Choosing the Function of Baseline Run-In Data for Use as a Covariate in the Analysis of Treatment Data from Phase III Clinical Trials in Hypertension

In this work, we simulated (1000 replications) diastolic blood pressure (DBP) data to model a Phase III clinical trial in newly diagnosed hypertensive patients that had 8 consecutive days of baseline run-in data followed by six months of double-blind, randomized treatment with either a drug or placebo. We considered six different patterns (3 linear, 3 non-linear) of baseline run-in DBP data, prior to randomizing patients to treatment with a drug or placebo in balanced fashion (50 per group). We defined 11 functions of the baseline run-in data for use as covariates. Comparative statistical analyses were performed using both repeated measures linear ANCOVA models and longitudinal data analysis models—with or without treatment-by-time interaction and with or without the covariates. Further we assumed the DBP data followed a truncated Normal distribution on the interval (80; 120] mmHg with covariance structure specified as either AR (1) or compound symmetry (CS) with correlation coefficients of 0.1, 0.5 and 0.9. Our objective was to determine the best function of the baseline run-in data to use as a covariate in the comparative statistical analysis of the monthly treatment period data.