Subject-Level Trend Analysis in Clinical Trials

Alexandru-Ionut PETRISOR
Liviu DRAGOMIRESCU
Cristian PANAITE
Alexandru SCAFA-UDRISTE
Anamaria BURG

Keywords

simple linear regression, slope, coefficient of determination, binomial test

Abstract

In particular situations, clinical trials researchers could have a potential interest in assessing trends at the level of individual subjects. This paper establishes a common approach and applies it in two different situations, one from nutritional medicine and one from cardiovascular medicine. The approach consists of running as many regression models as the number of subjects, looking at the behavior of some parameter of interest in time. The regression parameters, particularly the slope of the regression line, offer the general sense of the trend and allow for testing its statistical significance. Extrapolation at the level of the entire sample is possible using some version of the binomial test. In both cases, significant results were obtained despite of small sample sizes.