PROC MIXED and early phase trials

Abstract

Listen

Translate article

PROC MIXED is commonly being used to compare treatment or other differences in phase 1 crossover trials. In such trials there is variation between subjects and also variation within subjects — these two sources of variation can be described by random effects. PROC MIXED is used because it can accommodate for random effects and as opposed to other SAS regression procedures, subjects with missing observations can be handled without removing all of their data from the analysis. A fixed effect is a parameter which is modelled in the same way as in PROC GLM — there are pre-specified levels of that effect, e.g. treatment group which is pre-defined in a trial because the aim is to compare responses among the fixed groups. In contrast a random effect is a parameter whose values cause a random variability within a trial and whose values are not known pre-trial, e.g. the subjects' responses in a trial. So commonly, subject is declared as a 'random' parameter in PROC MIXED to account for this random variation in the statistical analysis. The purpose of this paper is to provide an introduction to the PROC MIXED procedure by describing some code that has been used to carry out an analysis on a phase 1 crossover trial.