fit anything to short term results - bit too much of the "fiddlers elbow" about them for me. I have contemplated trying some fourier analysis to see if there is any periodicity in the way things go up and down but it is so long since I did that sort of stuff I would not know where to start these days.

Somewhere back in the forum is a post where I reported the results of tests taken on all 10 digits in quick succession. The variation in those results rather put me off trying to get too involved in trying to interpret anything other than quite big changes in individual readings.
 
Never really been tempted to try and fit anything to short term results - bit too much of the "fiddlers elbow" about them for me. I have contemplated trying some fourier analysis to see if there is any periodicity in the way things go up and down but it is so long since I did that sort of stuff I would not know where to start these days.

Somewhere back in the forum is a post where I reported the results of tests taken on all 10 digits in quick succession. The variation in those results rather put me off trying to get too involved in trying to interpret anything other than quite big changes in individual readings.
1750174723166.pngthese are my MBA1C values.Diagnosed & started medication and reduced mainly sugar intake on 19th November. Looks like all blood cells replaced by mib Feb which is 90ish days later. Then reduced Gliclazide to zero so looks like it did nothing as HBA1C remained the same. Next steps reduce Carbs or more Metformin.
 
Interesting @paul.a.thompson .

Were you checking blood glucose during this period? I have found very good correlations between HbA1c results and the average blood glucose over the 90 days prior to it being taken. I have never looked for correlations with averages for less than 90 days. Might try that tomorrow and see what I get.

The large drop in HbA1c... what do you put that down to? Are their factors other than medication that might account for it?
 
Just done a quick and dirty check and find this...

1750186035223.png

It is a correlation between my HbA1c results and my average waking reading over the 30 days prior to the HbA1c being taken. The correlation is better than the 90 day data. The data covers a 6 year period. Must think about this.
 
Interesting @paul.a.thompson .

Were you checking blood glucose during this period? I have found very good correlations between HbA1c results and the average blood glucose over the 90 days prior to it being taken. I have never looked for correlations with averages for less than 90 days. Might try that tomorrow and see what I get.

The large drop in HbA1c... what do you put that down to? Are their factors other than medication that might account for it?
Basically as soon as I was diagnosed I stopped having any sugar at all in drinks, deserts etc.
Just done a quick and dirty check and find this...

View attachment 35831

It is a correlation between my HbA1c results and my average waking reading over the 30 days prior to the HbA1c being taken. The correlation is better than the 90 day data. The data covers a 6 year period. Must think about this.
Only got 6 months of data and 3 months of that is when I'm bringing my blood sugars down from 126. I plotted my waking values. My finger prick values seem to be going up slightly recently but my HBA1C is level with the last reading a little less but I don't think there are enough data points as yet.1750190932389.png
 
I'm guessing that your diet changes had as much to do with your drop in Hba1c as did the medication. There are so many factors involved that it is almost impossible to sort out what is affecting what, it is what makes looking at BG data both frustrating and intriguing at the same time.

What might be interesting is to plot your waking readings against date to see if they reflect your drop in HbA1c. I have long had the idea that a waking reading was probably the better guide to just where things were with respect to BG control but never really got a decent numerical handle on it.

PS... apologies for treating your system somewhat dispassionately, as if it were a chemical plant. Used to do that sort of thing for a living and old habits die hard. For me the answers (if there are any) are more likely to be found in the numbers than in opinions
 
@Docb

Please elaborate on your point about the answer being in the numbers. I agree with that but they need interpretation in terms of a theory, model or even an option. In themselves numbers do not tell you much and can lead to spurious correlations.
 
I think it is rather wrapped up in your use of the term, spurious correlation. For me a correlation has to be backed up by numerical data which has been statistically analysed. That will give you probability of a correlation being due to a real effect or random chance. There is no such thing as a spurious correlation in that world.

Take the graph I showed above. Mk 1 eyeball would suggest that there is a correlation between my HbA1c results and the mean of my waking results in the preceding 30 days. The R^2 value tells me that there appears to be linear relationship with a very high probability that it is not due to random chance. I can use the numerical relationship with confidence in predicting my Hba1c results. I have looked at this data further and found that the correlation coefficient increases until you get to the BG average in the last 30 days. Beyond 30 days and up to 90 days it changes very little.

This tells me that my average waking reading over the last 30 days can be used to predict, with a very high degree of confidence, my current HbA1c, something shown by my last couple of Hba1c tests where my predictions were very close to the actual result.

As always, this conclusion applies to me and my data. I would love to know whether it has wider validity but do not have enough data to express an opinion on the point. 😉
 
I'm fairly new to this but from what I've seen 30 days of data (either 30 finger pricks or HBA1C at beginning and after 30 days) can give you a good prediction of finishing HBA1C value. The most frustrating thing I've found is the medical insistance on not repeating HBA1C tests until after 3 months. If you change medication or diet then after a month you can predict and be fairly sure what difference it is going to make. Ok if you've got things stable then no need to test more regularly. But if you're trying to get things under control as quickly as possible its wasting time.
 
I think it is rather wrapped up in your use of the term, spurious correlation.
Yes, I used the term cover a multitude of sins and save time before going to bed. In particular the interpretation of statistics by non-statisticians to support points of view without understanding the underlying 'mechanisms' or even looking at the data.

During my postgraduate studies I was asked to analyse the key factors causing children from deprived areas in a metropolitan region to be stunted in comparison with children from more affluent areas. There had been a major survey designed to show this was the case and a huge amount of data had been collected. A statistician had done some form of cluster analysis showing the children fell into two distinct groups: case proven and accepted by the authorities. A few minutes looking at the data revealed that one group was only girls and the other only boys.

Coming closer to home in this forum, studies have shown that fatty liver is a risk factor for T2D and T2D is a risk factor for fatty liver. They have also shown about 60%-70% of T2Ds have fatty liver. However Prof Roy Taylor and his team have demonstrated that all T2Ds they have measured, other than those with pancreatic and other complications, have a fatty liver. The difference between 60%-70% and 95%-100% could well be due to undetected cases of fatty liver.

Anyway from the papers I have seen it is only just being generally accepted that fatty liver is a precursor to T2D, and then T2D only increases the risk of cardiovascular disease, liver disease and other serious conditions associated with metabolic syndrome. In other words we need models as well as data and statistics.

Take the graph I showed above. Mk 1 eyeball would suggest that there is a correlation between my HbA1c results and the mean of my waking results in the preceding 30 days. The R^2 value tells me that there appears to be linear relationship with a very high probability that it is not due to random chance. I can use the numerical relationship with confidence in predicting my Hba1c results. I have looked at this data further and found that the correlation coefficient increases until you get to the BG average in the last 30 days. Beyond 30 days and up to 90 days it changes very little.

This tells me that my average waking reading over the last 30 days can be used to predict, with a very high degree of confidence, my current HbA1c, something shown by my last couple of Hba1c tests where my predictions were very close to the actual result.

As always, this conclusion applies to me and my data. I would love to know whether it has wider validity but do not have enough data to express an opinion on the point. 😉
I should have complimented you on your graph. To my mind it's evidence that HbA1c and BG are likely to closely related at a personal level. We know they are loosely related in a population due the factors which cause 'everyone to be different'.

It also supports the 50% of HbA1c due to the last 30 days rule of thumb. BG levels for previous 60-90 days might have to be included if BG levels were (say) coming down due to dietary changes.

Bayes tells me your conclusion is highly like to prove of wider validity.
 
The other interesting thing is that I found that correlation in the numbers I had collected ..... I did not set out to collect numbers to prove some theory I might have. In my days of production performance analysis, historical data was all I had to work with, along with strongly held but differing opinions of the individuals involved. Using the data to try and reconcile the arguments between people was part of the process. Generally everybody had a point, and looking at the numbers showed when and where each individuals point was dominant.

I agree with you that using waking data to predict HbA1c might have wider validity - not easy to prove since it requires a lot of individual data, both daily readings and HbA1c results over a prolonged period.
 
In my days of production performance analysis, historical data was all I had to work with, along with strongly held but differing opinions of the individuals involved. Using the data to try and reconcile the arguments between people was part of the process. Generally everybody had a point, and looking at the numbers showed when and where each individuals point was dominant.
As it happens my thesis was about a similar situation. Not production but competitive bidding in an industrial contract market. The people involved could not understand the volatility of their competitors' bids. Some number crunching indicated each used a different method of costing most of the time. That was a long time ago and I imagine AI could do better today.
 
Back
Top