Page 164 - Glucose Monitoring Devices

Numerical (statistical) accuracy 165

Hypoglycemia (BG≤70 mg/dl) Euglycemia (70<BG≤180 mg/dl) Hyperglycemia (BG > 180 mg/dl)
Point Error-Grid Zones Point Error-Grid Zones Point Error-Grid Zones
A P D P E P A P B P C P A P B P C P D P E P
A R
Rate Error-Grid Zones uC R
B R

lC R
uD R
lD R
uE R
lE R
Accurate Readings Benign Errors Erroneous Readings
FIGURE 8.5
Error matrix combining P-EGA and R-EGA.

matrix, clinical accuracy occurs when both P-EGA and R-EGA values are within
zones A or B. Clinically, benign errors are those where P-EGA values fall within
zones A or B and there are zone C, D, or E errors in the R-EGA that would be
unlikely to result in clinical action or negative outcomes. Hypoglycemic BG values
always require treatment regardless of rate data. Therefore, when the P-EGA value
is in the A zone during hypoglycemia, only R-EGAvalues that fail to detect rapidly
falling CGM rates (upper D zone) or signify rapidly rising CGM rates when the
reference rate is actually rapidly falling (upper E zone) are erroneous readings.
When the CGM accurately detects euglycemia, only R-EGA values in the E
zone (rapidly rising or falling when reference BG is rapidly changing in the
opposite direction) would suggest a clinically negative outcome. When the
CGM accurately detects hyperglycemia, only failure to detect a rapid rise in BG
(lower zone D error) or detecting a rapid rise or fall in BG when the opposite
change is occurring (E zone erroneous rate errors) would lead to a clinically
negative situation.

Numerical (statistical) accuracy
Numerical accuracy of CGM has been analyzed using familiar statistical tech-
niques including average BG, standard deviation, linear regression, correlation co-
efficients, and coefficients of variation. Most data have been presented in terms of
differences between CGM and reference BG values expressed as mean and median
relative difference and/or mean absolute relative difference. The FDA has affirmed
the Clinical and Laboratory Standards Institute (CLSI) guidelines on performance
metrics for CGM POCT05-A [54] as the appropriate criteria for premarket regula-
tory evaluation of CGM. Those guidelines state that “there is no consensus on