Back to the Top
Hi
I would like to ask community for opinion about the way Kinetica 4.4
performs bioequivalence in a conventional two-treatment, two-period,
two-sequence (2 x2) randomized crossover design.
From my point of view, the proper way to calculate it according to
FDA-Guideline is (SAS program):
proc mixed data=sas2;
class subject sequence period treatment;
model lnAUC = sequence period treatment;
random subject(sequence);
estimate 'treatment difference' treatment -1 1 / alpha 0.1 c1;
run;
However, Kinetica method "Latin square" doesn't use subjects nested
in sequence (in difference to comment given in Kinetica's help, which
refers to FDA-conform calculation). One can simulate the results of
Kinetica's bioequivalence calculation by following SAS programs:
(1)
proc glm data=sas2;
class sequence period treatment;
model lnAUC= sequence period treatment;
estimate 'treatment difference' treatment -1 1 / alpha 0.1 c1;
run;
a delivers non correct effects for period, sequence and treatment
(2)
proc glm data=sas2;
class subject sequence period treatment;
model lnAUC= subject sequence period treatment;
estimate 'treatment difference' treatment -1 1 / alpha 0.1 c1;
run;
a delivers correct effects of period and treatment, but no results
for sequence
In summary, Kinetica's Latin square-method provide effects of period
and treatment from (2) and sequence effect from (1). Especially,
power calculation provided by Kinetica's approach might suffer from
this calculation.
I would like to your opinions, whether Kinetica's way to calculate
bioequivalence can be considered as an acceptable and FDA-conform
procedure or not.
Best regards Thomas Keller
www.addstat.com, Leipzig, Germany
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)