Using Regression Analysis to Predict A1c Measures

My wife, correctly, made fun of me for running a regression analysis using my morning AM blood sugar readings to predict my A1c readings.

I have data from every single morning to match up with my 7 A1c readings.

So, I tested to see if the A1c reading was most closely correlated with the 90-day, 60-day, or 30-day average of the morning readings.

It turned out that it was most correlated with the 90-day reading. For stats minded folks, here were the R-squared results:
90-day: 81%
60-day: 69%
30-day: 66%

I think I was expecting a much higher reading. Even 81% didn't seem that high given that both of these data points measured blood sugar.

On my last reading, the regression predicted a 6.6 and the actual reading was 6.3. It should be noted that the 95% highest and lowest bound was a prediction of 3.2 to 10.1!

Looks like we'll need more data to draw any conclusions.