Unit:
Postgraduate Programme Biostatistics & Health Science DataLibrary of the School of Health Sciences
Supervisors info:
Δημήτριος Καρλής, Καθηγητής, Στατιστική ΟΠΑ
Aπόστολος Μπουρνέτας, Καθηγητής, Μαθηματικό ΕΚΠΑ
Φώτης Σιάννης, Επ.Καθηγητής, Μαθηματικό ΕΚΠΑ
Original Title:
Validation of surrogacy for survival outcomes
Translated title:
Validation of surrogacy for survival outcomes
Summary:
Background: The potential use of surrogate markers has become a trend in
clinical research. A surrogate is intended to replace a clinical endpoint, so that it
overcomes several issues during the assessment of new treatments. The first objective
of the thesis is to review the available methods for surrogacy evaluation existing
in the literature. Thereafter, an application using simulated data is presented. We
aimed to investigate whether disease free survival (DFS) is an acceptable statistical
surrogate for overall survival (OS), in early breast cancer. Information theory was
applied in the case of failure endpoints. A two-stage model was fitted, and the hospital
center was the unit of the analysis. At the trial level, we focus on the association
between the treatment effects on both endpoints, using independent proportional
hazard Cox models. Weighted (by hospital center size) least squares regression of
Cox model effects was performed. The individual level association between the endpoints
was measured through time dependent Cox models.
Results: At the trial level, the squared correlation between treatment effects on
DFS and OS (R2h
, t) was 0.547 (s.e. =0.3). At the individual level, R2
ind,QF found
to be equal to 0.95 (0.948-0.969). The significant covariates for both endpoints were
included in the equations.
Conclusions: Information theory provides estimates with less computational issues
and high interpretability. DFS strongly correlates with OS at the individual
level. However, at the trial level, a moderate relationship between the endpoints was
observed, providing too wide confidence limits to be informative.
Main subject category:
Health Sciences
Keywords:
Surrogate endpoints, Surrogacy
File:
File access is restricted only to the intranet of UoA.
PETROS_LEKKAS master.pdf
1 MB
File access is restricted only to the intranet of UoA.
