Estimating the prevalence of hypertension corrected for the effect of within-person variability in blood pressure

• Published in: Journal of clinical epidemiology 2000-11, Vol.53 (11), p.1158-1163
• Main Authors: Klungel, Olaf H, de Boer, Anthonius, Paes, Arsenio H.P, Nagelkerke, Nico J.D, Seidell, Jacob C, Bakker, Albert
• Format: Article
• Language: English
• Subjects: Adult; Arterial hypertension. Arterial hypotension; Biological and medical sciences; Blood and lymphatic vessels; Blood Pressure; Cardiology. Vascular system; Clinical manifestations. Epidemiology. Investigative techniques. Etiology; Confidence Intervals; Cross-Sectional Studies; Humans; Hypertension; Hypertension - epidemiology; Medical sciences; Middle Aged; Prevalence; Prevalence estimation; Reference Values; Reproducibility of Results; Reproducibility study; Within-person-variability
• ISSN: 0895-4356; 1878-5921
• DOI: 10.1016/S0895-4356(00)00228-6

The objective of our study was to assess the applicability of using estimates of within-person variance (WPV) from reproducibility studies for a correction of blood pressure values in another study to improve the accuracy of the prevalence estimation of hypertension. Data were collected from two cross-sectional population-based studies on cardiovascular disease risk factors conducted from 1987 to 1995 among 55,026 subjects aged 20–59 years. Correction factors were calculated from a reproducibility study among 924 subjects who were examined in 1989 and 1990. All other studies with repeated measurements of blood pressure were searched in MEDLINE from 1966 onward. Six studies satisfied the inclusion criteria. The prevalence of hypertension, uncorrected and corrected with factors from other studies, were compared with the prevalence of hypertension corrected with the factor from our study. The uncorrected prevalence of hypertension was 17.3% [95%CI:17.0–17.7]. The prevalence of hypertension after correction for WPV with the factor from our study was 13.5% [95%CI:13.2–13.8]. Correction for WPV with factors from the appropriate studies (depending on factors such as number of measurements taken per visit, and time interval between visits) resulted in prevalences ranging from 13.9% to 14.7%. The bias that occurs when no correction for WPV is performed is much larger (29% overestimation) than the bias that occurs when correction factors are derived from other studies (3.1–8.4% overestimation). When repeated measurements of blood pressure are not available in a population study for a sample of that same study, it is advisable to use data from another study to correct for WPV.