Computer simulations of blood pressure tracking

Individual blood pressures (BP) tend to remain in the same relative position within a population BP distribution ('tracking'). Individuals in the upper range of the distribution may be predisposed to future hypertension, but this has not been proven epidemiologically. In this study tracking was simulated by computer using 4 simple empirical models to assess commonly reported criteria of BP tracking (correlation coefficient (r) and quintile analysis) and the relationship between the rank position of BP and the subsequent development of 'hypertension' or 'death'. The four models were applied to a population (n = 1500) of normally or log-normally distributed values corresponding to BPs with a mean of 75 and SD of 10 mmHg, also generated by the computer, and to real BP data from a longitudinal population survey. Each iteration of the model represented one year. Changes in BP at each iteration were cumulative. With both normally and log-normally distributed data, r for paired initial and final BPs decreased progressively but remained statistically significant for all models, up to 30 'years'. However, r was low. The patterns resembled that found in the longitudinal survey and data obtained by submitting the initial real BP data from the same survey to the simulations. In agreement with epidemiological studies, there was frequent movement of BP values between distribution quintiles, confirming that tracking is inefficient. In the computer-derived data, the relative risk of 'death' or 'hypertension' was greatest in the topmost quintile. The results suggest that though physiological BP control mechanisms are complex, adequate description of BP trends with time in a population is possible using simple models.(ABSTRACT TRUNCATED AT 250 WORDS)