FW: st: Parametric survival analysis with competing risks

["Seed, Paul" <paul.seed@kcl.ac.uk>]

Dear Statalist, 

Sally Hinchliffe contacted me offline about this issue
(and what I had said), and I replied.

I am copying the messages to Statalist for anyone else who is interested.

-----Original Message-----
From: Seed, Paul 
Sent: 01 May 2012 18:29
To: Hinchliffe, Sally R.
Subject: RE: st: Parametric survival analysis with competing risks

Dear Sally, 

Thank you for taking the trouble to reply.
I think I am OK in this instance.  
I know the gestation at delivery for every pregnancy.
(That is usual for obstetric data, particularly for studies of high-risk women).

There is no loss to follow-up or other early censoring in the data (only administrative censoring at 37 weeks gestation for all women).
If there had been, I would have treated LTFU as censored at the last date on which the pregnancy was known to be ongoing.
 
Every woman has one of 3 outcomes: 
* Pregnancy continues to 37 weeks 
	( treated as censored at 37 weeks).
* Delivery with Spontaneous onset before 37 weeks 
	(mainly spontaneous labour, but some Preterm Prelabour Rupture of Membranes[PPROM]) - treated as an event
* Delivery with Iatrogenic onset before 37 weeks 
	(either induction or elective C/Section)

My concern is to accurately estimate the probability of spontaneous onset before various dates - not in the hazard ratios, while properly allowing for the competing risk of iatrogenic onset.  

(I am carefully distinguishing onset from delivery.  
Emergency C/S, for instance is spontaneous onset, but iatrogenic delivery).

BW

Paul Seed


> -----Original Message-----
> From: Hinchliffe, Sally R. [mailto:srh20@leicester.ac.uk]
> Sent: 01 May 2012 10:00
> To: Seed, Paul; 'statalist@hsphsun2.harvard.edu'
> Cc: 'cnm100@hotmail.com'; 'enzo.coviello@tin.it'
> Subject: Re: st: Parametric survival analysis with competing risks
> 
> Dear Paul,
> 
> I think you have misunderstood my work. The slides given in the link 
> attempt to simplify some of the concepts of competing risks. When I 
> presented these I actually went into more detail.
> 
> What you are suggesting would not work unless you observe no censoring 
> in your data. So for example, if we had a cohort of patients that are 
> followed for a period of 5 years and there was no censoring within 
> that five year period other than administrative censoring. By 
> administrative censoring I mean a patient being censored if they have 
> not had an event by the end of the 5 year follow-up period.
> 
> In this scenario, if we were considering an event of interest and one 
> competing event, then patients that experience the competing event 
> will have their event time inflated to 5 years which would have been 
> their administrative censoring time.
> 
> However, in most data we actually observe some censoring. In this 
> case, patients that experience the competing event could actually have 
> had a chance of being censored before the end of the 5 year follow-up 
> period had they not experienced this event. Therefore, by just 
> inflating their time to five years we are not accounting for the 
> censoring distribution in the data.
> 
> 
> Kind Regards
> 
> Sally Hinchliffe
> 
> 
> 
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner- 
> statalist@hsphsun2.harvard.edu] On Behalf Of Seed, Paul
> Sent: 30 April 2012 17:19
> To: statalist@hsphsun2.harvard.edu
> Subject: RE: st: Parametric survival analysis with competing risks
> 
> Dear Enzo, Cameron,
> 
> Thank you for your comments, and reading list. Several the papers are 
> about estimating effects of treatments or exposures.  Unbiased 
> estimates are not my concern here.
> I am solely interested in arriving at an accurate parametric estimate 
> of the hazard function that can be tuned for different test results.
> It is important that the resulting survival function should correspond 
> to that imagined by the clinicians using it (once I have established 
> what this is); and this, it appears, I can achieve by choosing 
> appropriate censoring times for competing events.
> 
> I relied mainly on Sally R. Hinchliffe "Competing Risks - What, Why, 
> When and How?" (http://www2.le.ac.uk/departments/health-
> sciences/research/biostats/youngsurv/pdf/SHinchliffe.pdf/view)
> which draws on Fine & Gray (1999) and Putter (2007) - references given.
> The subdistibution hazard that she describes appears to have the 
> properties I want.  In particular, (slide 23):
> 
> * The difference between cause-specific and subdistribution hazards is 
> the risk set.
> 
> * For the cause-specific hazard the risk set decreases each time there 
> is a death from another cause - censoring.
> 
> * With the subdistribution hazard subjects that die from another cause 
> remain in the risk set and are given a censoring time that is larger 
> than all event times.
> 
> So that's two of the cited references which "recommend to apply the 
> ordinary survival regressions treating the subjects with competing 
> events as censored at the end of the study period."
> 
> Thanks again,
> 
> 
> Paul Seed
> 
> 
> 
> 
> Date: Sun, 29 Apr 2012 08:49:55 -0400
> From: Cameron McIntosh <cnm100@hotmail.com>
> Subject: RE: st: Parametric survival analysis with competing risks
> 
> Correct -- the results will be biased if you use a naive Kaplan-Meier 
> (censoring out the competing events as you say).
> Cox or Fine-Gray models are preferable in this case, but not perfect.
> Happy reading, Cam
> 
> > Date: Sun, 29 Apr 2012 08:40:42 +0200
> > From: enzo.coviello@tin.it
> > To: statalist@hsphsun2.harvard.edu
> > Subject: Re: st: Parametric survival analysis with competing risks
> >
> > Hi,
> >
> > thanks for the references.
> > Let me consider that none of the cited references recommend to apply 
> > the ordinary survival regressions treating the subjects with
> competing
> > events as censored at the end of the study period.
> >
> > Enzo
> >
> >
> >
> > Il 28/04/2012 22.27, Cameron McIntosh ha scritto:
> > > Enzo,
> > > I also suggest you take a look at:
> > >
> > > Andersen, P.K., Geskus, R.B., de Witte, T.,&  Putter, H. (2012).
> > > Competing risks in epidemiology: possibilities and pitfalls.
> > > International Journal of Epidemiology, Advance
> > >
> Access.http://ije.oxfordjournals.org/content/early/2012/01/08/ije.dy
> > > r213.abstracthttp://192.38.117.59/~pka/avepi11/Research_Report_11-
> 2.
> > > pdf
> > >
> > > Dignam, J.T., Zhang, Q.,&  Kocherginsky, M. (2012). The Use and
> Interpretation of Competing Risks Regression Models. Clinical Cancer 
> Research, 18, 2301-2308.
> > >
> > > Dignam, J.T.,&  Kocherginsky, M. (2008). Choice and Interpretation
> of Statistical Tests Used When Competing Risks Are Present. Journal of 
> Clinical Oncology, 26(24), 4027-4034.
> > >
> > > Fine, J. P.,&  Gray, R.J. (1999). A proportional hazards model for
> the subdistribution of a competing risk. Journal of the American 
> Statistical Association, 94(446), 496-509.
> > >
> > > Gichangi, A.,&  Vach, W. (2005). The analysis of competing risks
> data: a guided tour. Statistics in Medicine, 132(4), 1-41.
> > >
> > > Lambert, P. C., Dickman, P. W., Nelson, C. P.,&  Royston, P.
> (2010). Estimating the crude probability of death due to cancer and 
> other causes using relative survival models. Statistics in Medicine, 
> 29(7-8), 885-895.
> > >
> > > Williamson, P. R.,  Kolamunnage-Dona, R.,&  Smith, C.T. (2007). 
> > > The influence of competing-risks setting on the choice of 
> > > hypothesis test for treatment effect.  Biostatistics, 8(4),
> > > 689-
> 694.http://biostatistics.oxfordjournals.org/content/8/4/689.full
> > >
> > > Putter, H., Fiocco, M.,&  Geskus, R.B. (2007). Tutorial in
> > > biostatistics: competing risks and multi-state models. Statistics
> in
> > > Medicine, 26(11),
> > > 2389-2430.http://web.inter.nl.net/users/rgeskus/CompRisk.pdf
> > >
> > > Cam
> 
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/