Biology Reference
In-Depth Information
E
XERCISE
5-5
Table 5-2(A) demonstrates that the standard deviation of SMBG data is
significantly higher for T1DM. Give reasons why the standard deviation
cannot be effectively used as a risk measure for hyper- and
hypoglycemia separately?
2. Validation of the Low Blood Glucose Index as a Predictor of
Upcoming Severe Hypoglycemia
Recall that in 1997, the DCCT's consensus statement concluded that
only about 8% of future SH episodes could be predicted from SMBG
data. In contrast, we now present the results of a validation trial for the
LBGI that predicted about 40% of future SH episodes within the
following 6 months.
This LBGI validation trial was performed with a data set containing
about 13,000 SMBG readings from 96 adults having T1DM for at least
two years and using insulin for BG control. Participants measured
their BG three to five times per day for 1 month. Upon completion of the
data collection, the LBGI of each participant was computed. For the
next 6 months, patients recorded the date and time of all SH episodes
on diary sheets they mailed in monthly. Patients were instructed to
telephone the investigators whenever an SH episode occurred to
schedule a structured interview designed to verify that an episode of
SH had, in fact, taken place.
Regression analysis was applied to determine the significance of the
LBGI as a predictor of SH episodes.
2
This analysis showed the LBGI was
the most significant predictive variable for SH, predicting 40% of the
variance of the SH episodes in the subsequent 6 months. In addition,
among patients who reported at least one SH episode during the
12 months before the study, this rose to 43%.
The LBGI also provided a means for classifying the subjects for their risk
for SH in the subsequent 6 months: subjects with a LBGI below 2.5
experienced, on average, 0.6 SH episodes; subjects at moderate risk
(LBGI of 2.5-5) experienced 1.5 SH episodes; and subjects at high risk
(LBGI above 5) experienced 5.8 SH episodes (see Figure 5-11).
Given these results and a review of the available literature, we conclude
the LBGI is the best predictor to date of SH using SMBG
data.
Finally, an exploratory analysis confirmed the LBGI was capable of
differentiating patients at near-term risk for SH from patients at risk for
future, but not imminent, SH. We compared patients who reported SH
2. The details of this standard statistical procedure are too technical to present
here, but can be found in Draper and Smith (1998).
