time dependent cox regression

[Deniz]

Hi all,

I have the following problem:
I want to look at effects of a certain treatment on survival of patients. I have data from a follow-up study of patients with a certain neurological disease. They all entered the study at some point during their disease. At a certain point about two third of the patients started receiving treatment.

I wasn't sure how to go about this problem, so I was thinking about putting the treatment time as a time dependent covariate in the cox regression. Therefore I have created an extra variable Xt which stands for the time until treatment for those who received treatment at a point during follow up and for those who never were treated I have set Xt=10 because the range of total fu time has a maximum of 10 years.

I created this syntax:

TIME PROGRAM.
COMPUTE T_COV_ = T_ > Xt.
COXREG   Follow_uptime
  /STATUS=death(1)
  /CONTRAST (T_COV_)=Indicator(1)
  /METHOD=ENTER T_COV_
  /CRITERIA=PIN(.05) POUT(.10) ITERATE(20).

I would like to know if you think this is the right way to analyze this data for this specific question? Also, there is another issue: Is it possible to put more than 1 time dependent variable in the model? Eg the disease severity may also change over time and this might muddy the treatment effect.

Thanks,

Deniz

[Hector Maletta]

The value of the "time until starting treatment" variable is not a value
that changes over time. It is fixed for every patient. If John started
treatment 3 months after beginning to be observed, he would have this fixed
3-month interval for all subsequent times of observation during the study.
It varies from patient to patient, but it does not varies over time for each
individual patient. Therefore it is not a time dependent covariate. A time
dependent covariate is some attribute of patients that changes as times goes
by, i.e. as the study progresses from start to finish.

Hector

-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de Deniz
Enviado el: Friday, September 23, 2011 05:34
Para: [hidden email]
Asunto: time dependent cox regression

Hi all,

I have the following problem:
I want to look at effects of a certain treatment on survival of patients. I
have data from a follow-up study of patients with a certain neurological
disease. They all entered the study at some point during their disease. At a
certain point about two third of the patients started receiving treatment.

I wasn't sure how to go about this problem, so I was thinking about putting
the treatment time as a time dependent covariate in the cox regression.
Therefore I have created an extra variable Xt which stands for the time
until treatment for those who received treatment at a point during follow up
and for those who never were treated I have set Xt=10 because the range of
total fu time has a maximum of 10 years.

I created this syntax:

TIME PROGRAM.
COMPUTE T_COV_ = T_ > Xt.
COXREG   Follow_uptime
  /STATUS=death(1)
  /CONTRAST (T_COV_)=Indicator(1)
  /METHOD=ENTER T_COV_
  /CRITERIA=PIN(.05) POUT(.10) ITERATE(20).

I would like to know if you think this is the right way to analyze this data
for this specific question? Also, there is another issue: Is it possible to
put more than 1 time dependent variable in the model? Eg the disease
severity may also change over time and this might muddy the treatment
effect.

Thanks,

Deniz

--
View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/time-dependent-cox-regression-
tp4832813p4832813.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 10.0.1410 / Virus Database: 1520/3914 - Release Date: 09/23/11

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Deniz]

Thanks Hector,

But do you also have suggestions for a valid analyis? Is it for example valid to do a regular cox regression with entering all of the treated patients into the analysis at the time treatment started (i.e., ignore all of their pre-treatment time on study)?

Deniz

[Hector Maletta]

Of course you could do that, but I do not know the details of the study.
Perhaps things observed before treatment started are important as regards
the outcome, and also the length of time before starting treatment may be
important. Suppose, for instance, that during the pre-treatment period there
might be some events of interest (symptoms, whatever): you may want to
include those events in the analysis (as events occurred before treatment)
just as you treat things occurring after starting the treatment. It all
depends on the specifics of your study.
And by the way, if the study did not start with the start of treatment, when
did it begin? With diagnosis? There should have been something causing every
subject to start being observed.

Hector

-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de Deniz
Enviado el: Friday, September 23, 2011 15:14
Para: [hidden email]
Asunto: Re: time dependent cox regression

Thanks Hector,

But do you also have suggestions for a valid analyis? Is it for example
valid to do a regular cox regression with entering all of the treated
patients into the analysis at the time treatment started (i.e., ignore all
of their pre-treatment time on study)?

Deniz

--
View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/time-dependent-cox-regression-
tp4832813p4834636.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 10.0.1410 / Virus Database: 1520/3914 - Release Date: 09/23/11

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD