you want a study population you have a different calibration. I just raise it as a problem that I never saw until I started doing that research.

DONALD BODENNER: It was a very big consideration, and it still remains one. The idea, I should stress, is between an individual calibration, which is what you are discussing, where every time you do one of these analyses on individual patients you have to do a new individual calibration, versus a general calibration that you can apply to the entire population, off-the-shelf technology. That was a main concern that we had as well, and we still do.

JOAN CONWAY: The other comment I have is I know that this is being used to measure oxygen concentration in blood.

DONALD BODENNER: In a variety of places.

DONALD McCORMICK: Isn't it true that you have to fingerprint based on absorptions … My point is, is it not quite severely limited by the number of metabolites that you can reliably know you are looking at?

It is going back to Johanna Dwyer's question, because I do not think you can select a large number of things and filter them out.

DONALD BODENNER: Once again, getting back to how the equations are made, if you can find an overtone or any kind of a band in the NIR region that absorbs adequately that is representative of interfering substances, then their contribution to the glucose absorbance can be accounted for.

ROBERT WOLFE: Yes, but the question is how many do you need for a generalizable calibration equation?

DONALD BODENNER: We found that you can get by with about 200 in terms of glucose measurement, but we have not looked at a lot of the different potential contributors, especially in diabetics, or people in dialysis. What, for instance, happens when the BUN, which is a measure of renal failure, is 150? Or when patients are profoundly anemic? There are a large number of pathophysiolgic states where it is unclear whether this type of analysis will hold true.