Page 46 - Glucose Monitoring Devices

CHAPTER

Clinical evaluation of

SMBG systems 3

William L. Clarke, MD
Profesor, Emeritus of Pediatric Endocrinology, Department of Pediatrics, University of Virginia,
Charlottesville, VA, United States
Systems for self-monitoring of blood glucose (SMBG) can be assessed in terms of
their analytical or clinical accuracy. Analytical accuracy is a method to quantita-
tively assess how close the SMBG measurement is from a reference value; clinical
accuracy qualitatively describes the clinical outcome of making treatment decisions
using the measurements provided by the SMBG device. Error grid analysis (EGA) is
a tool largely used to evaluate SMBG systems in terms of their clinical accuracy.
EGA was originally designed as a metric to describe the clinical accuracy of blood
glucose (BG) estimates by subjects with type 1 diabetes (T1D) participating in a
research project where they estimated their BG level and then immediately
performed SMBG [1]. Conventional statistical analyses of individual subjects’
data revealed significant correlations between estimates and measurements but did
not address the type of error made nor the clinical significance of estimation errors.
Similarly, calculating percent deviation between estimates and actual BG levels
(i.e., examining their analytical accuracy) failed to address these issues. In fact, a
30% deviation between an estimated BG of 120 mg/dL and the corresponding actual
BG of 156 mg/dL would not be as clinically significant as a 30% deviation of an
estimate of 80 mg/dL when the actual BG is 56 mg/dL. Thus a different type of anal-
ysis of the clinical implications of differences between estimated and actual BG
levels was needed. EGA takes into account (1) the actual BG value, (2) the absolute
difference between the estimate and the actual BG, and (3) the clinical implications
of self-treatment decisions based on the estimated BG. EGA is always based on the
individual’s therapeutic or target BG range and assumes that treatment of BG values
outside the treatment range will always occur.
The error grid (Fig. 3.1) is a two-dimensional graph divided into five zones of
clinical accuracy. Zones A and B are considered clinically accurate and acceptable,
respectively. Zones C, D, and E are clinically inaccurate or unacceptable. In zone A,
estimated BG values fall within 20% of the reference BG (SMBG or laboratory
determined BG) or are 70 mg/dL or less when the reference BG is also 70 mg/dL
or less. The estimated BG values (“benign errors”) in zone B are greater than
20% different from the reference, but would most likely not result in dangerous
clinical decisions, that is, hypoglycemia or severe hyperglycemia. Zone C estimates
are designated “overcorrection errors” because treatment decisions based on these

Glucose Monitoring Devices. https://doi.org/10.1016/B978-0-12-816714-4.00003-X 43
Copyright © 2020 Elsevier Inc. All rights reserved.