Biology Reference
In-Depth Information
V. SYMMETRIZATION OF THE BLOOD GLUCOSE
MEASUREMENT SCALE
BG fluctuations are often the object of statistical description and
various data analyses in research and clinical practice. Most statistical
techniques, however, require assumptions about the shape of the
underlying distribution of the data being analyzed. For example, the
routine statistical practice of reporting in the format ''mean value
standard deviation'' assumes a symmetric distribution of the data
readings. This is not the case with BG readings. For example, Figure 5-4
presents a typical BG data distribution for a subject with T1DM (186
readings downloaded from his memory meter). The distribution is
substantially skewed, and the superimposed bell curve (normal density)
describes the data poorly.
This problem is not new, and it arises often in statistics. There are well-
developed techniques that provide transformations converting
nonsymmetric samples to approximately symmetric ones (Box and Cox
[1964]). The statistical analyses are performed with the symmetric data,
and then an inverse transformation is used to translate the results so that
they correspond to the original sample. It is important to be aware that
such transformations are sample-dependent (i.e., different samples will be
symmetrized by different transformations). Therefore, this approach will
be impractical for implementing in a SMBG device, because the
transformation should be known in advance.
An alternative approach that eliminates sample dependency is to change
the scale of the BG readings so that in the new scale the BG sample is
symmetric. In Figure 5-4, notice the following:
30
Hyperglycemia
Hypo-
Target Range
25
20
15
10
5
0
0
1
2
3
4
5
6
7
8
9
10
11
12 13
14
15 16 17 18 19
20
BG (mmol/L)
FIGURE 5-4.
The distribution of BG levels. The bar graph represents the BG readings of a person with T1DM.
A normal distribution (the line graph) does not fit the data well (
#
1997 American Diabetes
Association. From Diabetes Care, 20, 1655-1658. Reprinted with permission from The American
Diabetes Association.)
