Sample size for short-term trials of antihypertensive drugs

1 Controlled trials of antihypertensive drugs published in the British Journal of Clinical Pharmacology during 1979 and 1980 were examined. Studies comparing two or more active drugs or dosage regimens nearly always failed to separate and treatments significantly. The sample size (mean 19 patients) and power of these studies were too low. 2 When planning such studies the aims should be a power of at least 0.8; significance 0.05 or less; and to detect a difference between treatments of 10/5 mmHg, or 6.7 mmHg mean arterial pressure (MAP). The sample size needed can be derived readily from a nomogram if the standard deviation of differences (SDD) between BP measurements under trial conditions is known. 3 In five studies the SDDs were fairly constant despite different observers, patient groups and measuring devices, at approximately 14 mmHg systolic, 9 mmHg diastolic, and 9 mmHg MAP. Use of three BP measurements at each visit reduced the SDD by about 1 mmHg, and would reduce the sample size required by about 20%. Replicate BP measurements at separate visits would be expected to have a larger effect on the power of the study. 4 Published studies with negative results should give an estimate of the power of the study.