310 Introduction to Human Nutrition
Correlation is the statistical method to use when
studying the association between two continuous
variables. The degree of association is ordinarily mea-
sured by Pearson’s correlation coeffi cient. This calcu-
lation leads to a quantity that can take any value from
−1 to +1. The correlation coeffi cient is positive if
higher values of one variable are related to higher
values of the other and it is negative when one vari-
able tends to be lower while the other tends to be
higher. The correlation coeffi cient is a measure of the
scatter of the points when the two variables are
plotted. The greater the spread of the points, the lower
the correlation coeffi cient. Correlation involves an
estimate of the symmetry between the two quantita-
tive variables and does not attempt to describe their
relationship. The nonparametric counterpart of
Pearson’s correlation coeffi cient is the Spearman rank
correlation. It is the only nonparametric method that
allows confi dence intervals to be estimated.
To describe the relationship between two continu-
ous variables, the mathematical model most often
used is the straight line. This simplest model is known
as simple linear regression analysis. Regression
analysis is commonly used not only to quantify the
association between two variables, but also to make
predictions based on the linear relationship. Nowa-
days, nutritional epidemiologists frequently use the
statistical methods of multivariable analysis (Table
13.3). These methods usually provide a more accurate
view of the relationship between dietary and non-
dietary exposures and the occurrence of a disease or
other outcome, while adjusting simultaneously for
many variables and smoothing out the irregularities
that very small subgroups can introduce into alterna-
tive adjustment procedures such as stratifi ed analysis
(Katz, 2006).
Most multivariate methods are based on the
concept of simple linear regression. An explanation
of the variation in a quantitative dependent variable
(outcome) by several independent variables (expo-
sures or predictors) is the basis of a multiple-
regression model. However, in many studies the
dependent variable or outcome is quite often dichoto-
mous (diseased/nondiseased) instead of quantitative
and can also be explained by several independent
factors (dietary and nondietary exposures). In this
case, the statistical multivariate method that must be
applied is multiple logistic regression. In follow-up
studies, the time to the occurrence of disease is also
taken into account. More weight can be given to
earlier cases than to later cases. The multivariate
method most appropriate in this setting is the pro-
portional hazards model (Cox regression) using a
time-to-event variable as the outcome (Table 13.3).13.3 In vitro studies
Scientifi c research involves studies across a reduction-
ist spectrum. As studies become more reductionist,
more and more confounding factors are stripped
away. In vitro studies represent part of the reduction-
ist approach in nutrition research. The range of tech-
niques used is large.
● Chemical analysis studies provide data on nutrient
and nonnutrient content of foods.
● Digestibility techniques, in which a substrate is
exposed to enzymes capable of digesting the sub-
strate, help to refi ne the gross chemical analytical
data to predict nutritional potential.
● Intact organs such as the liver of experimental
animals can be used in studies such as perfused
organ studies. In such studies, the investigator can
control the composition of material entering an
isolated organ and examine the output. Sections of
organs can also be used, such as the everted gut sac
technique. A small section of the intestine is turned
inside out and placed in a solution containing some
test material. Uptake of the test material into the
gut can be readily measured.
● Another approach is the construction of mechani-
cal models that mimic an organ, usually the gut (in
nutrition research). Many of these models success-
fully predict what is observed in vivo and have
advantages such as cost and fl exibility in altering
the experimental conditions with great precision.
System biology is a recently launched platform to
integrate metabolic pathways using computational
biology.
The application of molecular biology techniques to
tissue and cell culture systems has provided research-
ers with powerful strategies to evaluate and establish
metabolic pathways and regulatory roles of nutrient
and nonnutrient components of food. Thus, North-
ern, Southern and Western blotting techniques to
quantitate specifi c RNA, DNA, and proteins in tissues
in response to nutrients are common tools in the
nutrition laboratory. The infl uence of some nutrients
or nutritional conditions on ribosomal dynamics as
well as on cell hyperplasia or hypertrophy processes