Characterization of hypertension through multivariate analysis utilizing linear and nonlinear methods

Publication Year:

No metrics available.

Repository URL:
Donnelly, Diane L.
Nonlinear analysis; Linear analysis; Hypertension; Systolic blood pressure; Biomedical Engineering and Bioengineering
thesis / dissertation description
Analysis of blood pressure by nonlinear methods is vastly underutilized in current research. As such, 24-hour ambulatory blood pressure data from a small cohort of borderline hypertensive and normotensive subjects were analyzed using linear and nonlinear methods. Data were collected and provided by researchers from Columbia University. The cohort size was twelve subjects, consisting of two groups of six each. Although disease state was known, group membership for individual subjects was not. Therefore, one aspect of this research was to separate the cohort, a-priori, into two distinct, evenly sized groups, based solely on analysis results. Separation was accomplished by the long-term scaling exponent α2 from detrended fluctuation analysis and parameter SD2 from Poincaré analysis. Linear results did not aid in subject separation. Linear analyses consisted of heart rate and blood pressure variability and baroreflex response. The nonlinear analysis methods included approximate entropy (ApEn), detrended fluctuation analysis (DFA) and Poincaré mapping. Analysis was performed hourly, averaged by group and presented as the mean +1- the standard deviation. Results from linear analysis agree with previous published reports of elevated blood pressure variability, decreased heart rate variability and decreased baroreflex response in hypertension. Results from nonlinear analyses of systolic blood pressure data revealed elevated approximate entropy values in borderline hypertension indicating an increased randomness. The long-term scaling exponent, a2, from the detrended fluctuation was less in borderline hypertension indicating a break down of the scaling properties of systolic blood pressure in the very early stages of hypertension. Poincaré plots alone revealed little difference between subject groups however, the quantitative parameter SD2, which is an indication of the long-term variability, was on average greater in borderline hypertension. Approximate entropy, the long-term scaling exponent α2 from detrended fluctuation analysis and SD2 from Poincaré analysis were statistically significant to p < 0.05 between groups. Statistical significance was determined by paired t-tests over the 24-hour recording period. The broader impact of this work was the finding that nonlinear analysis methods alone facilitated a-priori subject separation. Characterization of hypertension in a close physiologically cohort was achieved through application of nonlinear analysis methods. Linear analysis methods did not aid in determining group membership during any phase of this research. This work and the results that follow are unique due to the use of nonlinear methods in the analysis of systolic blood pressure, specifically in a cohort of borderline hypertensive and normotensive subjects. In addition to the novel use of