[R] Competing risks Kalbfleisch & Prentice method

[Ravi Varadhan]

Hi Eleni,

I would like to take a look at your R function for obtaining the cumulative
incidence function (CIF) from individual Cox models for cause-specific
hazards (CSH).  Does your code predict the CIF (with pointwise confidence
intervals and global confidence bands) for different sub-groups?  Have you
seen the paper by Cheng, Fine, and Wei (Biometrics 1998) that does this? 

A major advantage of the F&G model is that you can get a direct, numerical
measure of the effect of a covariate on the CIF.  This cannot be obtained by
modeling all the CSHs and then combining them.  The idiosyncratic assumption
concerning risk set in F&G model is made mainly for mathematical purposes so
that a proportional hazards form may be obtained for the CIF.  You can test
this assumption by plotting schonefeld-type residuals (this is available in
cmprsk).  Fine (Biostatistics 2006) provides a different approach that
relaxes this assumption (it also uses a different estimation approach), but
I don't know if there is an R implementation for that.

Thanks,
Ravi.
 
----------------------------------------------------------------------------
-------

Ravi Varadhan, Ph.D.

Assistant Professor, The Center on Aging and Health

Division of Geriatric Medicine and Gerontology 

Johns Hopkins University

Ph: (410) 502-2619

Fax: (410) 614-9625

Email: rvaradhan at jhmi.edu

Webpage:  http://www.jhsph.edu/agingandhealth/People/Faculty/Varadhan.html

 

----------------------------------------------------------------------------
--------


-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
Behalf Of Eleni Rapsomaniki
Sent: Thursday, March 26, 2009 10:18 AM
To: Arthur Allignol
Cc: r-help at r-project.org
Subject: Re: [R] Competing risks Kalbfleisch & Prentice method

Thank you for your reply.

It wasn't too hard to code actually, which is probably why it doesn't have a
special package dedicated to it. The results are almost identical to Fine &
Gray regression model. The problem with the latter is that my colleagues are
not convinced that the model assumptions (people who die from competing
causes remaining in the risk set) are theoretically sound. 

If anybody is interested in the Kalbfleisch & Prentice based cumulative
incidence adjusting for competing risks with covariates, I'm happy to supply
the code.

Eleni Rapsomaniki
 
Research Associate
Tel:     +44 (0) 1223 740273
Strangeways Research Laboratory
Department of Public Health and Primary Care University of Cambridge
 

-----Original Message-----
From: Arthur Allignol [mailto:arthur.allignol at fdm.uni-freiburg.de]
Sent: 26 March 2009 10:36
To: Eleni Rapsomaniki
Cc: r-help at r-project.org
Subject: Re: [R] Competing risks Kalbfleisch & Prentice method

I don't think there is a package to do that.

But you could have a look at ?predict.crr.

Best regards,
Arthur Allignol

Eleni Rapsomaniki wrote:
>  
> 
> Dear R users
> 
>  
> 
> I would like to calculate the Cumulative incidence for an event 
> adjusting for competing risks and adjusting for covariates. One way to 
> do this in R is to use the cmprsk package, function crr. This uses the 
> Fine & Gray regression model. However, a simpler and more classical 
> approach would be to implement the Kalbfleisch & Prentice method
(1980,
> p 169), where one fits cause specific cox models for the event of 
> interest and each type of competing risk, and then calculates
incidence
> based on the overall survival.  I believe that this is what the cuminc 
> function in the aforementioned package does, but it does not allow to 
> adjust for a vector of covariates.
> 
>  
> 
> My question is, is there an R package that implements the Kalbfleisch
&
> Prentice method for competing risks with covariates?
> 
>  
> 
> for example, if k1 is the cause of interest among k competing causes:
> 
> P_k1(t; x)=P(T<=t, cause=k1|x)=Sum(u=0, ..., u=t)
{hazard_k(u;x)*S(u;x)}
> 
> where S(u;x) = exp{-sum_of_k(sum(hazard_k(u))}
> 
>  
> 
> I have searched extensively for an implementation of this in many 
> packages, but it appears that more complex approaches are more
commonly
> implemented, such as timereg package. 
> 
>  
> 
> Eleni Rapsomaniki
> 
>  
> 
> Research Associate
> 
> Strangeways Research Laboratory
> 
> Department of Public Health and Primary Care
> 
>  
> 
> University of Cambridge
> 
>  
> 
>  
> 
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at 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.
> 

______________________________________________
R-help at 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.