Page 47 - Glucose Monitoring Devices

44 CHAPTER 3 Clinical evaluation of SMBG systems

FIGURE 3.1
Error griddzone Adclinically accurate, zone Bdbenign errors, zones Cdovercorrection
errors, zone Ddfailure to detect and treat errors, zones Ederroneous errors.

values might result in BG values outside of the intended therapeutic range. Zone D
estimates are considered “dangerous failure to detect and treat” errors: the estimated
BG is within the therapeutic range while the actual BG is either above or below that
range; treatment is needed but would be omitted. Finally, zones E values are consid-
ered “erroneous treatment” errors. Estimates in these zones would suggest the need
for a treatment exactly opposite to that needed, that is, giving insulin for a BG less
than 70 mg/dL or giving glucose for a BG greater than 180 mg/dL.
Even before the results of the Diabetes Complications and Control Trial were
known, patient-determined SMBG was becoming a routine management tool among
those persons injecting insulin to manage their disease. The commercial develop-
ment of self-blood glucose monitors began to grow significantly and numerous
descriptions of their user-friendly attributes and statistical accuracy began to appear
in the scientific literature. Early reports of accuracy (SMBG values vs. Reference
values) included correlation coefficients, linear regression equations, percent devia-
tion, precision indexes, and standard error of measurements, but no analyses of
clinical accuracy [2,3]. The application of EGA to previously published statistical
analyses demonstrated that clinical accuracy did not always correlate with the statis-
tical data. In fact, no SMBG system whose data were examined using EGA produced
a high percentage of clinically inaccurate or unacceptable BG readings [3]. The most
common errors in these early studies (less than 5%) were zone D failures to detect
glucose values outside the treatment range. Reporting clinical accuracy in
performance trials of SMBG systems using EGA became more widely used and