Biology Reference
In-Depth Information
Analyzing the Data
In statistical regression, it is suggested to not have as many variables in the calculations as
there are individuals in the study. To reduce the number of variables, a statistical variable
selection method called McHenry's algorithm (1978) was used independently for the
cross-sectional CT variables and DEXA density variables in order to choose the best variables
to develop the multiple regression equations. The best variables for each sex were then sub-
jected to a second variable selection. The best two or three variables were used to create
robust multiple regression equations in NCSS (Number Crunching Statistical Software,
1997). To avoid problems with multicollinearity, only the strongest density variable was
included to develop each multiple regression equation. Multicollinearity is the statistical
problem of two variables reflecting the exact same effect, thus overemphasizing its influence
on the end result.
The results of this study demonstrate that bone mineral density has a strong correlation
with body mass in the proximal human femur for both European American males and
females. Furthermore, there are significant differences in bone mineral density between
different weight classifications in both males and females. This correlation is not as strong
in males as it is for females between the average weight and obese individuals. From the
CT scans, the shape variables that were important for body mass in both males and females
were the cross-sectional area at most proximal cross-section, polar moments of area (i.e.,
torsional strength) at several locations, and the area moments of inertia in various directions.
The shape indices (I
max
/I
min
and I
y
/I
x
) surprisingly do not show any clear relationship with
body mass, but may still in fact reflect activity. Further, there does not appear to be a clear
correlation between the canal area and body mass, as was predicted by the fact that endosteal
apposition occurred at a higher rate in obese individuals in other studies (
Ruff, et al., 1994;
Pearson and Lieberman, 2004; Moore et al., 2007
).
One advantage of this study over previous studies was that by using skeletal femora with
a uniform depth of soft-tissue equivalent material, any inconsistencies that arose from
different thicknesses of living tissue were removed. Another advantage was the ability to
directly compare the results from DEXAwith the cross-sectional geometry of the same bones
gathered from CT scanned data.
5
In a living population, this would have required exposing
patients to excessive amounts of radiation. By using a skeletal sample, the resolution of the
images could be increased to provide more accurate models. This allowed the addition of
a cross-sectional analysis to better predict body mass, which improved upon previous
research and increased the power of the model. Both CT and DEXA use X-rays to develop
an image of the internal and external structure of bone for better interpretation of the material
properties (bone density) and macrostructure to provide a more holistic view of the func-
tional adaptation that results from variations in body mass.
There were many logistical obstacles to overcome during this multidisciplinary disserta-
tion project. Through collaboration with multiple departments, valuable resources were
shared to that mutual benefit of all parties concerned. The CT scans that I helped produce
5
The CT scans are now part of the UT collection, and can be made available for many potential research
projects. Contact the Forensic Anthropology Center at UT for information:
www.utk.edu
.
