Abstract
In this study, we applied a general linear mixed model to demonstrate the usefulness of multilevel analysis to assess the association between aging and longitudinal blood pressure variations. The study population which was followed between 1997 and 2000 consisted of 5, 574 male employees over the age of 40 from two worksites in Aichi Prefecture. We adopted both the general linear regression model and the general linear mixed model in statistical analyses where systolic and diastolic blood pressures (BP) were regressed by the observation period of one year, with the baseline body mass index, age, preference for salty taste, daily alcohol consumption, smoking status, leisure time physical activity, and family history of hypertension treated as covariates.In the regression model, aging showed a significant relationship with the diastolic BP increase, but not with the systolic BP increase. In the mixed model, aging was found to be a significant predictor of the longitudinal rise in both systolic and diastolic BP. Random-effect analysis showed a significant inverse relationship between baseline BP and the slope of the regression line of longitudinal BP increase. These findings suggested that the effects of the regression to the mean could be separated at the upper level in the hierarchical model, thus resulting in improvement of the statistical power. We concluded that multilevel modeling was a useful approach to effectively detect the relationship of aging with longitudinal BP variations.
