Re: st: pk analysis suite: elementary questions about -pkequiv-, -pkcross-, and -pkshape-

["Joseph Coveney" <jcoveney@bigplanet.com>]

Michael McCulloch wrote:

I'm studying the manual (v11) for the pk suite, and have 3 questions:
1. Regarding the "carry" variable created by -pkshape-, in the manual  
example (p. 1325), is "carry" simply an indicator that shows whether  
or not a measurement was made in the same period/treatment  
combination, so that if
	"carry"==0, then no carryover is possible, and if
	"carry"==A, then a carryover effect is possible, but no evidence yet?

2. Can -pkcross- be used in to compare treatment response (e.g. pain  
score) in an AB-BA crossover study design, that was not a test of a  
biopharmaceutical drug concentration?

3. -pkequiv- performs bioequivalence tests using serum levels of  
biopharmaceutical drug concentration. Can the command also be used  
instead to measure *any* continuous outcome, measured following some  
other non-pharmaceutical intervention?

--------------------------------------------------------------------------------

1.  Carryover is a term that indicates what the treatment (drug formulation) was
in the immediately preceding period.  Carryover is zero during the first period;
afterward (where the potential for carryover effects is of concern) its pattern
distinguishes between sequences of first-through-penultimate treatments.  In the
two-sequence, two-formulation, two-period crossover design (i.e., AB BA), during
the second period it is 1 for one of the treatments and 2 for the other.  You
can get a handle on what it does by inspecting the variables -treat- and -carry-
after running the do-file below.

2.  There's no reason why not.  As far as I'm aware, -pkcross- is nothing but a
convenience command that calls -anova- and displays the ANOVA table in a format
more familiar to biopharmaceutics scientists.  Crossover study designs are not
unique to pharmaceutical equivalence evaluation.  And although the phenomenon is
called "carryover" in biopharmaceutics, you will have the same concern in other
disciplines, namely, that the first exposure to a condition might change things
so that a second exposure will result in a different score ("learning
phenomenon", "sensitization", "desensitization", "stimulus fatigue", "acquired
tolerance").

3.  Again, there is no reason not to:  the term *equivalence* has more general
usage than for drug concentrations in blood plasma.  Keep in mind that the
potential for carryover effects (by whatever name) to confound crossover studies
is often more serious with outcomes other than chemical concentrations.  And it
sometimes gets tricky even with the latter.

Joseph Coveney

clear *
set more off
set seed `=date("2010-02-20", "YMD")'
input str1 trt
"A"
"B"
"C"
end
tempfile tmpfil0
save `tmpfil0'
rename trt seq
forvalues replicate = 1/2 {
    cross using `tmpfil0'
    replace seq = seq + trt
    drop trt
}
erase `tmpfil0'
generate byte id = 3
expand id
sort seq
replace id = _n
forvalues per = 1/3 {
    generate byte period`per' = floor(runiform() * 100)
}
pkshape id seq period1 period2 period3
sort id period
order id sequence period treat carry outcome
list, noobs sepby(id)
exit



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/