Logo cin2003

Discussion Board

Paneles de Discussión

Paneais de Discussio

Free Papers

Comunicaciones libres

Comunicaçoes livres

Home cin2003

Volver al Inicio cin2003

Voltar ao inicio cin2003

Methods for Quantitative Analysis of Ambulatory Blood Pressure Monitoring

A. Díez-Noguera*, M.A. Sans-Fuentes*, X. Sarrias**.

*Departament de Fisiología, Facultat de Farmacia. Universitat de Barcelona.
**Servei de Nefrología, Hospital de Bellvitge, Barcelona. España


The continued improvement of arterial blood pressure monitors and their availability has promoted the use of this exploration as a routine clinical practice 1. Nevertheless the great number of data that this method provides is not adequately analyzed, and a 10% of the information is hardly used. Often the analysis is reduced to the single ocular inspection of the daily profiles. Consequently it is of great interest to develop analytical methods suitable for diagnostic purposes, and propose selected indexes with an intuitive interpretation 2. Previously to study the diagnostic value of a specific index, it is necessary to evaluate the reliability and formal consistency of the different indexes one can calculate from the data obtained through the ambulatory monitors. This is just the aim of the present work. We studied systematically a wide number of numerical indexes (360) derived from 752 records, in order to study their formal properties and suggest the most informative ones.

Material and methods

To obtain the arterial blood ambulatory recordings (ABMR) 3 types of monitors were used: Spacelab 90207, Takeda 2020, and Profilomat. Each individual record elapsed a duration comprised between 23 and 24 hours. Sampling interval was fixed to 15 minutes from 6:00 to 23:00, and 30 minutes from 23:00 to 6:00. The study was carried out over a sample of 752 patients both normotensive and hypertensive, without any medication. Consequently this sample could be considered representative of the population of patients visited for the first time.

Previously to the analysis, each data series were filtered by dropping out those measurements laying out of the interval 50-250 for the S-ABP, 30-200 for the D-ABP and 30-200 for the heart rate. The data showing an absolute difference greater that 60 with respect to the preceding measurement were also removed.

Data analysis was carried out with the program GARAPA. Using this software we defined 12 different time intervals along the day (see the Table) in which the different indexes were calculated independently. In each interval the following indexes were calculated separately for systolic (S) and diastolic (D) blood pressures: MD mean data value, DO (distributed overload) is the area comprised between the arterial blood pressure (ABP) curve and the threshold divided by the time interval, TO (time of overload) is the real time that the ABP exceeds the threshold value, %T is the TO expressed as the percentage of the duration of the interval, and %D that represents the percentage of measurements over the threshold. In this way we obtain 10 values for each interval giving a total of 120 indexes.

Name of the interval

From ...

... to

Sys/Dias threshold



Wake up

Go to bed

135 / 85 mm Hg


Short day

Wake up +1 h

Go to bed –1 h

135 / 85 mm Hg


First half of the day

Wake up

Wake up +6 h

135 / 85 mm Hg


Second half of the day

Go to bed –6 h

Go to bed

135 / 85 mm Hg


Center of the day

Wake up +4 h

Go to bed –4 h

135 / 85 mm Hg



Go to bed

Wake up

115 / 75 mm Hg


Short night

Go to bed +1 h

Wake up –1 h

115 / 75 mm Hg


First half of the night

Go to bed

Go to bed +4 h

115 / 75 mm Hg


Second half of the night

Wake up –4 h

Wake up

115 / 75 mm Hg


Center of the night

Go to bed +2 h

Wake up –2 h

115 / 75 mm Hg


After placing short

After placing

After placing +2 h

135 / 85 mm Hg


After placing long

After placing

After placing +4 h

135 / 85 mm Hg


This procedure was applied to the original data (previously filtered), then the same procedure was repeated over a data series obtained after fitting the values to a sinusoidal wave by mean of squares regression (Cosinor method 3-4). The equation used is: , where y is the variable to be analyzed (S-ABP or D-ABP), t is the time in hours, M is the mesor, or mean value around which the variable oscillates, A is the amplitude of the oscillation and is the acrophase, that indicates the moment of the day at which the function reach the maximum. Knowing the parameters M, A and one can generate the new series of estimated data over which the 120 indexes can be calculated. It is important to point out that the new estimated series retains the main characteristics of the original one but eliminates small details and local variations occurred along the day.

The last set of indexes is obtained in the same way but analyzing another estimated series of data. In this case we used Fourier analysis to calculate and estimated function containing six sinusoidal functions (harmonics 5-7). In this case the equation is: . Similarly to the case of Cosinor analysis, the main characteristics of the original series are retained and also those local changes that are present for intervals longer than four hours.

As a result of all these calculations 360 indexes were obtained. All these indexes were systematically analyzed calculating means, standard deviations and studying its correlation with the other indexes (Perason’s r). Mean value comparisons were done by the Student’s t test.


Discussion of results and conclusions

Due to the extension of results only some of the must relevant will be discussed. The first thing to discuss is the fact that practically in all overloads calculated from the Cosinor analysis the day-night differences are bigger than those calculate over the original data. This can be explained after considering the details of the calculation: the fitted sinusoidal curve (used for the analysis) normally shows its maximum during the day and the minimum at night. When calculating the area between the curve and the threshold during the night, we have a geometric figure that exaggerates the differences at the extremes of the interval, and consequently the corresponding overload is slightly augmented. Inversely, during the day the overload is reduced. So we see that this method introduces a bias, but, in this case, is a known bias that emphasizes one of the most interesting values: the night overloads.

It is important to point out that all the results obtained for the S-ABP are practically equal to the ones obtained for D-ABP. It is also important to consider the values of thresholds. We used the values commonly recommended in the literature 8, and we observed that there is a tendency to find slightly higher values of overload in the nocturnal intervals of S-ABP. This result clearly suggest that the nocturnal threshold for the S-ABP (115 mm Hg) is probably a bit low, since the overloads calculated on this basis are bigger that the others.

Finally, the commonly used index %D 9-11 has been analyzed studying its correlation with %T, TO and DO. The results show a high correlation between %D and its homologous %T, but they also show that this correlation is reduced when considering the nocturnal intervals. In addition during these intervals the regression coefficient becomes smaller than one, indicating that the values calculated with %D are a bit higher than the real ones obtained with %T. This confirms the convenience of using %T instead of %D, specially at night, when the differences are bigger depending on the methods.

In summary: there are no differences to justify the use of different indexes for the analysis of S-ABP and D-ABP. In general it would be recommendable the quantification of overload on basis to the indexes DO and %T, since their interpretation has a clear biological meaning and shows values much more stables that other indexes currently used, as the %D. The usage of the Cosinor analysis, besides providing information about the amplitude and the acrophase of the circadian rhythm, gives a realistic approximation to the real daily fluctuating values, emphasizing the nocturnal overloads. On the other hand, the stability of the indexes is higher at night, probably due to the increased variability of the recordings during the day.


  1. Mallion JM, Baguet JP, Siché JP, Tremel F, De Gaudemaris R. Clinical value of ambulatory blood pressure monitoring. J Hypertens 1999; 17:585-595.
  2. Cotas AJS, Clark SJ, Conway J. Analysis of ambulatory blood pressure data. J Hypertens 1991; 9:S19-S21.
  3. Bingham C, Arbogast B, Cornelissen GG, Lee G, Halberg F. Inferential statistical methods for estimating and comparing cosinor parameters. Chronobiologia 1982; 9: 397-439.
  4. Díez-Noguera A, Cambras T. Determinación de las características del ritmo en variables biológicas. Método del Cosinor. Inf Med Biol 1989; 1(3): 25-30.
  5. Maching T, Engels G, Schöbel S, Gellert J, Bachmann K. A new aproach for evaluating circadian blood pressure levels and variability from 24h ambulatory blood pressure profiles. Am J Hypertens 1993; 6:248-252.
  6. Chau NP, Mallion JM: Twenty-four-hour ambulatory blood pressure in shift workers. Circulation 1989; 80: 341-347.
  7. Díez-Noguera A, Cambras T, Vilaplana J, Casamitjana N. Cronobiología. Edicions de la Universitat de Barcelona, pp25-26. Barcelona 1997.
  8. Myers MG, Haynes RB, Rabkin SW. Canadian hypertension society guidelines for ambulatory blood pressure monitoring. Am J Hypertens 1999; 12:1149-1157.
  9. Mulè G, Nardi E, Andronico G, Cottone S, Raspanti F, Piazza G, Volpe V, Ferrara D, Cerasola G. Relationships between 24 h blood pressure load and target organ damage in patients with mild-to-moderate essential hypertension. Blood Press Monit 2001; 6:115-123.
  10. Tsioufis C, Stefanadis C, Goumas G, Pitsavos C, Toutouzas P. Relation of blood pressure load with left ventricular geometry in untreated patients with mild-to-moderate hypertension. J Hum Hypertens 1999; 13:667-682.
  11. White WB, Lund-Johansen P, Omvik P. Valor de la sobrecarga tensional de 24 horas en la evaluación del tratamiento antihipertensivo. J Hypertens 1993; 11:S75-S80.