It's all About the Data and MetaData

Software coding for a covariate adaptive randomization

 Email Article  

Tweet  

  

  

  

 delicious  

  

Here's a question we were asked regarding clinical trial coding that involved SAS programming.

Question? Can you code a software for covariate adaptive randomization? We are conducting a parallel double blind, placebo-controlled, 5-arm (1 control, 4 intervention) randomized study. We have a list of 225 subjects. For each subject we want the SAS program to put the subject in one of the five arms considering two covariates. The two covariates are: The Center where the patients are recruited (VA, UHS, and Center3), and the other is the Severity of Kidney Disease (Moderate, Severe). The random assignments using this method should prevent a skew of how the subjects get distributed in the arms (that is for example there should not be a large number of VA patients in one arm instead of being evenly distributed in all arms).

This can be done with covariate randomization models with normal distribution using Proc Lifetest, Proc GLM or Proc PHREG etc.

**Suppose there are 100 patients in 3 centers

  classified by sever 1 or 2.**;

data orig;

  do i=1 to 100;

    patNo=i;

      if i<33 then center=1;

      else if i>32 and i<65 then center=2;

      else center=3;

      if i<5 or (i>17 and i<22) or (i>77 and i<92) then sever=1;

      else sever=2;

      output;

  end;

  drop i;

run;

proc sort data=orig; by patNo;

**Randomly Assigning Subjects to Treatments we can use the PLAN procedureto design a completely randomized design to assign one of two  treatments to each patients.**;

**Permuted-block randomization**;

ods listing close;

proc plan seed=54321;

   factors patNo=100 cell=1;

   treatments trt=2 random;

   output out=rand;

run;

ods listing;

proc sort data=rand(drop=cell); by patNo;

data randomized; merge orig(in=a) rand(in=b); by patNo;

  if a;

run;

You can find good articles and further explanations at the SAS support website when typing 'randomization' in the search field. But basically PROC PLAN and PROC FATEX SAS procedures can do it all.

http://www.clinplus.com/report-landing-clinplus-report-tables-listings?utm_campaign=blog

Hope this was helpful! DZS Software Solutions

DZS/ClinPlus can also give you a hand producing the Statistical Analysis Tables of the trial results in the course of the trial. (Safety and Efficacy reports etc.).