Page 72 - Glucose Monitoring Devices

70 CHAPTER 4 Consequences of SMBG systems inaccuracy

Patient behavior plays a critical role in these assessments. Variations of the
in silico study presented here where patients are always compliant (bolus correctly
and on-time) are needed. Here, we see that the effect of error disappears, likely due
to a law of large numbers: in the absence of systematic bias, errors tend to cancel
each other out. Alternatively, one can conclude that noncompliant behavior
increases the effect of the inaccuracy: noncompliant patients have less opportunities
to correct, highlighting the importance of accurate measurements during those
times.

Limitations
The in silico study does not accommodate any long-term behavioral adaptations.
Although short-term behavior (meals, bolus, etc.) were considered in the study, it
is unlikely that a patient that experiences frequent hypoglycemic events will not
adapt. These adaptations could include switching BGM systems, adjusting their
insulin therapy, modify eating and exercise behaviors. In addition, our study
assumes that BGM system accuracy remains constant throughout the progression
of a patient’s disease. This ignores developments that might happen during the life-
time of the patient. Finally, the results were limited to CSII. It is likely that the
results can be extended to patients using MDI therapy, at least at a qualitative level.

Conclusions and future work
Assessing the effects of BGM system accuracy is a challenging task. On one hand,
the effects of poor decisions span a long time, from the immediate to the very long
term, affecting at the same time many aspects of a patient’s life. To complicate
things, behavior, technology, and environmental considerations impose limits on
how effective glycemic control can become. Isolating the role of BGM system
accuracy in this complex environment is nontrivial.
Simulation and systems modeling can shed light into this process. The ability to
stimulate metabolism, behavior, and technology as they interact in practice is invalu-
able in understanding how this complex system interacts to produce outcomes.
However, many challenges still exist. Better metabolic models, particularly models
that properly account for long-term metabolic variations are needed. This is partic-
ularly true in type 2 diabetes where poor glycemic control leads to the progression of
the disease. Models of the interplay between a failing glucose-insulin metabolism
and treatment options for a type 2 patient are still in their infancy. The amount of
treatment options and combinations make this a challenging combinatorial
modeling problem.
We have shown that behavior has a strong effect on the overall ability to achieve
good control. More work is necessary to advance our understanding of the patient’s
behavior. What are the patient’s goals? What is the best use of information to help
this patient achieve his/her goals? Recent reports [94] show that little progress has
been made, despite clear improvements in accuracy, insulin formulations, and