Evidence-based proposal for the number of ambulatory readings required for assessing blood pressure level in research settings: an analysis of the IDACO database

Background: Guidelines on the required number of ambulatory blood pressure (ABP) readings focus on individual patients. Clinical researchers often face the dilemma of applying recommendations and discarding potentially valuable information or accepting fewer readings. Methods: Starting from ABP recordings with ≥30/≥10 awake/asleep readings in 4277 participants enrolled in eight population studies in the International Database on Ambulatory Blood Pressure in Relation to Cardiovascular Outcomes (IDACO), we randomly selected a certain number of readings (from 30 to 1 awake and 10 to 1 asleep readings) at a time over 1000 bootstraps at each step. We evaluated: (i) concordance of the ABP level; (ii) consistency of the cross-classification based on office blood pressure and ABP; and (iii) accuracy in predicting cardiovascular complications. For each criterion, we fitted a regression line joining data points relating outcome to the number of readings covering the ranges of 30-20/10-7 for awake/asleep readings. Results: Reducing readings widened the SD of the systolic/diastolic differences between full (reference) and selected recordings from 1.7/1.2 (30 readings) to 14.3/10.3 mm Hg (single reading) during wakefulness, and from 1.9/1.4 to 10.3/7.7 mm Hg during sleep; lowered the κ statistic from 0.94 to 0.63, and decreased the hazard ratio associated with 10/5 mm Hg increments in systolic/diastolic ABP from 1.21/1.14 to 1.06/1.04 during wakefulness and from 1.26/1.17 to 1.14/1.08 during sleep. The first data points falling off these regression lines during wakefulness/sleep corresponded to 8/3 and 8/4 readings for criteria (i) and (iii) and to 5 awake readings for criterion (ii). Conclusions: 24-h ambulatory recordings with ≥8/≥4 awake/asleep readings yielded ABP levels similar to recordings including the guideline-recommended ≥20/≥7 readings. These criteria save valuable data in a research setting, but are not applicable to clinical practice.