Randomization Schema: How do I assign subjects to blocks with a ratio of treatment and control?

michaeljf16 — Mon, 20 Feb 2017 18:38:49 GMT

I need to write code to create a reproducible randomization scheme for a three site clinical trial. At each site, the randomization is 1:1 treatment:control but in blocks of 6. 24 subjects are assigned to each site. In other words, there are a total of 72 subjects, 24 subjects randomly assigned to each site, 4 blocks with a size of 6 at each site, with 3 treatment subjects and 3 control subjects within each block. The code has to be reproducible in a sense that you can change the number of sites, number of subjects per site, number of blocks, size of blocks and the ratio of control to treatment.

I am fine with randomly assigning subjects to different sites. Where I struggle is randomly assigning subjects within each site to blocks with a ratio of treatment to control. For some reason, when I run the code to assign subjects to blocks with treatment and control, each subject within each site produces 24 observations (4 blocks of 6). So a total of 576 observations in each site. What I am trying to accomplish is randomly assigning each subject within a site to one of 4 blocks of 6 with 3 treatment and 3 control subjects in each block (or whatever the specified block size or treatment/control ratio is).

Here is what I have so far:

data values;

nsites=3;

subjectspersite=24;

/* to calculate the total sample size based on the number of sites and number of subjects assigned to each site*/

samplesize=(nsites*subjectspersite);

blocksize=6;

TreatmentR=1;

nblocks=round((subjectspersite)/(blocksize));

na=round(blocksize/(1/(treatmentR/(1+treatmentR))));

nb=blocksize-na;

run;

data values1;

set values;

/*to assign each subject*/

DO Subject = 1 to (samplesize) by 1;

OUTPUT;

END;

RUN;

DATA values2;

set values1;

/*to randomly assign a number 1 to 0 to each subject*/

random=ranuni(27407349);

RUN;

/*to sort each subject by their randomly assigned number created in values2*/

PROC SORT data=values2;

by random;

RUN;

PROC FORMAT;

/*to format values to different sites*/

value siteno 1 = 'Site 1'

2 = 'Site 2'

3 = 'Site 3'

4 = 'Site 4'

5 = 'Site 5'

6 = 'Site 6'

7 = 'Site 7'

8 = 'Site 8'

9 = 'Site 9'

10 = 'Site 10';

RUN;

data values3;

set values2;

/*if else statements to assign subjects to their sites, based on the number of sites and total sample size*/

/*for 1 site*/

if nsites=1 then site=1;

/*for 2 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=2 then site=1;

else if _ gt ((1/nsites)*samplesize) and nsites=2 then site=2;

/*for 3 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=3 then site=1;

else if _N_ gt ((1/nsites)*samplesize) and _N_ le ((2/nsites)*samplesize) then site=2;

else if _N_ gt ((2/nsites)*samplesize) then site=3;

/*for 4 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=4 then site=1;

else if _N_ gt ((1/nsites)*samplesize) and _N_ le ((2/nsites)*samplesize) then site=2;

else if _N_ gt ((2/nsites)*samplesize) and _N_ le ((3/nsites)*samplesize) then site=3;

else if _N_ gt((3/nsites)*samplesize) then site=4;

/*for 5 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=5 then site=1;

else if _N_ gt ((1/nsites)*samplesize) and _N_ le ((2/nsites)*samplesize) then site=2;

else if _N_ gt ((2/nsites)*samplesize) and _N_ le ((3/nsites)*samplesize) then site=3;

else if _N_ gt((3/nsites)*samplesize) and _N_ le ((4/nsites)*samplesize) then site=4;

else if _N_ gt((4/nsites)*samplesize) then site=5;

/*for 6 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=6 then site=1;

else if _N_ gt ((1/nsites)*samplesize) and _N_ le ((2/nsites)*samplesize) then site=2;

else if _N_ gt ((2/nsites)*samplesize) and _N_ le ((3/nsites)*samplesize) then site=3;

else if _N_ gt((3/nsites)*samplesize) and _N_ le ((4/nsites)*samplesize) then site=4;

else if _N_ gt((4/nsites)* samplesize) and _N_ le ((5/nsites)*samplesize) then site=5;

else if _N_ gt((5/nsites)*samplesize) then site=6;

/*for 7 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=7 then site=1;

else if _N_ gt ((1/nsites)*samplesize) and _N_ le ((2/nsites)*samplesize) then site=2;

else if _N_ gt ((2/nsites)*samplesize) and _N_ le ((3/nsites)*samplesize) then site=3;

else if _N_ gt((3/nsites)*samplesize) and _N_ le ((4/nsites)*samplesize) then site=4;

else if _N_ gt((4/nsites)* samplesize) and _N_ le ((5/nsites)*samplesize) then site=5;

else if _N_ gt((5/nsites)*samplesize) and _N_ le ((6/nsites)*samplesize) then site=6;

else if _N_ gt((6/nsites)*samplesize)then site=8;

/*for 8 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=8 then site=1;

else if _N_ gt ((1/nsites)*samplesize) and _N_ le ((2/nsites)*samplesize) then site=2;

else if _N_ gt ((2/nsites)*samplesize) and _N_ le ((3/nsites)*samplesize) then site=3;

else if _N_ gt((3/nsites)*samplesize) and _N_ le ((4/nsites)*samplesize) then site=4;

else if _N_ gt((4/nsites)* samplesize) and _N_ le ((5/nsites)*samplesize) then site=5;

else if _N_ gt((5/nsites)*samplesize) and _N_ le ((6/nsites)*samplesize) then site=6;

else if _N_ gt((6/nsites)*samplesize) and _N_ le ((7/nsites)*samplesize) then site=7;

else if _N_ gt((7/nsites)*samplesize) then site=8;

/*for 9 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=9 then site=1;

else if _N_ gt ((1/nsites)*samplesize) and _N_ le ((2/nsites)*samplesize) then site=2;

else if _N_ gt ((2/nsites)*samplesize) and _N_ le ((3/nsites)*samplesize) then site=3;

else if _N_ gt((3/nsites)*samplesize) and _N_ le ((4/nsites)*samplesize) then site=4;

else if _N_ gt((4/nsites)* samplesize) and _N_ le ((5/nsites)*samplesize) then site=5;

else if _N_ gt((5/nsites)*samplesize) and _N_ le ((6/nsites)*samplesize) then site=6;

else if _N_ gt((6/nsites)*samplesize) and _N_ le ((7/nsites)*samplesize) then site=7;

else if _N_ gt((7/nsites)*samplesize) and _N_ le ((8/nsites)*samplesize) then site=8;

else if _N_ gt((8/nsites)*samplesize) then site=9;

/*for 10 sites*/

if _N_ le ((1/nsites)*samplesize) and nsites=10 then site=1;

else if _N_ gt ((1/nsites)*samplesize) and _N_ le ((2/nsites)*samplesize) then site=2;

else if _N_ gt ((2/nsites)*samplesize) and _N_ le ((3/nsites)*samplesize) then site=3;

else if _N_ gt((3/nsites)*samplesize) and _N_ le ((4/nsites)*samplesize) then site=4;

else if _N_ gt((4/nsites)* samplesize) and _N_ le ((5/nsites)*samplesize) then site=5;

else if _N_ gt((5/nsites)*samplesize) and _N_ le ((6/nsites)*samplesize) then site=6;

else if _N_ gt((6/nsites)*samplesize) and _N_ le ((7/nsites)*samplesize) then site=7;

else if _N_ gt((7/nsites)*samplesize) and _N_ le ((8/nsites)*samplesize) then site=8;

else if _N_ gt((8/nsites)*samplesize) and _N_ le ((9/nsites)*samplesize) then site=9;

else if _N_ gt((9/nsites)*samplesize) then site=10;

format site siteno.;

run;

PROC SORT data=values3;

by site;

RUN;

data sites1 sites2 sites3 sites4 sites5 sites6 sites7 sites8 sites9 sites10;

set values3;

if nsites=1 then output sites1;

else if nsites=2 then output sites2;

else if nsites=3 then output sites3;

else if nsites=4 then output sites4;

else if nsites=5 then output sites5;

else if nsites=6 then output sites6;

else if nsites=7 then output sites7;

else if nsites=8 then output sites8;

else if nsites=9 then output sites9;

else if nsites=10 then output sites10;

run;

Data site1of1;

set sites1;

if site=1 then output site1of1;

run;

data site1of2 site2of2;

set sites2;

if site=1 then output site1of2;

else if site=2 then output site2of2;

run;

data site1of3 site2of3 site3of3;

set sites3;

if site=1 then output site1of3;

else if site=2 then output site2of3;

else if site=3 then output site3of3;

run;

data site1of4 site2of4 site3of4 site4of4;

set sites4;

if site=1 then output site1of4;

else if site=2 then output site2of4;

else if site=3 then output site3of4;

else if site=4 then output site4of4;

run;

data site1of5 site2of5 site3of5 site4of5 site5of5;

set sites5;

if site=1 then output site1of5;

else if site=2 then output site2of5;

else if site=3 then output site3of5;

else if site=4 then output site4of5;

else if site=5 then output site5of5;

run;

data site1of6 site2of6 site3of6 site4of6 site5of6 site6of6;

set sites6;

if site=1 then output site1of6;

else if site=2 then output site2of6;

else if site=3 then output site3of6;

else if site=4 then output site4of6;

else if site=5 then output site5of6;

else if site=6 then output site6of6;

run;

data site1of7 site2of7 site3of7 site4of7 site5of7 site6of7 site7of7;

set sites7;

if site=1 then output site1of7;

else if site=2 then output site2of7;

else if site=3 then output site3of7;

else if site=4 then output site4of7;

else if site=5 then output site5of7;

else if site=6 then output site6of7;

else if site=7 then output site7of7;

run;

data site1of8 site2of8 site3of8 site4of8 site5of8 site6of8 site7of8 site8of8;

set sites8;

if site=1 then output site1of8;

else if site=2 then output site2of8;

else if site=3 then output site3of8;

else if site=4 then output site4of8;

else if site=5 then output site5of8;

else if site=6 then output site6of8;

else if site=7 then output site7of8;

else if site=8 then output site8of8;

run;

data site1of9 site2of9 site3of9 site4of9 site5of9 site6of9 site7of9 site8of9 site9of9;

set sites9;

if site=1 then output site1of9;

else if site=2 then output site2of9;

else if site=3 then output site3of9;

else if site=4 then output site4of9;

else if site=5 then output site5of9;

else if site=6 then output site6of9;

else if site=7 then output site7of9;

else if site=8 then output site8of9;

else if site=9 then output site9of9;

run;

data site1of10 site2of10 site3of10 site4of10 site5of10 site6of10 site7of10 site8of10 site9of10 site10of10;

set sites10;

if site=1 then output site1of10;

else if site=2 then output site2of10;

else if site=3 then output site3of10;

else if site=4 then output site4of10;

else if site=5 then output site5of10;

else if site=6 then output site6of10;

else if site=7 then output site7of10;

else if site=8 then output site8of10;

else if site=9 then output site9of10;

else if site=10 then output site10of10;

run;

proc print data=site1of3;

run;

data treatmentsite1of3;

set site1of3;

do block=1 to nblocks by 1;

nna=0;

nnb=0;

do i=1 to blocksize;

subject2=i+((block-1)*blocksize);

if nna=na then treatment="B";

if nnb=nb then treatment="A";

else do;

aprob=(na-nna)/(na+nb-nna-nnb);

u=ranuni(0);

if (0<=u<=aprob) then do;

treatment="A";

nna=nna+1;

end;

if (aprob<u<=1) then do;

treatment="B";

nnb=nnb+1;

end;

keep subject2 treatment site subject site nblocks block;

output;

end;

run;

proc print data=treatmentsite1of3;

id subject2;

var treatment site subject block;

title1 'Site 1 of 3 sites';

title2 "Randomization Schema for Specified Ratio of Treatments A to Control B";

run;

Any help would be greatly appreciated. Also, Any correction or suggestions with my current code would be extremely helpful.

Thank you.

Re: Randomization Schema: How do I assign subjects to blocks with a ratio of treatment and control?

art297 — Mon, 20 Feb 2017 19:12:08 GMT

Would proc plan do what you want? e.g.:

data Unrandomized;
  do site=1 to 3;
   do block=1 to 4;
     do subject=1 to 6;
       id+1;
       if (subject <= 3) then treatment=1;
       else                   treatment=2;
       output;
     end;
   end;
 end;
run;

proc plan seed=27371;
   factors site=3 block=6 treatment=2;
   output data=Unrandomized out=Randomized;
run;

proc sort data=Randomized;
   by id;
run;

Art, CEO, AnalystFinder.com

Re: Randomization Schema: How do I assign subjects to blocks with a ratio of treatment and control?

michaeljf16 — Mon, 20 Feb 2017 21:57:11 GMT

This has been a huge help. Thank you!

topic Re: Randomization Schema: How do I assign subjects to blocks with a ratio of treatment and control? in SAS Programming

Randomization Schema: How do I assign subjects to blocks with a ratio of treatment and control?

Re: Randomization Schema: How do I assign subjects to blocks with a ratio of treatment and control?

Re: Randomization Schema: How do I assign subjects to blocks with a ratio of treatment and control?