Bioequivalence and Bioavailability Forum

Hi biostat experts,

Is there a significant difference between using PROC GLM or PROC MIXED for 3-way (2T vs R), 4-way (3T vs R) and 5-way crossover studies (4T vs R) when 1) all subjects completed the study, 2) not all subjects completed all treatment arms in a study.

Would there be an issue if I elect to go PROC GLM all the way? My feeling is to go with PROC MIXED for >2 treatments since the chance of unbalanced could happen due to subject dropouts (correct me if I am wrong).

Update: I did a comparison in SAS with PROC GLM and PROC MIXED using n=24 unbalanced 4-way study data (3T vs R) with some subjects missing from certain treatments. I just ran Ln(Cmax).

PROC GLM (Ratios; 90%CI)
T1 vs R: 1.0089; 0.9035-1.1264
T2 vs R: 0.9584; 0.8580-1.0203
T3 vs R: 0.8701; 0.7788-0.9713

PROC MIXED (Ratios; 90%CI)
T1 vs R: 1.0094; 0.9030-1.1273
T2 vs R: 0.9646; 0.8606-1.0812
T3 vs R: 0.8710; 0.7798-0.9728

Thanks
John