Components of blood pressure variability in the elderly and effects on sample size calculations for clinical trials

This study investigated components of blood pressure variability in the elderly using both ambulatory blood pressure monitoring (ABPM) and casual clinic blood pressure measurement. These were then used to determine sample size requirements for clinical trials of different design scenarios in the elderly. Twenty-six elderly subjects not receiving antihypertensive medication were recruited from general practices and seen on four occasions at weekly intervals. On each occasion of blood pressure was measured in the clinic using a standard mercury sphygmomanometer and then for 24 h using a noninvasive ambulatory monitoring device. The between subject and between subject/within occasion components of blood pressure variability were determined by analysis of variance and used to calculate to sample size requirements for parallel and crossover trials respectively. The between subject variance of mean blood pressure was 1/3 greater with clinic readings, except within a subgroup of subjects who had isolated systolic hypertension (ISH). Increasing the number of readings or occasions on which measurement was performed in a parallel group trial only reduced the variability substantially when the number of subjects involved was small. With crossover designs, the between subject component of variance is eliminated resulting in substantial reduction in sample size. Whereas 60 subjects with ISH would be required to detect a 10 mm Hg difference in systolic blood pressure between two treatments in a parallel design using casual readings, only 18 are required with a crossover trial. If ABPM is used the number of subjects required are 54 and 14, respectively. Reducing variability with ABPM involves a trade-off between the increased number of readings available with the technique against the highly uniform and standardized conditions used to determine clinic blood pressures. ABPM appears most useful as a strategy for reducing sample size in parallel group trials involving small numbers of subjects measured on one occasion.