I have what I *think* should be a simple problem in R, and hope
someone might be able to help me.
I'm working with cancer survival data, and would like to calculate
adjusted survival figures based on the age of the patient and the
tumour classification. A friendly statistician told me I should use
Cox proportional hazards to do this, and I've made some progress with
using the coxph function. However, there doesn't seem to be a simple
way to get adjusted survival figures, and I'm still a bit unsure
whether I'm using the function correctly.
I've found plenty of examples of how to plot survival curves for a
single catagorical variable, like tumour stage. However, its a bit
less obvious how to properly combine multiple variables (like stage
and grade) - examples I've found aren't very helpful. For instance,
can someone explain what these two function calls are doing differently?
cox.model1 <- coxph(Surv(time,outcome)~stage+grade,method="breslow")
cox.model2 <- coxph(Surv(time,outcome)~stage+grade
+stage:grade,method="breslow")
Also, am I right to think that this is how I would plot the survival
curves for stage in a 50 year old patient?
cox.model3 <- coxph(Surv(time,outcome)~stage+I
(age-50),method="breslow")
plot(survfit(cox.model3,newdata=data.frame(stage2=c
("III","IV"),age=c(50,50))))
Finally, assuming I manage to correctly call the coxph function in
the first place (!), what is the sensible way to adjust the survival
for each patient? My understanding of what I'm trying/hoping to do
has led me down the following dark alley;
1) pick a sensible centre point (median age, modal stage/grade) as
my 'standard' case
2) obtain the survival probabilities for this 'standard' case from
the Cox model
3) for each patient find the probability of their observed
survival, for their specific age/stage/grade
4) the adjusted survival is the survival time from the 'standard'
case that matches the observed survival probability
This seems long-winded, and for all I know might be completely the
wrong way to do it..... hopefully someone will be able to offer some
friendly advice!
Apologies for the long-ish post, and thanks in advance for any help.
Chris Jones.
-----------------------
Gynaecological Cancer Research Laboratories,
UCL EGA Institute for Women's Health,
University College London,
Paul O'Gorman Building,
72 Huntley Street,
London
WC1E 6DD
United Kingdom
Telephone; 020 3108 2007
Fax; 020 3108 2010
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
