Dear SAS Statisticians,
I have an investigator that has a data set of around 195 patients that each
had taken drug A with escalating doses over 3 months. Of these 195
patients, 19 patients left the study early due to no efficacy, AE, or cost.
The Physician collected such variables as time to withdrawal, nature of the
AE, demographics of patients (gender, age etc), pre-existing psychological
problems, dose of drug and whether sleep apnea was present.
The physician would like to be able to predict which of the covariates he
has collected would/may predict (1) discontinuation from the study and (2)
Whether certain pre-existing conditions would predict or be indicative of
certain AEs (eg sleap apnea or certain psychological condition).
Given the size of the data set from which he has to work, would anyone care
to offer their opinion as to the feasability of modeling his ideas. Not
being a professional statistician I was thinking he may need something like
loglinear analysis using logits or nominal gereneralized logits models with
a polytomous response.
Any thoughts on the usefulness of this approach would be appreciated.
Alex M. Gray, Ph.D.
Clinical Development Scientist
Five Moore Drive
RTP, NC 27709-3398
* (919) 483 1820 * Room MAI C-2610