Thank you very much. Welcome. It's my great pleasure
to talk to you about surrogate endpoints.
I'm Geert Molenberghs, based in Belgium in two universities,
the University of Hasselt and the University of KU Leuven,
will share the Institute for Biostatistics and Statistical Bioinformatics.
In the first lecture on surrogate endpoints,
what I would like to do is go back to the very beginning,
start with Prentice's definition,
because that's where it all start.
Then, from there, go over subsequent developments up to the time when we have
passed over what we call the single-trial
validation framework, the meta-analytic framework.
When that's done, we will have a good tool enhanced,
show us to evaluate the surrogate that is continuous,
a true endpoint that's continuous,
and that's been recorded not just in one trial but in a variety of trials.
The first question is, why do we use surrogate endpoints?
What is the main motivation?
Well, in a clinical trial,
we have a endpoint that we agree upon.
For example, the survival of the patient,
the death of the patient.
For example, the disappearance of symptoms or symptoms going below a certain threshold.
Well, it might well be that the true endpoint
especially when it's time to event like survival takes too long,
it's too distant in time in other words, and therefore,
studies would take too long which is uncomfortable for the patient and is also expensive.
Survival especially in certain clinical trials where we have censoring,
may not happen over the course of the study in all patients.
In other words, we would have censoring,
and therefore, we lose information.
So the idea is that a surrogate endpoint would be used instead of a true endpoint,
show that it solves or at least minimizes one or both of these issues.
Some additional motivations for using a surrogate endpoint are;
the true endpoint may be invasive,
it may be an invasive undertaking towards the patient,
it may be uncomfortable,
it may cost a lot of money.
Of course, if it takes a long time before we can observe a surrogate,
it may be confounded with secondary treatments,
there may be competing risks.
In other words, all sort of things may happen over the course of a study.
To summarize it in simple terms,
at the beginning of the study,
we have the benefit of randomization,
but randomization is a bit like a bottle of milk.
It's fine in the beginning, but then,
if you wait too long, it grows tale.
So we would have that very same issue of
course here that we need to take that into account.