TIME-WEIGHTED VS CONVENTIONAL QUANTIFICATION OF 24 HOUR AVERAGE AMBULATORY SYSTOLIC BLOOD PRESSURE

Original article 459

Time-weighted vs. convention
al quantification of 24-haverage systolic and diastolic ambulatory blood pressuresJose A. Octavioa,b, Jesus Contrerasb, Pablo Amairb, Bernardo Octavioa,Domenico Fabianoa, Federico Moleiroa, Stefano Ombonic,Antonella Groppellic, Grzegorz Biloc, Giuseppe Manciac,d
and Gianfranco Paratic,d

Background Conventional calculation of mean 24-h

ambulatory blood pressure (BP), SBP and DBP based on

the average of all BP readings disregards the fact that a

larger number of measurements is usually scheduled

during the daytime than at night, an imbalance possibly

leading to an overestimation of 24-h average BP. The aim of

our study was to quantify this possible bias and to explore

its determinants.

Methods Four hundred and fifty untreated individuals were

subdivided into three groups (150 individuals each) with

three different ambulatory blood pressure measurement

schedules for day/night: group I, four (day)/two (night)

readings/h; group II, four (day)/three (night) readings/h;

and group III, with BP readings every 30 min throughout

24 h. Hourly and 24-h averages were computed. The

conventional 24-h averages of all SBP and DBP values were

compared with the averages of hourly SBP and DBP mean

values (time-weighted quantification). The difference

between 24-h conventional and 24-h time-weighted BP was

computed in each group and related to the degree of

nocturnal BP dip and to the ratio between the number of

readings of day and night.

Result In the three groups, 24-h conventional and 24-h

time-weighted BP values were highly correlated (r > 0.99),

24-h conventional SBP and DBP being significantly higher

(P < 0.01) than the corresponding 24-h time-weighted

values in groups I and II but not in group III (Bland–Altman

Copyright © Lippincott Williams & Wilkins. Unaut

0263-6352 � 2010 Wolters Kluwer Health | Lippincott Williams & Wilkins

analysis). The bias magnitude was related to the day/night

ratio in number of readings and to nocturnal BP dip in

groups I and II (P < 0.01) but not in group III.

Conclusion The higher number of readings/h during

daytime leads to an overestimation of conventional 24-h

average BP, particularly in individuals with preserved

nocturnal BP dipping. This can be avoided either by

scheduling the same number of readings/h throughout

24 h or by performing a time-weighted quantification

of 24-h BP. The clinical implications of these different

approaches deserve further investigation. J Hypertens

28:459–464 Q 2010 Wolters Kluwer Health | Lippincott

Williams & Wilkins.

Journal of Hypertension 2010, 28:459–464

Keywords: ambulatory blood pressure monitoring, average 24-h bloodpressure, nocturnal blood pressure fall, time-weighted analysis

Abbreviation: ABPM, ambulatory blood pressure monitoring

aExperimental Cardiology, Tropical Medicine Institute, Central University,bHospital de Clinicas Caracas, Caracas, Venezuela, cDepartment of Cardiology,IRCCS San Luca Hospital, Istituto Auxologico Italiano and dDepartment ofClinical Medicine and Prevention, University of Milano-Bicocca, Milan, Italy

Correspondence to Professor Gianfranco Parati, MD, IRCCS Ospedale SanLuca, Istituto Auxologico Italiano, via Spagnoletto 3, 20149 Milan, ItalyTel: +39 02 619112949/890; fax: +39 02 619112712/956;e-mail: [email protected]

Received 2 October 2009 Accepted 2 November 2009

IntroductionAmbulatory blood pressure (BP) monitoring (ABPM)

carries important information in the management of

hypertensive patients [1–12], 24-h average systolic

(SBP) and diastolic pressures (DBP) being among the

most important parameters considered. In the majority of

cases, these 24-h averages are computed as the mean of

the total number of readings obtained during the whole

24-h recording time. On the contrary, ABPM devices are

programmed with a variety of BP measurement schedules

in different centers. In some of them, ABPM devices are

programmed to obtain BP measurements evenly spaced

throughout the 24 h, with BP readings every 60, 30, 20 or

15 min [6,13,14]. In other centers, a different frequency of

measures is scheduled during the daytime and the night-

time, with BP measurements programmed every 15, 20

or 30 min during the day and every 20, 30, 40 or 60 min

during night [15–18]. When using more frequent

measurements per hour during the day than during the

night, the calculation of 24-h SBP and DBP based on the

average of all 24-h values introduces a bias related to

the greater relative weight of the higher daytime over the

lower night-time number of BP values. To our know-

ledge, the magnitude of this bias has not been specifically

addressed before and is not usually taken into account

when interpreting ABPM data. The aim of our study was

to investigate the practical relevance of this methodo-

logical problem by comparing the 24-h SBP and 24-h

DBP values obtained by averaging all readings (conven-

tional method) with those obtained by the average of

horized reproduction of this article is prohibited.

DOI:10.1097/HJH.0b013e328334f220

mailto:[email protected]

http://dx.doi.org/10.1097/HJH.0b013e328334f220

C

460 Journal of Hypertension 2010, Vol 28 No 3

Table 1 Number of blood pressure readings for the daytime, thenight-time and the 24-h period, and the day/night ratio of suchnumbers

Daytime Night-time 24-h 24-h D/N ratio

Group I 61.5�5.2 13.9�0.5 75.4�5.3 4.4�0.4Range 49–71 12–17 63–85

Group II 63.2�4.0 21.8�2.2 85.0�4.8 2.9�0.3Range 40–73 15–28 56–95

Group III 34.7�2.6 13.9�0.6 48.5�2.5 2.5�0.2Range 29–39 13–16 42–54

Values are expressed as mean�SD, with the addition of the range betweenmaximum and minimum number of readings for each of the three groups con-sidered in our study. 24-h D/N ratio, ratio between 24-h daytime and night-timemeasurements.

hourly BP means, that is, by a method that takes the

different frequency of BP measures per hour during

daytime and night-time into account (time-weighted

method). This comparison was made by considering data

obtained from ABPMs performed with different BP

measurement schedules during the day and night.

MethodsFor our study, we considered good quality ABP recordings

consecutively performed in 450 patients who were referred

for the first time to our three centers (one in Milan, Italy,

and two in Caracas, Venezuela) for suspected arterial

hypertension and, thus, were not under treatment at the

time of the study. Because of the differences in the usual

procedures followed by the three different centers in

programming ABPM devices, we were able to obtain three

sets of recordings, including data from 150 individuals

each, characterized by three different BP measurements

scheduled over the 24 h: group I, with BP readings every

15 min during the daytime (0600–2300 h) and every

30 min during nighttime (2300–0600 h); group II, with

BP readings every 15 min during the daytime and every

20 min during night-time; and group III, with BP readings

scheduled every 30 min during the whole 24-h period.

Recordings with less than 75% of the total number of

readings programmed over the 24 h, the daytime and the

night-time, were excluded from further analysis. Record-

ings with 2 or more consecutive hours without successful

measurements were also excluded.

Data analysisFor each recording, we computed the number of readings

obtained over the 24 h, the daytime and the night-time.

The ratio between the number of BP readings obtained

during daytime and night-time was also computed. The

average 24-h SBP and DBP values were estimated with a

dedicated software in two different ways: by the conven-

tional method (average of all values) and by calculating the

mean of hourly averages during the 24 h (time-weighted

method). The degree of nocturnal dip of SBP and DBP was

also calculated as the percentage difference between day-

time (0600–2300 h) and nighttime (2300–0600 h) mean

values. In each group, we assessed the correlation between

the 24-h SBP and 24-h DBP values obtained by the two

methods (conventional vs. time-weighted). The differ-

ences in overestimation bias for SBP and DBP between

the three groups were calculated and compared by one-

way analysis of variance (ANOVA) with Bonferroni’s cor-

rection. The difference in 24-h BP values obtained by the

two methods was also assessed by the Bland–Altman

analysis. For each group, the correlation of the between-

method difference in 24-h SBP or DBP with the degree of

nocturnal dip of SBP and DBP was also explored. For each

group, Pearson correlation coefficients between the con-

ventional vs. time-weighted method difference in 24-h BP

and the day/night ratio in the number of readings during

the 24 h were also assessed. A P value of less than 0.05 was

opyright © Lippincott Williams & Wilkins. Unauth

taken as the minimum level of statistical significance

throughout the study. Values are expressed as mean�SD,

SD, unless otherwise indicated.

ResultsTable 1 shows the number of BP readings obtained in each

group during the daytime, the night-time and the entire

24-h period. The day/night ratio in reading number was

much higher in group I than in groups II and III. Although

24-h SBP or 24-h DBP obtained by the conventional

method and by the time-weighted method in each group

was closely correlated (r values always>0.99), the average

values obtained by the two methods were significantly

different. This is shown in Fig. 1 by Bland–Altman

analysis of the differences between conventional and

time-weighted 24-h SBP or 24-h DBP values obtained

in the three groups. In group I, in which the day/night ratio

in the number of measurements was 4.4þ 0.4 (Table 1),

the mean overestimation biases associated with the use of

conventional analysis were 1.95� 1.4 and 1.5� 1.0 mmHg

for 24-h SBP and DBP, respectively. In group II, in which

the day/night ratio in the number of measurements was

around 2.9þ 0.3 (Table 1), the corresponding mean over-

estimation biases carried by conventional analysis were

0.5� 0.7 and 0.4� 0.5 mmHg for 24-h SBP and DBP,

respectively, whereas in group III, in which the day/night

ratio in the number of measurements was around 2.5þ 0.2

(Table 1), they amounted to 0.3� 0.8 and 0.2� 0.5 mmHg

for 24-h SBP and DBP, respectively. The among-group

differences were all statistically significant (P< 0.05

ANOVA), except for the overestimation bias of SBP

between groups II and III. In Fig. 2, the overestimation

bias associated with conventional assessment of average

24-h SBP and DBP values is plotted vs. the nocturnal SBP

and DBP dips, separately for each group. In group I, the

correlation coefficient between the two sets of values was

very high (r¼ 0.88 and r¼ 0.86 for 24-h SBP and DBP,

respectively, P< 0.001), whereas in group II, it was lower

(r¼ 0.56 and r¼ 0.48 for 24-h SBP and DBP, respectively),

although still significant (P< 0.001). Conversely, in group

III, the estimation bias was unrelated to SBP and DBP

nocturnal dip. In Fig. 3, the magnitude of the 24-h average

SBP and DBP overestimation, associated with their

conventional assessment, is plotted as a function of the

orized reproduction of this article is prohibited.

Conventional vs. time-weighted BP analysis Octavio et al. 461

Fig. 1

6

4

2

0

80 100 120 140 160 180

80 100 120 140 160 180

(24hBP conventional - 24BP time weighted)/2 (mmHg)

Systolic blood pressure Diastolic blood pressure

(24-

h B

P c

onve

ntio

nal -

24-

h B

P t

ime

wei

ghte

d)/

2 (m

mH

g)

100 120 140 160 180 200

−2

6

4

2

0

0

1

2

3

806040 100 120

1008060 120 140

806040 100 120

−2

−2

−1

0

1

2

3

−2

−3

−1

0

1

2

3

−2

−3

−1

0

1

2

3

−2

−1

Bland–Altman analysis of the differences between 24-h blood pressures obtained by conventional and time-weighted analysis. Top panels: group I;intermediate panels: group II; bottom panels: group III. Left panels and black circles: 24-h SBP. Right panels and white circles: 24-h DBP. Solid linein each panel: mean value of overestimation bias. Dotted lines: 2 SD over and under mean. BP, blood pressure.

day/night ratio in BP readings number, separately for each

group. In group I, there was a weak correlation between the

two sets of values only for 24-h DBP, whereas in groups II

and III, the bias magnitude was always significantly and

consistently correlated with the day/night ratio in the

number of readings, both for SBP and DBP.

DiscussionOur study provides an important contribution to the

clinical interpretation of average 24-h ambulatory BP

data by demonstrating that the usual estimate of 24-h

SBP and DBP mean values obtained by averaging all BP

readings obtained during 24 h (conventional method)

leads to an overestimation of the 24-h BP mean levels

computed according to a time-weighted approach able to

account for the different number of values obtained

during daytime and nighttime, respectively. Our data

indeed suggest that the overestimation bias affecting

the conventional approach is at least, in part, due to

the higher number of readings per hour commonly sched-

uled during the day than at the night. Indeed, when the

ratio between the number of BP measurements per-

formed during the day and night was higher, as in our

group I, the overestimation bias associated with use of


conventional assessment of 24-h average BP levels was

also significantly higher. Conversely, when the difference

in the number of BP readings scheduled to be performed

during the daytime and the night-time was minimal, the

24-h BP overestimation bias was negligible, as observed

in our study in groups II and III.

The size of the ratio between the number of BP measure-

ments performed during the day and night depends on

two parameters, that is, on the number of readings per

hour scheduled during day and night and on the differ-

ence in the number of hours included in the day and night

subperiods. It is interesting to note that the difference in

the average 24-h BP values computed by the conven-

tional and by the time-weighted methods disappears

when ABPM contains the same number of readings

per hour along the 24 h. We have to acknowledge that

our ‘time-weighted’ method accounts for a possible

difference in the number of readings per hour during

the day and night, but it does not account for the different

number of hours between daytime wakefulness and

night-time sleep. The latter difference is related to the

‘physiological’ longer duration of the awake as compared

with the asleep periods during 24 h in daily life. Indeed,


C


Fig. 2

6

4

2

0

−2−20 −10 0 10 20 30 40 −20 −10 0 10 20 30

−10 0 10 20 30−10 0 10 20 30

40

−20 −10 0 10 20 30 40 −20 −10 0

Nocturnal blood pressure fall (%)

Systolic blood pressure

r = 0.88P < 0.001

r = 0.86P < 0.001

r = 0.48P < 0.001

r = 0.56P < 0.001

r = 0.15ns

r = 0.12ns

Diastolic blood pressure

Ove

rest

imat

ion

bia

s (m

mH

g)

10 20 30 40

3

2

1

0

−2

−1

0

1

2

3

−1

6

4

2

0

−2

3

2

1

0

−2

−1

0

1

2

3

−1

Correlation between nocturnal dip and calculated ‘bias’. Superior panels: group I; intermediate panels: group II; lower panels: group III. Left panelsand black circles: 24-h SBP. Right panels and white circles: 24-h DBP. Inserts in each panel: correlation coefficients and their statistical significance.NS, not significant.

daytime values physiologically exert a greater influence

than night-time values on the 24-h BP average level, and

this imbalance has to be accepted simply because we

usually spend more time awake than asleep.

As a corollary of the above observations, another import-

ant result of our study that deserves to be discussed is

the observed relationship between the magnitude of

the overestimation bias and the degree of nocturnal BP

fall. This relationship appears to be largely a function of

the day–night difference in the number of BP readings.

In fact, when the BP measurements are significantly more

frequent during the day than during the night, as in

groups I and II of our study, the bias associated with

the conventional calculation of 24-h BP averages

increases with the increase in magnitude of the nocturnal

‘dip’ (Fig. 2), the overestimation bias being as high as

5 mmHg in individuals with a pronounced nocturnal

BP fall.

Twenty-four-hour ABPM is increasingly used in the

diagnosis of hypertension, in the stratification of BP-

related cardiovascular risk and in the assessment of the

24-h BP coverage by antihypertensive treatment [1]. Our

finding that a pronounced difference in the number of BP


readings per hour between day and night may lead to an

overestimation of 24-h average BP by conventional

analysis is thus clinically relevant and should be taken

into account when assessing the prognostic information

carried by this parameter. It also emphasizes the possib-

ility of a clinically relevant error both in the interpretation

of 24-h ABP recordings and in the definition of 24-h

average normal reference values [8,9]. Another practical

implication comes from the observation that, due to the

limited reproducibility of nocturnal BP dipping [19], also

the overestimation of 24-h average BP levels by the

conventional approach may vary between ABP recordings

as a function of the varying difference between day and

night BP levels. Our findings need to be considered both

in clinical and research applications of ABPM because to

the best of our knowledge, with only a few exceptions

[15,20], the software used by most ABPM devices for data

processing and calculation of 24-h average BP values does

not take into account the different frequency of BP

measures scheduled during the day and night.

In conclusion, our results may have implications for

analysis of ABPM data in clinical practice. In particular,

our data indicate that, in order to reliably quantify

24-h average ABP values, ABPM devices should be


Conventional vs. time-weighted BP analysis Octavio et al. 463

Fig. 3

6

4

2

0

3 3.5 4 4.5 5 5.5 6−2

0

2 2.5 3 3.5 4

2 2.2 2.4 2.6 2.8 3

Ratio of day/night readings

Ove

rest

imat

ion

bia

s (m

mH

g)

Systolic blood pressure

r = 0.18ns

r = 0.29P < 0.001

r = 0.52P < 0.001

r = 0.44P < 0.001

r = 0.62P < 0.001

r = 0.45P < 0.001

Diastolic blood pressure

1

2

3

4

−2

−1

0

1

2

3

−2

−1

6

4

2

0

3 3.5 4 4.5 5 5.5 6−2

0

2 2.5 3 3.5 4

2 2.2 2.4 2.6 2.8 3

1

2

3

4

−2

−1

0

1

2

3

−2

−1

Correlations between total day/night ratio in number of readings and calculated ‘bias’. Superior panels: group I; intermediate panels: group II; lowerpanels: group III. Left panels and black circles: 24-h SBP. Right panels and white circles: 24-h DBP. Insertions in each panel: correlation coefficientsand their statistical significance. NS, not significant.

F ¼ s2 bet

s2 wit¼MSbet

MSwit

¼ 121:6

1:03¼ 118:08 P ¼ 0:000

Multiple comparisons – Bonferroni t-test

Comparison Difference of means P Pcrit P<0.05

1 vs. 3: 1.95�0.3¼1.65 14.080 2.403 Yes1 vs. 2: 1.95�0.5¼1.45 12.373 2.403 Yes2 vs. 3: 0.5�0.3¼0.2 1.707 2.403 No

Degrees of freedom 447.

programmed to minimize the difference in number of

readings per hour between daytime and night-time

periods. Whenever this is not done, calculation of 24-h

SBP and DBP mean values should then be based on a

‘time-weighted’ approach, that is, on estimates of 24-h

mean BP value based on the average of hourly BP means

or on other ‘weighting’ methods [21].

AddendumOverestimation bias: SBP


ANOVA

Group n Mean SD SEM

I 150 1.95 1.4 0.1143II 150 0.5 0.7 0.05715III 150 0.3 0.8 0.06532

ANOVA, analysis of variance.

Source of variation SS DF Variance estimated (MS)

Between groups 243.2 2 121.6Within groups 460.4 447 1.03Total 703.7 449

DF, degree of freedom; MS, Mean Square; SS, Sum of Squares.

Overestimation bias: DBP


ANOVA

Group n Mean SD SEM

1 150 1.5 1 0.081652 150 0.4 0.5 0.040823 150 0.2 0.5 0.04082

ANOVA, analysis of variance.

Source of variation SS DF Variance estimated (MS)

Between groups 147 2 73.5Within groups 223.5 447 0.5Total 370.5 449

DF, degree of freedom.

C


F ¼ s2 bet

s2 wit¼MSbet

MSwit

¼ 73:5

0:5¼ 147:00 P ¼ 0:000

Multiple comparisons – Bonferroni t-test

Comparison Difference of means P Pcrit P<0.05

1 vs. 3: 1.5�0.2¼1.3 15.922 2.403 Yes1 vs. 2: 1.5�0.4¼1.1 13.472 2.403 Yes2 vs. 3: 0.4�0.2¼0.2 2.449 2.403 Yes

AcknowledgementWe are in debt with Juan Perez-Gonzalez for his kind

revision of the manuscript.

References1 Mancia G, De Backer G, Dominiczak A, Cifkova R, Fagard R, Germano G,

et al. 2007 guidelines for the management of arterial hypertension. TheTask Force for the Management of Arterial Hypertension of the EuropeanSociety of Hypertension (ESH) and of the European Society of Cardiology(ESC). J Hypertens 2007; 25:1105–1187.

2 Pickering T, O’Brien E, O’Malley K. Second international consensusmeeting on twenty-four-hour ambulatory blood pressure measurement:consensus and conclusions. J Hypertens 1991; 9 (Suppl 8):S2–S6.

3 Verdecchia P, Porcellati C, Schillaci G, Borgioni C, Ciucci A, Battistelli M,et al. Ambulatory blood pressure. An independent predictor of prognosis inessential hypertension. Hypertension 1994; 24:793–801.

4 Staessen J, Thijs L, Fagard R, O’Brien E, Clement D, de Leeuw P, et al., forthe Systolic Hypertension in Europe Trial Investigators. Predictingcardiovascular risk using conventional vs ambulatory blood pressure inolder patients with systolic hypertension. JAMA 1999; 282:539–546.

5 Verdecchia P. Prognostic value of ambulatory blood pressure: currentevidence and clinical implications. Hypertension 2000; 35:844–881.

6 Kario K, Pickering TG, Matsuo T, Hoshide S, Schwartz JE, Shimada K.Stroke prognosis and abnormal nocturnal blood pressure falls in olderhypertensives. Hypertension 2001; 38:852–857.

7 Mancia G, Sega R, Bravi C, De Vito G, Valagussa F, Cesana G, et al.Ambulatory blood pressure normality: results from the PAMELA study.J Hypertens 1995; 13:1377–1390.

8 Ohkubo T, Imai Y, Tsuji I, Nagai K, Ito S, Satoh H, Hisamichi S. Referencevalues for 24-h ambulatory blood pressure monitoring based on aprognostic criterion: the Ohasama Study. Hypertension 1998; 32:255–259.

9 O’Brien E, Asmar R, Beilin L, Imai Y, Mallion JM, Mancia G, et al. EuropeanSociety of Hypertension recommendations for conventional, ambulatoryand home blood pressure measurement. J Hypertens 2003; 21:821–848.

10 Zachariah PK, Sheps SG, Ilstrup DM. Blood pressure load: a betterdeterminant of hypertension. Mayo Clin Proc 1988; 63:1085–1091.

11 Hermida RC, Fernandez JR, Mojon A, Ayala DE. Reproducibility of thehyperbaric index as a measure of blood pressure excess. Hypertension2000; 35:118–125.

12 Mancia G, Ferrari A, Gregorini L, Parati G, Pomidossi G, Bertinieri G, et al.Blood pressure and heart rate variabilities in normotensive andhypertensive human beings. Circ Res 1983; 53:96–104.

13 Schillaci G, Pasqualini L, Verdecchia P, Vaudo G. Prognostic significanceof left ventricular diastolic dysfunction in essential hypertension. J Am CollCardiol 2002; 39:2005–2011.

14 Sega R, Cesana G, Bombelli M, Grassi G, Stella M, Zanchetti A, Mancia G.Seasonal variations in home and ambulatory blood pressure in the PAMELApopulation. J Hypertens 1998; 16:1585–1592.

15 Mancia G, Zanchetti A, Agabiti Rosei EG, De Cesaris R, Fogari R, PessinaA, et al. Ambulatory blood pressure is superior to clinic blood pressure inpredicting treatment-induced regression of left ventricular hypertrophy.Circulation 1997; 95:1464–1470.

16 Rodrıguez-Roa E, Octavio JA, Mayorca E, Castro P, Miranda R, Valecillos E,Gonzalez M. Blood pressure response in 24 h in patients with high bloodpressure treated with two nifedipine formulations once a day. J HumHypertens 2002; 16 (Suppl 1):S151–S155.

17 Schettini C, Bianchi M, Nieto F, Sandoya E, Senra H. Ambulatory bloodpressure normality and comparison with other measurements.Hypertension 1999; 34 (Pt 2):818–825.

Degrees of freedom 447.


18 Kario K, Schwartz J, Pickering T. Ambulatory physical activity as adeterminant of diurnal blood pressure variation. Hypertension 1999;34:685–691.

19 Omboni S, Parati G, Palatini P, Vanasia A, Muiesan ML, Cuspidi C, et al.Reproducibility and clinical value of nocturnal hypotension: prospectiveevidence from the SAMPLE study. J Hypertens 1998; 16:733–738.

20 Hansen TW, Jeppesen J, Rasmussen S, Ibsen H, Torp-Pedersen C.Ambulatory blood pressure and mortality: a population-based study.Hypertension 2005; 45:499–504.

21 Stanton A, Cox J, Atkins N, O’Malley K, O’Brien E. Cumulative sums inquantifying circadian blood pressure patterns. Hypertension 1992;19:93–101.


Documents

TIME-WEIGHTED VS CONVENTIONAL QUANTIFICATION OF 24 HOUR AVERAGE AMBULATORY SYSTOLIC BLOOD PRESSURE