Done it again.

Status
Not open for further replies.
Docb

Docb

Moderator
Relationship to Diabetes
Type 2
Just had an HbA1c done. The algorithm I have in my database to predict HbA1c from my blood glucose readings predicted 53.3, and the lab result was 54. That's two years in a row that my prediction has agreed with the lab.

So if you have 5000 data points, half a dozen HbA1c results, a decently organised data base and a bit of jiggery pokery you can quite accurately predict your HbA1c from finger prick readings as far as I am concerned.
 
Docb said:
Just had an HbA1c done. The algorithm I have in my database to predict HbA1c from my blood glucose readings predicted 53.3, and the lab result was 54. That's two years in a row that my prediction has agreed with the lab.

So if you have 5000 data points, half a dozen HbA1c results, a decently organised data base and a bit of jiggery pokery you can quite accurately predict your HbA1c from finger prick readings as far as I am concerned.
Click to expand...
I found the same as I test every day. Sounds like you've got way more info at your finger tips. Mine was a pretty basic average using a convertion chart. Still, interesting stuff.
 
Docb said:
Just had an HbA1c done. The algorithm I have in my database to predict HbA1c from my blood glucose readings predicted 53.3, and the lab result was 54. That's two years in a row that my prediction has agreed with the lab.

So if you have 5000 data points, half a dozen HbA1c results, a decently organised data base and a bit of jiggery pokery you can quite accurately predict your HbA1c from finger prick readings as far as I am concerned.
Click to expand...
That's pretty impressive. I've got 3,000-and-something readings so I'm a bit behind where you are, and they start from a few days after my diagnosis which means there are some high numbers in there. I've allowed +/-2 when I've tried to estimate my HbA1c since then and although my HbA1c result is always in that range I've never been spot on.
 
That’s amazing. The Libre estimated Hba1c isn’t anywhere near as accurate as that!
 
@Martin.A If you want to follow my trail, my algorithm is not that sophisticated. Fundamentally I have taken all my HbA1c results and calculated the mean of all the finger prick results taken in the 90 days preceeding the date of the HbA1c. Plotting one against the other gives a reasonable correlation. A regression line then gives you an equation relating one to the other and I use that equation to compute a current predicted HbA1c from the mean of finger prick results from the last 90 days.

I think it helps that I do quite a few random tests and so get a resonable picture of how my blood glucuse varies through the day. If I use the same algorithm but only using my waking reading, the prediction is not as good.

It also helps if you can construct a data base and write queries. That way you can automate most of the processes and cut out most of the fag. Anyone who is good at spreadsheets could do it no doubt but I used it as a means of teaching myself how to use a database.
 
Docb said:
I think it helps that I do quite a few random tests and so get a resonable picture of how my blood glucuse varies through the day. If I use the same algorithm but only using my waking reading, the prediction is not as good.
Click to expand...

That’s really interesting Doc. I was going to ask what your testing routine was. Sounds like the data points you have fairly accurately reflect your overall glucose profile!
 
You're on the best-fit HbA1c vs avg BG line. That's not very informative for somebody else's situation.
 
Agree with you @Eddy Edson. The average BG will be dependent on testing regime and that will vary from person to person. There is also the possibility that I have just been a bit lucky. Have to see if i get it right next year!
 
Docb said:
Agree with you @Eddy Edson. The average BG will be dependent on testing regime and that will vary from person to person. There is also the possibility that I have just been a bit lucky. Have to see if i get it right next year!
Click to expand...
Fundamentally it's not the testing regime but (mainly) difference in average red blood cell age.

No matter how accurately you measure avg BG, for most people the standard conversion won't deliver a result as close to their actual HbA1c as you get.

For reference, paper by some of the Harvard and MassGen researchers responsible for the empirical study leading to the standard conversion, looking at the variability due to differences in RBC average age: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5714656/pdf/nihms916510.pdf

The glycated hemoglobin assay (HbA1c) is essential for the diagnosis and management ofdiabetes because it provides the best estimate of a patient’s average blood glucose (AG) over thepreceding 2–3 months and is the best predictor of disease complications.

However, there is substantial unexplained glucose-independent variation in HbA1c that makes AG estimation inaccurate and limits the precision of medical care for diabetics.

The true AG of a non-diabetic and a poorly-controlled diabetic may differ by less than 15 mg/dL, but patients with identical HbA1c and thus identical HbA1c-based estimates of AG may have true AG that differs by more than 60mg/dl.

We combine a mechanistic mathematical model of hemoglobin glycation and red blood cellflux with large sets of intra-patient glucose measurements to derive patient-specific estimates ofnon-glycemic determinants of HbA1c including mean red blood cell age (MRBC). We find that interpatient variation in derived MRBC explains all glucose-independent variation in HbA1c. We then use our model to personalize prospective estimates of AG and reduce errors by more than 50% in four independent sets of more than 200 patients.

The current standard of care provided AG estimates with errors > 15 mg/dL for 1 in 3 patients. Our patient-specific method reduced this error rate to 1 in 10.
The paper reproduces the summary graph from the original study:

1704439889906.png

You seem to be very close to the best-fit regression line. Most people are further away.
 
I need to read that paper Eddy Edson - might take me some time.....
 
Docb said:
Just had an HbA1c done. The algorithm I have in my database to predict HbA1c from my blood glucose readings predicted 53.3, and the lab result was 54. That's two years in a row that my prediction has agreed with the lab.

So if you have 5000 data points, half a dozen HbA1c results, a decently organised data base and a bit of jiggery pokery you can quite accurately predict your HbA1c from finger prick readings as far as I am concerned.
Click to expand...
I wish I was as whizzy as others with spreadsheets etc re my HbA1c but I am not. So, my very amateurish way is to check with my monitor quite regularly and then see what the 90 day memory gives me. I then add another 3 for good measure. That does almost match any HbA1c I receive. I appreciate not scientific... I am more a student of English and History student. Or I could simply wait until my HbA1c appears on Patient Access. But am not patient enough boom boom.
 
I agree with the view that prediction algorithms correlating A1c with finger prick readings can be very accurate but are individualised by factors like RBC turnover. I too have applied regressions to my spreadsheet over three years and obtained highly consistent results - for me, an average of blood glucose readings of 5.4 over any 3-month period predicted an A1c in the range 38-42 at the end of that period, always born out by the actual outcomes. But it might predict an A1c of 33 for someone 30 years younger than me having a quite different RBC profile.
 
Yes, my correlation applies to me, to my method of testing and determining average BG and above all to my system and how it chooses to glycate its haemoglobin and kill off its old red cells. It is interesting to see that others can get their own very good correlations but it is a bit much to expect a universal correlation which can be applied to all.
 
Status
Not open for further replies.
Back
Top