Blood pressure measurement in epidemiological studies: a comparative analysis of two methods. Data from the EPIC-Potsdam Study. European Prospective Investigation into Cancer and Nutrition

Abstract

Objective: To investigate the impact of a device replacement in blood pressure measurement in epidemiological studies on comparison and interpretation of epidemiological data by replacing traditional aneroid manometry with automated oscillometric devices.

Design: Within the context of a continuing epidemiological study (EPIC-Potsdam Study), blood pressure measurements were performed simultaneously with an aneroid sphygmomanometer and an automated oscillometric device for each subject. We randomly selected 400 men and women from the main study population and one observer performed three consecutive blood pressure measurements for each subject according to a standardized procedure. In total, 10 oscillometric devices of the same type were used. Demographic and anthropometric data for each subject were obtained by trained interviewers.

Results: The mean difference between the aneroid and the oscillometric measurements of systolic blood pressure was 0.2 +/- 5.6 mmHg (aneroid value greater, NS), whereas the mean difference in diastolic blood pressure, 0.5 +/- 3.5 mmHg (aneroid value smaller), attained statistical significance (P = 0.0001). Estimates of prevalence for hypertension differed by 0.4% for women, and by 2.9% for men. Associations with the differences between methods were observed with age, blood pressure, lean body mass, upper arm circumference and specific devices without indicating a strong and consistent pattern.

Conclusions: The use of automated oscillometric devices in epidemiological studies introduces a bias of very small magnitude compared with use of the aneroid method. The effect of the change to this automated measurement procedure on prevalence estimates is small but might affect comparability of data. Minor differences in measurement performance between devices of one type might affect population parameters.