Dear all,
maybe anybody has an idea to solve the
following survival analysis problem.
I have a large breast cancer dataset. In
the metastacic stadium of breast cancer the oncologists tried a lot of
treatments using the "play the winner drop the loser strategy". The aim of the
study having survival time since metastases as the primary outcome should be to
find the "optimal treatment sequence" which maximizes the survival
time.
The data situation is the following:
In addition to all well known prognostic factors,
the ordering, the duration and the success of the different
treatments each patient received during metastatic stadium was
documented. Most of the patients received between three and five
treatments.
in detail:
first-line therapie: t1:
from .. to ..., th. success (PD;NC;PR;CR)
second-line
th.: t2: from .. to ..., ts2
third-line
th. t3: from .. to,
ts3
...
...
I want to find a method which uses the whole
information from the dataset and evaluates the effects of the
different treatment sequences on the individual's survival time.
Maybe an approach with Markov chains and cox-ph
with time-dependent
variables would be successful.
Any hints, suggestions, comments or
references would be greatly appreciated!..
Thanks
Bernd Genser
