Page 84 - Glucose Monitoring Devices

Literature models of SMBG measurement error 83

CA) device. Nevertheless, none of these models was validated, for example, by com-
parison of error realizations generated by the model versus an independent set of
observations.
Another critical point of the simple Gaussian PDF model concerns the symmetry
assumption of SMBG measurement error distribution. In fact, the histograms
reported in Chan et al. [39] show that some SMBG devices present an asymmetric
error distribution, calling for the use of PDF models allowing for nonzero skewness.
This is visible also in Fig. 5.1, where the SMBG relative error distribution is reported
for a sample collected with the One Touch Ultra 2 (Lifescan Inc., Milpitas, CA)
device.
To overcome the limitations of the simple Gaussian model, our research group at
the University of Padova proposed a new methodology to model the SMBG mea-
surement error’s PDF, which was presented in Vettoretti et al. [40]. Specifically,
this method (i) deals with the variability of SMBG error characteristics with BG
by using multiple PDF models in different zones of the glucose range and (ii) takes
into account the asymmetry of the SMBG error distribution by using PDF models
allowing for nonzero skewness. A similar approach was used in a recent work by
Campos-Naneez et al. [42], where the error of 43 commercial SMBG devices was
modeled using two Johnson distribution PDF models, one to describe the absolute

FIGURE 5.1
Histogram (absolute frequencies) of the SMBG relative error (in percentage) for a sample
collected with the One Touch Ultra 2 device.
Adapted from Vettoretti M, Facchinetti A, Sparacino G, Cobelli C. A model of self-monitoring blood glucose
measurement error. Journal of Diabetes Science and Technology 2017;11(4):724e735.