[R] Re : left truncated data survival analysis package

phguardiol at aol.com phguardiol at aol.com
Mon Mar 2 08:55:27 CET 2009


This kind of analysis is new to me, s oI m sorry if I m asking "stupid" 
questions below...

In my study, patients might have been sick for years before the 
diagnosis is made, then they can die few months later, once diagnosed, 
either from their disease or because they are elderly patients; 
however, they can also survive for decades.
I d like to know the impact of a "newcovariate" on their survival, and 
in a second step, adjust for know prognostic factors to check whether 
this new covariate is an independent prognostic factor as well.

So according to my readings it seems to me that using left truncated 
right censored models is the best way to test my hypothesis ... I hope 
I m right there !

after reading the help file of the survival package am I right to say 
that I have to use the coxph function to do this kind of analysis ?

using the following formulation: coxph(Surv(time, time2, event, type = 
interval, origin = 0))~ newcovariate,....

in this case if a patient is 54 years at the time of diagnosis, and 
dies at 55, is it right to code this one
time= 54, time2 = 55, event=1

then if I have patient diagnosed at 66 and followed up to 68 without 
any relevant event (censored) I m also right to code this one:
time= 66, time2= 68, event=3

I dont fully understand the option of using "interval2" for "type" 
option, could you comment a little more=2
0on this please...?

last question: what is the best way to check that the left truncation 
data and the failure times are independent ? just a scatter plot ? 
regression ?

thanks for your help

yours sincerely

Philippe Guardiola

-----E-mail d'origine-----
De : Thomas Lumley <tlumley at u.washington.edu>
A : phguardiol at aol.com
Cc : R-help at stat.math.ethz.ch
Envoyé le : Jeudi, 26 Février 2009 8:24
Sujet : Re: [R] left truncated data survival analysis package

On Thu, 26 Feb 2009 phguardiol at aol.com wrote: 


> Hello, 

> I d like to run a survival analysis with "left truncated data". Could
> you recommend me a package to do this please ? 

The 'survival' package. 


Thomas Lumley           Assoc. Professor, Biostatistics 

tlumley at u.washington.edu    University of Washington, Seattle 

Reçevez AOL Mail sur votre téléphone. Vos e-mails accessibles à tout 
moment! Créez un e-mail gratuit aujourd’hui.

More information about the R-help mailing list