Date:         Tue, 25 Apr 2006 13:43:15 +0100
Reply-To:     Margaret MacDougall <margmacd@talk21.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Margaret MacDougall <margmacd@talk21.com>
Subject:      Re: Kaplan Meier curves using SPSS
In-Reply-To:  <1031517093.20060422194546@terra.es>
Content-Type: text/plain; charset=iso-8859-1

Dear Marta

  Thank you very much for this info.  I took some time out just a moment ago and can see why this little trick does not change my calculations, given the definition for the cummulative survival function. I am most grateful for your useful suggestions. However, I am tempted to suggest that the SPSS curves should not require this little trick of the trade.

  Best wishes

  Margaret

Marta García-Granero <biostatistics@terra.es> wrote:
  Hi Margaret,

MM> Thank you for this kind reply. I am sure that it will prove
MM> helpful in addressing the problem I have raised. However, can you
MM> also advise on how I could apply the little trick of adding a
MM> false censored case when relying solely on menu commands whilst
MM> still performing an accurate log-rank test (which does not include
MM> the censored case)?

You can't use pull down menus for that. I use a syntax for that (a
MATRIX file creates a file with the false censored case and then I use ADD
FILES to add it to the current dataset:

* Example dataset *.
DATA LIST FREE/time status (2 F8.0).
BEGIN DATA
6 1 6 1 6 1 6 1 8 1 8 1 12 1 12 1
12 0 15 0 16 0 18 0 18 0 20 1 22 0 24 1
28 0 28 0 28 0 30 1 30 0 33 0 42 1
END DATA.
VALUE LABEL status 0'Censored (Alive)' 1'Uncensored (Dead)'.

* "Ugly" graph *.
KM time /STATUS=status(1)
/PRINT TABLE MEAN
/PLOT SURVIVAL .

* Adding false censored data *.
MATRIX.
COMPUTE time=0.00001.
* Assuming censored value is 0 *.
COMPUTE status=0.
COMPUTE data={time,status}.
COMPUTE vnames={'time','status'}.
SAVE data /OUTFILE='c:\Temp\FalseCensored.sav'/NAMES=vnames.
END MATRIX.
ADD FILES /FILE=*
/FILE='C:\temp\FalseCensored.sav'.

* Graph starts from Cum Survival=1 *.
KM time /STATUS=status(1)
/PRINT TABLE MEAN
/PLOT SURVIVAL .

It is long to explain the math reasons, but, believe me, any censored
case with a follow up time shorter than the first uncensored case is
ignored in KM analysis, it's like tied data in McNemar test: you get
the same result with or without them. Compare the output from the
first KM with the 2nd, and you'll see that the statstistics are
exactly the same, the only difference is that in the second KM, the
number of censored is 12 instead of 11.

MM> Whilst I am happy to use the SPSS commands
MM> language, it would be particularly good to learn if it is possible
MM> to address the survival plots problem by selecting appropriate
MM> options from the menus provided.

No, unless you go to Data View and add the extra data manually instead
of authomatically with syntax.

HTH

Marta



---------------------------------
To help you stay safe and secure online, we've developed the all new Yahoo! Security Centre.