Reproducibility of ambulatory blood pressure and assessing treatment withdrawal in hypertension trial

Ambulatory blood pressure monitoring (ABPM), like casual blood pressure (BP), is imperfectly reproducible between visits. The reproducibility of ABPM is less than the reproducibility predicted from a simple model of random variation of BP. This suggests either increased technical error in the technique, an element of increased biologic variability, or both. In addition, ABP shows a between-day variation (add about 6 to 7 mm Hg diastolic, similar in magnitude to what is observed with casual BP) which cannot be obviated by the average of many within-day measurements. The reliability of ABPM, although conventionally judged excellent at about RI = 0.75, is insufficient to avoid misclassification errors and regression to the mean when BP cutpoints are employed. About 40% of patients apparently failing treatment withdrawal at a threshold of +2 SD, or about 95 mm Hg diastolic ABP, will be false-positive failures when categorized by a single ABP. Despite the inconvenience, obtaining an average of two or more ABP measurements on different days may be cost-saving to a clinical trial because of the improved reproducibility; other costs may exceed $1000 per entrant.