Page 80 - Glucose Monitoring Devices

P. 80

CHAPTER

Modeling the SMBG

measurement error 5

Martina Vettoretti, PhD
Department of Information Engineering, University of Padova, Padova, Italy

SMBG measurement error
Measurements collected by portable self-monitoring of blood glucose (SMBG)
devices approximate blood glucose (BG) concentrations. The difference between
SMBG measurement and true BG concentration represents the SMBG measurement
error (or observational error). SMBG measurement error is typically assessed by
comparing SMBG measurements to temporally matched reference measurements
collected by a laboratory measurement system known to have small measurement
errors. As all measurement instruments, SMBG devices are affected by both system-
atic and random errors. The systematic error is the predictable error component,
which inﬂuences observations consistently in one direction. For example, a glucose
meter that consistently under/overestimate glucose concentration is affected by a
systematic error. Commonly, systematic errors are quantitatively assessed by calcu-
lating the average absolute or relative difference between SMBG measurements and
reference values, also called bias. The magnitude of the systematic error deﬁnes the
accuracy of the measurement instrument. The random error is the unpredictable
error component, which leads to inconsistent values when repeated measurements
of the same quantity are performed. Random errors are usually quantiﬁed by assess-
ing the standard deviation (SD) or the coefﬁcient of variation (CV) of the differences
between SMBG measurements and reference values. The magnitude of the random
error deﬁnes the precision of the measurement instrument.
Several studies have assessed the accuracy and precision of SMBG devices on
the market, by comparing SMBG measurements with reference samples collected
by high-accuracy and precision laboratory instruments [1e8]. These studies show
that performance in terms of accuracy and precision can be very different for
different devices. For example, Freckmann et al. [2] compared the accuracy and
precision of 43 SMBG devices and demonstrated how the error’s bias and CV varied
signiﬁcantly with the considered device (see, e.g., Fig. 2 in Ref. [2]). A similar result
was obtained in a recent work by Klonoff et al. [8] (see, e.g., Table 4 in Ref. [8]).
Besides the type of device used, the characteristics of the SMBG measurement error
may be inﬂuenced by many other factors, such as the hematocrit level [7,9] and
ambient conditions [10,11], and can vary when different lots of test strips are
used [12]. Literature studies also suggest that, even when these factors are ﬁxed

75 76 77 78 79 80 81 82 83 84 85