Methods of data analysis of ambulatory blood pressure monitoring in clinical trials
de Gaudemaris R; Chau
Blood Pressure Monitoring 1(3): 259-262
1996
ISSN/ISBN: 1359-5237 PMID: 10226240 Document Number: 461214
The objectives of the analysis are to compare the blood pressure levels at different phases of the trial and to be sure that the effect will continue until the next dose. There are problems before the analysis for which consensus is necessary. If the overall 24 h study is the first generally recognized stage, it is difficult to know how to take into account the circadian rhythm: one could determine a day phase and a night phase arbitrarily, one could rely upon the diary of the patient or calculate by using a model of the two phases of individual diurnal rhythms. Analysis of shorter periods reduces the reproducibility of ABPM: what minimum period os sufficiently reproducible (3-4 h for example)?. The methods of statistical analysis of the levels of blood pressure are dictated by the conditions of application of the tests. If a normal distribution is confirmed, the use of parametric tests is required (Student's paired t-test in cross-over studies, analysis of variance for repetitive measures and to investigate the interaction between treatment and time). Non-parametric tests are required for non-continuous distributions (the Wilcoxon test for example). The study of effect over time as performed classically by the peak-trough method poses many problems of analysis. Over what time period should one quantify the peak and the trough?. How should one integrate the placebo effect into a parallel group study?. Is it necessary to calculate the individual peaks and troughs and then take the mean of these or would it suffice to perform a global calculation?. It might be necessary to perform subgroup analyses in order to evaluate the real effects of treatment. How then should one define responders?. Taking into account the 'white-coat' effect and compliance with treatment demands more rigid inclusion and exclusion criteria than do other methods of analysis and limits the validity of the sample.