Assessment of ambulatory blood pressure monitoring: better reproducibility with polynomial analysis

Objective: Ambulatory blood pressure monitoring (ABPM) data using values of arbitrarily separated day- and nighttime hours are poorly reproducible, undermining the validity of this diagnostic tool. Previous studies from our group have demonstrated that polynomial curves can be produced of ABPM data from both normo- and hypertensive groups, and that these polynomial curves are within the 95% confidence intervals of the data means. However, intra-individual reproducibility of this approach has not been assessed, and is an important prerequisite for further implementing this approach. Methods: Reproducibility was studied in 10 untreated mildly hypertensive patients by performing 24 hour ABPM monitoring in each of them 2 times, intervals at least one week. ABPM monitoring was performed using validated Space Lab Medical Inc portable equipments, polynomial regression analyses of the systolic blood pressures using Harvard Graphics 3 as well as SPSS Statistical Software. Polynomes were compared,vith the actual data as measured. Results: Reproducibility of means of the population: Polynomes of duplicate 24 hour observations were not significantly different from each other (p=0.44), The duplicate standard errors of the polynomes of the data were significantly better reproducible (P <0.001) than those of the actual data (1.86 mmHg and 15.9 mmHg, respectively). So was intra-class correlation (98.6% and 46%, respectively, P <0.001). Reproducibility of the individual data: Duplicate standard errors of raw data were generally more than twice the size of those of the polynomes, while intraclass correlations of raw data were accordingly generally almost half the size of those of the polynomes, Pooled differences were statistically highly significant both for duplicate standard errors and for intraclass correlations (P <0.001 and P=0.009, respectively). Conclusions: Reproducibility of polynomial analysis of ABPM data is fundamentally better than that of actual data, and this is so not only with means of populations but also with individual data. We assume that the difference in reproducibility is due to the potential of polynomial analysis to remove exogenic components from the data and thus visualize the true endogenic circadian rhythm of blood pressures