st: Re: which statistical analysis to evaluate vitals signs

["Joseph Coveney" <stajc2@gmail.com>]

Dao Kim wrote:

Hi, which statistical analysis would you use to analyse tolerance to a drug, for
evaluation of blood pressure, oxygen saturation, pulsation before and after a
drug, to detect a 10% difference between both values?

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

Because your description calls for only a single administration, I'm guessing
that you meant drug tolerability (absence of side effects) and not drug
tolerance (attenuation of effects upon repeated exposure).  How about MANOVA?
You could then jointly test for a difference in before-and-after change scores
for each of the three vital signs.  The null hypothesis would be that the
absolute value of the mean change score for at least one of them exceeds 10%
(you'll need to decide the denominator for that), that is, the drug has poor
tolerability.  The alternative hypothesis would then be that none of the
absolute values of the three mean change scores exceeds 10%, that is, the drug
at the examined dose level has good tolerability in terms of the three vital
signs.  (You would construct joint 90% confidence intervals and interpret them
just as in a therapeutic equivalence--a.k.a., bio-equivalence--study.)  Power
analysis could be by simulation (readily done in Stata) using whatever you've
got now on correlation structure.  You'd need to look into how well the
underlying assumptions for MANOVA hold up, and you might need to consider
transforming the oximeter measurements, but, again, it would allow for joint
testing of hypotheses for the correlated data.  . . . should be at least worth
suggesting to your site's biostatistician.

Joseph Coveney

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