Page 159 - Glucose Monitoring Devices

160 CHAPTER 8 Accuracy of CGM systems

yielding various results [18e21]. For example, it was hypothesized that if a
glucose fall is due to peripheral glucose consumption the physiologic time lag
would be negative, that is, fall in IG would precede fall in BG [15,22]. In most
studies, IG lagged behind BG (most of the time) by 4e10 min, regardless of the
direction of BG change [17,18]. The formulation of the pushepull phenomenon
offered reconciliation of these results and provided arguments for a more complex
BGeIG relationship than a simple constant or directional time lag [21,23e25]. In
addition, errors from calibration, loss of sensitivity, and random noise confound
CGM data [26]. Nevertheless, the accuracy of CGM has been steadily increasing
[27e31] and has reached a point where CGM could be used as a replacement
for traditional BG measurement [32]. Increased overall accuracy and reduction
of outliers were due to both improved sensor technology and advances in algo-
rithmic signal processing used to convert the raw electrochemical sensor data
into calibrated BG values [33e35].
In addition to presenting frequent data (e.g., every 5e10 min), CGM devices
typically display directional trends and BG rate of change and are capable of alerting
the patient of upcoming hypo- or hyperglycemia. These features are based on
methods that predict blood glucose and generate alarms and warning messages. In
the past 15 years, these methods have evolved from a concept [36] and early imple-
mentation in CGM devices [37,38], to elaborate predictive low glucose suspend
[39e41], advisory [42], and artificial pancreas systems [43e46].
Because the clinical adoption of CGM increases [47] and contemporary systems
use a CGM signal to discontinue insulin delivery (e.g., low glucose suspend), or
modulate insulin delivery up and down (e.g., artificial pancreas), it is imperative
that the information CGM provides be accurate and reliable. The methods for assess-
ment of CGM accuracy are reviewed below.

Clinical accuracy
Accuracy of CGM systems includes both clinical and numerical (statistical) compo-
nents and each must be assessed differently. Clinical accuracy always asks the ques-
tion “Will the treatment decision made on the basis of a particular BG monitoring
system reading be correct?” The information provided by CGM is much different
from that provided by a single self-monitoring of blood glucose (SMBG) result.
SMBG results are static. They represent a single point in time that may or may
not bear a relationship to previous test results. CGM readings are part of a process
in time, a moving dataset that includes speed and direction as well as a BG value
[48]. A good analogy is a difference between a still photo of a cue ball on a billiard
table and videos that show that cue ball rolling left or right, fast or slowly, toward the
eight balls or in a direction that will miss contact with any other ball. In other words,
BG fluctuations are a continuous process in time. Each point of that process is
characterized by its location, speed, and direction of change. Thus, at any point in
time, the BG value is not only a number but a vector with a specific bearing.
Fig. 8.1 presents this concept depicted over the grid of the original error grid