Here is a copy of the risk chart correlated with A1Cs originally posted by Ron Sebol, who has given permission for displaying it here. This also appears in Gretchen's book. It's long and a little technical but well worth a read. Here goes:
Ron Sebol wrote:
This is the second of the previously posted tables dealing with absolute risk of retinopathy versus A1c. The A1c is presumed constant over the 9 years addressed in the original data and the 15 years addressed by extrapolating along the curves that were fitted to the original data. Each A1c trace in the family of curves was determined to be of the form LN(Y)= a + bX where Y is the incidence of 3 step retinopathy progression (on a 9 step evaluation scale) in a particular year and X is the year itself. Seen on a semi log graph, each of the curves in the family is a straight line. Rather surprisingly, the coefficient in the a + bX is a straight line, though a change in sign is needed for low A1c values. The coefficient is described by a constant plus K*LN(A1c)/A1c, where K is also a constant. Thus, it was possible to arrive at an expression for A1c of 6 and of 12, even though they were not contained in the original data. The original charts made no distinction between 1st time occurrence and repeat occurrence. To answer the question about 1st time risk, I had to assume that in each year, those with a history of retinopathy were all going to have a repeat of the prior years progression. While that is unlikely to be literally true, it is probably close enough to true to put the calculation of 1st time risk in the ball park. Of those with no history, it was assumed that occurrence in any given year was random. That allowed an occurrence of 1st time problems to be calculated for each year by subtracting the occurrence of problems in the prior year. The growth in occurrence year to year, therefore, is all from the group having no prior retinopathy history. The probability of no 1st time retinopathy through a given year is the product of (1 minus the 1st time probability in year one) times (1 minus the 1st time probability in year two) times (1 - probability of 1st time in year three), etc. through the year in question. That figure is then multiplied by (1 - prior year incidence) and the result subtracted from one.
This mathematical background information is for the list archive record. For most on the list, it can be ignored.
The second of the two charts in the original posting is the main topic of this message. First, please discard the 2nd table of that original. Not only was it mangled by uninvited line wrapping, it was wrong. I have recently revised the probability equations in a way that I think is closer to correct. As before, the table will contain the cumulative probability that, at a particular A1c, there will be a sudden appearance of retinopathy progression of three steps for the first time. As time passes, that becomes ever more probable and, at the higher A1c levels, that increase in likelihood grows more rapidly even to the point of becoming a near certainty. In the real world, such data are not nearly as well behaved as the family of curves that were used as the source for this analysis. The actual data points lie above and below each line in the curve family and the curves only represent a least squares fit to the real data points. Since the true data is rather noisy, consider the curves and the tables derived from them to be approximate. That will account for exceptions to the table seen in the real world.
Nevertheless, it is undoubtedly true that for A1c, lower is better. That for A1c of 8 or above, the long term risk is entirely unacceptable because it becomes too high after 15 years and by my yardstick is too high even after 8 years. A one in ten chance of having those complications is more of a chance than I care to risk. I am a bit biased when it comes to such matters because I once had a paralytic disease that had one in four million odds in my favor. If you are in the ten percent waiting for your laser eye surgery appointment, it is scant comfort to know that the other 90% of your friends have a few years to go yet.
This chart and text used by permission of Ron Sebol at the DSM list (Diabetes Self Management).
To subscribe to DSM, send an e-mail message to
listserv@listserv.tbinet.org
Put a couple of asterisks in the subject line. In the message body write the words
Subscribe DSM-L and then your true first name and true last name. For
example, Subscribe DSM-L John Doe
Then they'll send you a confirmation letter and you're in.
* * *