Biology Reference
In-Depth Information
scientific impact of the study and the safety of the participants. Once the
protocols are approved, government guidelines strictly prohibit any
deviation from the outlined procedures. In particular, unless the need
for information-sharing between researchers and participants is
documented as essential for achieving the main objective of the study,
and made explicit in the protocol, no results of any experimental
measurements can be disclosed to the participants.
4. Validation of the High Blood Glucose Index
High BG levels not lasting very long are not nearly as dangerous as
similarly low BG levels that may trigger SH. The risk for hyperglycemia
only becomes significant when chronically high BG levels are
maintained over time. Therefore, because we hypothesized that HBGI
accounts for the trends towards hyperglycemia observed in patients'
SMBG records, one possible validation of this measure is to evaluate its
correlation with HbA
1c
—the standard for assessing high average BG
levels.
We analyzed SMBG and HbA
1c
data provided by Amylin
Pharmaceuticals (San Diego, CA), from 600 subjects with type 1
(N
323) diabetes. The subjects collected more
than 300,000 SMBG readings and had 4180 HbA
1c
assays taken over six
months. The overall correlation between the HBGI and HbA
1c
was 0.73,
p<0.0001, demonstrating a strong linear relationship between
these two variables. Further, we identified five categories for the HBGI
(below 7, 7-12, 12-15, 15-20, above 20) and computed 95% confidence
intervals for HbA
1c
corresponding to these categories, establishing
almost one-to-one correspondence between HBGI and HbA
1c
. (The
statistical details can be found in Kovatchev et al. [2000].)
¼
277) and type 2 (N
¼
In summary, we are now in a position to conclude that LBGI and HBGI
are valuable quantitative characteristics that could be used for
assessment and maintenance of glycemic control. Because these
parameters are directly quantifiable from routine SMBG data, they can
provide accurate information for analysis and assessments of the effects
of changes in therapeutic regimens. In addition, the LBGI and HBGI
could be used as building blocks for more sophisticated mathematical
models.
We present one such model next.
X. MORE COMPLEX MODELS
The 40% success rate for predicting SH from SMBG data using the LBGI
provided a substantial improvement over the 8% rate achieved by the
DCCT. In this section, we build a more complex model that utilizes the
LBGI, some basic probability laws, and curve-fitting to achieve even
greater success in predicting future SH episodes.
