Summary:
The purpose of the present thesis is an overview and application of methods of
Me-
diation Analysis. Mediation analysis has gained ground over the past few years,
due to
its ability to explain the true mechanisms that underlie an observed
relationship between
an exposure variable and an outcome variable. In particular, it examines how
the afore-
mentioned variables relate to one (or more), third, intermediate variable(s),
the so called
mediators. The thesis is divided into two parts.
In the rst part, we drew attention on certain methods, that can be found in the
current literature, in order to assess the e ect of one or more mediators on a
model
of interest. Speci cally, the rst section of the methods' overview was an
attempt to
present single and multiple mediator models, through traditional techniques, so
as to
estimate direct and indirect e ects and to quantify the uncertainty of these
measures,
by calculating standard errors and con dence intervals. We also reviewed
moderated
mediation, i.e. the case in which mediation and moderation are simultaneously
present
in a model. We then reviewed methods for mediation analysis within the
counterfactual
framework. The counterfactual framework, which is based on the causal inference
makes
it possible to examine causal relations, under certain assumptions. Moreover,
within this
framework it seems feasible to overcome many limitations of the traditional
approaches.
In the second part we focused on the application of these techniques on data
from a
cohort study and in particular the EPIC cohort study. Many studies have been
conducted
over the years associating healthy dietary habits with reduced incidence of
diseases such as
cancer, cardiovascular disease, diabetes etc. A great deal of attention has
been focused on
the Mediterranean diet, since it has been shown to be associated with increased
longevity.
Although certain hypotheses have been made regarding possible mechanisms
underlying
the indicated associations of Mediterranean Diet with various health outcomes,
conclusive
evidence regarding these mechanisms is still lacking. The main goal of our
analysis was the
investigation of the association between the Mediterranean diet and all cause
mortality
and how this relationship can be mediated by certain factors, such as type II
diabetes
and high blood cholesterol. After applying the aforementioned techniques we
found that
adherence to the Mediterranean diet was related with reduced odds of developing
diabetes,
which in turn was related with reduced hazard of all cause mortality. Contrary
to that,
adherence to the Mediterranean diet was related with increased odds of
developing high
blood cholesterol, which in turn was related with reduced hazard of all cause
mortality.
Keywords:
Mediation analysis, Overview, Direct effect, Indirect effect, EPIC
