|
Hi,
I need some help to analyse the following clinical interaction trial.
The study is a randomised, 3 period crossover with three treatments A,
B and (A+B). The wash-out between periods is really important.
The objective of this study is to assess the potential for
pharmacokinetic interaction between A and B when administered
together.
The randomization has been performed using 6 treatment sequences
according to the following scheme :
PERIOD 1 PERIOD 2 PERIOD 3
Sequence 1 Treatment A Treatment B Treatment A+B
Sequence 2 Treatment A Treatment A+B Treatment B
Sequence 3 Treatment B Treatment A Treatment A+B
Sequence 4 Treatment B Treatment A+B Treatment A
Sequence 5 Treatment A+B Treatment A Treatment B
Sequence 6 Treatment A+B Treatment B Treatment A
In each of the 3 treatments (A, A+B, B), PK blood samples have been
collected and the concentration of each drug was measured over period
when it was possible. My parameter of interest is the logarithm of the
AUC (Area under the curve). So, for each subject, I have only two
periods with measurements of the AUC and one period without
concentration for drug A (since when the subject took the drug B alone
the measurement of the concentration of drug A does not exist). So, my
data set has the following form :
Subject sequence period treatment log(AUC)
1 1 1 A 7.5
1 1 3 A+B 8.5
2 1 1 A 6.4
2 1 3 A+B 5.9
3 2 1 A 6.3
3 2 2 A+B 8.2
4 2 1 A 7.5
4 2 2 A+B 8.6
....etc
I want to do the following analysis in order to see if the
concentration is different when the subject took A alone or A and B
together:
PROC MIXED;
Class period treatment subject sequence;
Model AUC=treatment period;
Repeated period/ type=TOEP subject=subject(sequence);
estimate "treatment A versus A+B" treatment -1 1/ cl alpha=0.05;
lsmeans treatment / cl;
Run;
Is it correct since my subjects have only two measures of AUC but the
design has three periods?
In other words, is it a problem for PROC MIXED that the subjects 1 and
2 have values in periods 1 and 3 and no value in period 2 and the
subjects 3 and 4 have values in periods 1 and 2 but no value in period
3 ? Does PROC mixed consider that subjects 1 and 2 have missing values
in period 2? And in this case, is it a problem since the values are
absolutely not Missing At Random, isn'it?
Thanks in advance for your responses
Virginie
| 
