phguardiol@aol.com wrote:
Sir,
Let me try to be more precise !
Patients are elderly patients with leukemia. They are included in a
trial in which they all receive a first course of chemotherapy (high
dose one, induction).
At the initiation of this induction (day1), they are randomized between
androgen or no androgen (for consolidation and maintenance treatment,
total duration 18 months). Main question is: long term outcome modified
or not by the addition of androgen maintenance therapy ?
This treatment, androgen, will be started after day20 post-induction
start (in most cases at the time of evaluation of the efficacy of the
induction). So all patients will have induction of one type, and half of
those who will respond to induction will be given androgen).
Good; then you have an easy to interpret external time-dependent
covariate. A simple way to analyze the data if the treatment is given
on roughly the same day for all patients is to start the clock over and
do a regular survival analysis from the new baseline (randomization time
of secondary treatment).
Sorry I don't have time to look at what you wrote below.
Frank
Once the patient has recovered from induction, it is checked whether he
has failed or not.
If the patient achieves some kind of response (complete or partial) then
he receives 6 courses of intermediate-dose chemotherapy PLUS OR NOT
androgen according to the randomization over ap eriod of about 1 year.
(Androgen is given once the evaluation after induction has been done and
has shown that there was some kind of response.). androgens are given
for 18 months from induction recovery.
Androgen is supposed to maintain the remission for a longer time. I m
expecting (this is my view) that during the period20of chemotherapy,
androgen will not have a major effect but this effect should be clearer
once chemotherapy is stopped (at 12 months).
Androgen, is not supposed to have major side effects which could lead to
death. Toxic deaths will come first from the induction and the other
chemotherapy courses, more than from the use of androgen.
Besides as we are dealing with patients between 60 and 80 years of age,
some will die from natural causes or from toxicity of the chemotherapy
(not from androgen therapy; we have this information).
So I d like to model the thing this way as the PH assumption is violated
by the covariate androgen around 1 year from treatment initiation: use a
time dep strategy for androgen covariate < 1 year > 1 year, and combine
this with a competing risk model ...
Does it make sense ? Other option to consider ?
Nobody has this kind of situation to face before myself !!?
Thanks again, hoping that my english and statistical knowledge is "good"
enough for this audience
Yours sincerely
Philippe Guardiola
-----E-mail d'origine-----
De : Frank E Harrell Jr <f.harrell@vanderbilt.edu>
A : phguardiol@aol.com
Cc : s-news@lists.biostat.wustl.edu
Envoyé le : Samedi, 26 Juillet 2008 16:35
Sujet : Re: [S] competing risk model with time dependent covariates< br>
phguardiol@aol.com <mailto:phguardiol@aol.com> wrote:
> > Dear S users,
> > is there a way, I mean a package, to perform a competing risk model
> which can handle time dependent covariates ?
> > my main covariate (additional treatment to patients) appears not to
> follow the proportional hazards assumption, its effect being observed
> after one year of treatment but not before (this is expected / makes
> sense on a clinical point of view). SO I was planning to use a time >
dependent Cox model with 2 time intervals for patients < > 1 year etc...
Philippe,
As your main covariate is an internal time-dependent covariate, it will
be impossible to interpret its effect without having in the model (to
condition on) the variables that indicate the reason for the treatment.
These variables need to be measured before the auxiliary treatment.
This is to get around "treatment by indication" bias.
I don't know the answer to your question, though. I hope there is some
S-Plus or R software for time-dependent covariates with competing risks
as we may have need for that also.
Frank
> > However, a large number20of patients dies from treatment related >
complications of unrelated to the disease and the treatment (elderly >
patients) so to analyze the effect of my covariate of interest on >
disease free survival and relapse incidence, I d like to use a competing
> risk model such as the Gray CMPRSK.
> > My problem is that I cannot have multiple "lines" for the same
patient > with this package (for time dep covariates 2 or 3 lines with
same Id /UPN).
> How can I manage this !!!???
> Any help will be greatly appreciated
> Thanks
> Yours sincerely
> > Philippe Guardiola
> >
-- Frank E Harrell Jr Professor and Chair School of Medicine
Department of Biostatistics Vanderbilt University
--
Frank E Harrell Jr Professor and Chair School of Medicine
Department of Biostatistics Vanderbilt University
|