Bioequivalence and Bioavailability Forum

Dear All,

We are planning a BE study of a transdermal patch.
The adhesion will be measured at time points 0.0, 0.50, 24, 48 and 74 hours after application of the patch. Score to be used as follows:
0 = No evidence of irritation;
1 = Minimal erythema, barely perceptible;
2 = Definite erythema, readily visible; or minimal edema; or minimal popular response;
3 = Erythema and papules; and
4 =Definite edema;
5= Erythema, edema, and papules;
6= Vesicular eruption;
7= Strong reaction spreading beyond test (i.e., application) site.
Study design is randomised, double blind, repeat dose, single centre studies designed to assess the potential for Transdermal Patch cause skin irritancy in healthy volunteer.
For the information: FDA requires a comparison between the scores of test and reference treatments ( “The adhesion evaluation of the active test product and RLD must demonstrate that the upper bound of the one-sided 95% CI of the mean adhesion score for the test product minus 1.25 times the mean adhesion score for the RLD must be less than or equal to 0.”

Hence I have made dummy data set and SAS code for this type of study but I want to confirm and do the analysis. Otherwise the result could go wrong.

Anyone who has experience with this kind of statistical analysis, please suggest and guide.

Data one;
infile cards expandtabs;
input Sub  Day  Time Per  Seq $ Form $     Response;
cards;
1    4    0    1    A    R    0
1    4    0.5  1    A    R    3
1    4    24   1    A    R    4
1    4    48   1    A    R    5
1    4    72   1    A    R    7
1    4    0    2    A    T    2
1    4    0.5  2    A    T    3
1    4    24   2    A    T    4
1    4    48   2    A    T    6
1    4    72   2    A    T    7
.
.
.
.
.
20   4    0    2    B    R    2
20   4    0.5  2    B    R    3
20   4    24   2    B    R    6
20   4    48   2    B    R    5
20   4    72   2    B    R    7
20   4    0    1    B    T    2
20   4    0.5  1    B    T    3
20   4    24   1    B    T    4
20   4    48   1    B    T    5
20   4    72   1    B    T    0
;
run;
proc sql;
create table aml as
select Sub, Per, seq, form, avg(response) as avg_res
from one
/*where Time ge 0*/
group by Sub, Per, seq, form;
run;
data aml1;
set aml;
if form="R" then avg_res1=1.25*avg_res;
else avg_res1=avg_res;;
run;
ods trace on;
PROC MIXED data = aml1 ;
CLASS Sub Per seq form;
MODEL avg_res1 = Sub(seq) Per seq form;
LSMEANS form / PDIFF=control('R') CL alpha=0.025 ;
estimate 'T vs R' Form -1 1/CL alpha=0.025;
ods output Estimates=ep;
ods output LSMEANS=lm;
ods trace off;

Regards

---

Edit: Category changed; see also this post #1. [Helmut]