In longitudinal research studies with follow-up examinations, the devices used to measure phenotypes may change over time. When a device change occurs, the two devices should be calibrated to each other to ensure that measurements are comparable. This paper details the Jackson Heart Study (JHS) blood pressure (BP) comparability study. Two of the authors, Drs. Lisandro Colantonio and Paul Muntner, are in the department of epidemiology at University of Alabama at Birmingham School of Public Health.
During its second clinic exam (2005–2008), the JHS switched from a random-zero sphygmomanometer (RZS) BP measurement device to an oscillometric device (OD). During this exam, BP measurements from both an RZS and an OD were taken simultaneously in 2117 participants for the purpose of calibration. Five methods for calibrating systolic BP (SBP) and diastolic BP (DBP) were considered: ignoring the change, ordinary least squares regression, adding the average difference, Deming regression, and robust regression.
Using the RZS and OD, the mean (SD) SBP was 125.5 (19.2) and 126.5 (19.9), respectively, and the mean (SD) DBP was 76.4 (10.6) and 74.0 (11.0), respectively. The correlation between RZS and the OD was 0.90 for SBP and 0.80 for DBP. The prevalence of high BP and hypertension and associations with albuminuria were similar when applying each of the five calibration methods. Robust regression was chosen for calibration, giving the following equations:
These equations had a higher R2 statistic than using calibration equations from the Coronary Artery Risk Development in Young Adults Study and the Heinz Nixdorf Recall Study.
The JHS BP data have been calibrated using the above equations for use in future analyses.Friday Letter Submission