Procedure for ordered scale analysis in the crossover design (GLIMMIX?)

[AndreyMyslivets]

Hi to All!

Could you please help me with the choosing of a SAS procedure?

The study protocol states that: "The parameter 1 is ordered categorical variables. For these endpoints, the investigational product will be compared with the reference product using a mixed ordered logistic regression model. The model will include sequence, period and treatment as fixed effects and a repeated effect for patients within sequences, using a compound symmetry covariance structure. A 5% level of significance will be used to test for differences between the two treatments."
 
My endpoint is a scale and has values from 0 to 4.
This study is a crossover study, with 2 periods and 2 drugs.
 
After some literature review and seach I've realized that the best choice would be proc glimmix in this case with the following syntax:

proc glimmix data=t1 method=laplace;
   class SUBJID sequence period drug timepoint;
   model AVALC = sequence period drug/ CL dist=multinomial link=clogit solution ODDSRATIO;
   random timepoint/ subject=SUBJID(sequence) residual type=cs;
   estimate 'Trt vs. Control' drug 1 -1;
run;

What do you think?

[SteveDenham]

METHOD=LAPLACE and the residual option in the RANDOM statement are incompatible, as are the multinomial distribution and the residual option.  You will have to use something like;

proc glimmix data=t1 method=laplace;
   class SUBJID sequence period drug timepoint;
   model AVALC = sequence period drug/ CL dist=multinomial link=clogit solution ODDSRATIO;
   random timepoint/ subject=SUBJID(sequence) type=cs;
   estimate 'Trt vs. Control' drug 1 -1;
run;

Then comes the hard part - this looks like a clinical endpoint bioequivalence protocol, so that you may need to create a confidence interval on the difference as well as a hypothesis test of no difference at 5%.  You can do Schuirman's two one-sided tests approach with two ESTIMATE statements, with a lower and upper option on each statement..

This can be done by adding the following:

 estimate 'Trt vs. Control' drug 1 -1/lower alpha=0.1;
 estimate 'Trt vs. Control' drug 1 -1;/upper alpha=0.1;

SteveDenham