COMMENTARY

# The Problem With ADAGE (A1c-Derived Average Glucose Equivalent)

George Treviño, PhD

Disclosures

November 27, 2007

ADAGE is the acronym given to an international study performed at 10 centers in North America, Europe, and Africa, to confirm the relationship between average glucose and A1c levels. ADAGE was introduced at the ADA 2007 Scientific Sessions by Dr. David Nathan (Harvard University Medical School, Boston, Massachusetts). The goal of the study is to report glycohemoglobin results to patients as A1c-derived averages rather than strictly A1c values. This strategy has the advantage of reporting chronic glycemia in the same units as patients' self-monitoring, and presumably adds clarity for diabetic patients looking to manage their disease (see related article by O'Riordan for further elaboration). Problems, though, underlie this procedure.

To a diligent diabetic who measures and records blood sugar at regular intervals on a daily basis, a blood sugar average is simply the sum of N successive glucose values divided by N. The classical relationship between glucose and A1c, however, is a weighted average; ie, the sum of N successive glucose values, each multiplied by a known weight factor then added together, the weighted sum then divided by N. There's also a constant, whose units are the inverse of glucose units, introduced to nondimensionalize the determined A1c value. Both the patient average and the A1c average are generated from the same set of reference values save that the A1c average has the cumulative effect of the weighting (and the constant). As clearly stated in O'Riordan's article, the intent of ADAGE is to measure the A1c of a given patient and from that measurement deduce an estimate of the patient's average glucose.

Given, though, that average glucose is unaffected by "scrambling" (randomly rearranging) the glucose values while the value of A1c is not, the same average glucose can produce different A1c values. As an illustrative example consider 2 glucose values, 130 and 140, and 2 respective weights, 3 and 1. The related A1c value is proportional to (0.5)[(130)(3)+(140)(1)] = 265 while the patient average is 135. If the glucose values are reordered to instead read 140 and 130, the patient average remains unchanged but the related A1c value, because the time sequence of the weights is uninvertible, is now proportional to (0.5)[(140)(3)+(130)(1)] = 275. When averaging more than 2 glucose values, with more disparity in the weights and the glucose values, the difference is more pronounced.

The bottom line is that, in the weighted-average format, the (time) order of occurrence of glucose values affects the A1c value while in the patient-average format it doesn't. This nonuniqueness questions the viability of the ADAGE result.