Date:         Thu, 9 Nov 2006 19:27:40 -0500
Reply-To:     "data _null_;" <datanull@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "data _null_;" <datanull@GMAIL.COM>
Subject:      Re: Automatic Unblinding
Comments: To: Roland Rivers <seatedcoin@gmail.com>
In-Reply-To:  <200611092111.kA9HYuej024853@mailgw.cc.uga.edu>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Keeping in mind that I don't really know what I'm doing.

I my limited experience we don't usually do it that way.

We usually create dummy treatment codes and produce all the data
displays and summary tables using using those dummy codes, but
labeled(formatted) with the real treatment descriptions that will be
used in the final report.  That way we can make sure the columns fit
and have the proper labeling, alignment etc.   We can fully visually
the final report but the the treatments are made up.  When the blind
is broken we replace the dummy codes with the "real thing".

But to answer your question, yes it can be done but it may be
difficult for some displays.  If they are simple like the one you
suggest yes.  Here is one approach, using treatment as an ACROSS
variable in proc report.

To switch the program from treatments to blinded comment out one of
these lines below.

treatment cvals=('A' 'B' 'C')
treatment cvals=('Blinded' 'Blinded' 'Blinded')

I used proc plan to generate a bit more data but that was not necessary.


proc format;
   value disp 1='Completed' 2='Discontinued';
   value $trt(default=9) 'A' = 'Placebo' 'B'='Active 1' 'C'='Active 2';
   run;
proc plan seed=112233;
   factors
      patient = 100 ordered
      Treatment = 1 of 3 random
      Disposition = 1 of 5 random
         / noprint;
   output out=work.plan
      treatment cvals=('A' 'B' 'C')
/*      treatment cvals=('Blinded' 'Blinded' 'Blinded')*/
      disposition nvals=(1 1 1 1 2)
      ;
   format treatment $trt. disposition disp.;
   run;

proc report nowd headline list;
   columns ('--' disposition treatment);
   define disposition / group left;
   define treatment / across 'Treatment' '--';
   run;



On 11/9/06, Roland Rivers <seatedcoin@gmail.com> wrote:
> I want to be able to unblind a drug trial automatically without having to
> change the column headers by hardcoding. For example I might have the
> following blinded INPUT data:
>
> Patient #  Disposition   Treatment
> 1           Completed      Blinded
> 2           Discontinued   Blinded
> 3           Completed      Blinded
>
> Output
>            Treatment Counts
> Disposition     Blinded
> -----------     -------
> Completed          2
> Discontinued       1
>
> Once the study becomes unblinded (Treatments A and B rather than Blinded)
> I want to automatically be able to produce output without hardcoding the
> output columns.
>
> Unblinded Input:
>
> Patient #  Disposition   Treatment
> 1           Completed      A
> 2           Discontinued   B
> 3           Completed      A
>
> Dersired Output after the trial has been unblinded
>               Treatment Counts
>               ----------------
> Disposition        A    B
> ------------      ---   --
> Completed          2    0
> Discontinued       0    1
>