How Well Does a Shortened Time Interval Characterize Results of a Full Ambulatory Blood Pressure Monitoring Session?

Abstract

Ambulatory blood pressure monitoring (ABPM) is useful in evaluating cardiovascular risk but requires significant time. The authors examined how closely shortened time intervals correlate with the systolic blood pressure (BP) determined from a full 24‐hour ABPM session in 1004 ABPM recordings. After excluding the first hour, Pearson correlations performed for the mean systolic BP of the subsequent 3‐, 5‐, and 7‐hour periods (4, 6, and 8 hours total) with the entire, and remainder of the session, demonstrated greatest improvement in correlation when the session is increased from 4 to 6 hours. Bland‐Altman analysis of the 6‐hour time period revealed a mean difference of 5.41 mm Hg compared with the full session mean. The authors conclude that 6‐hour ABPM can approximate the overall mean BP obtained from full 24‐hour ABPM. However, shortened sessions do not characterize the influence of circadian variation on the 24‐hour mean BP and may overestimate the 24‐hour BP levels.

Twenty‐four‐hour ambulatory blood pressure monitoring (ABPM) is recognized as a more accurate measure of a patient's blood pressure (BP) than office‐based BP readings. Several assessments of important prognostic significance can be obtained from ABPM more readily than office BP measurement, including overall mean BP, assessment of circadian rhythm, and BP variability. ¹ Ambulatory BP values, as well as those obtained from repeated home BP monitoring, appear to correlate more closely with prediction of clinical outcomes than does office BP and are valuable in estimating the cardiovascular risk profile of an individual patient. ABPM is suggested to be of use in several situations, including assessment of the white‐coat effect, treatment resistance, masked hypertension, labile BP, and orthostasis. ²

Unfortunately, there are disadvantages to ABPM as compared with office or home recordings that prevent it from being widely used in nonspecialist practices. These disadvantages include expense, limited insurance reimbursement, as well as the requirement of a 24‐hour commitment by patients to wear the monitor during both waking and sleeping hours. Although patient discomfort and inconvenience appear minimal, they have been well documented. ³ Abbreviated ABPM sessions, if representative of the results of the entire session, could result in similar clinical decision making and could be more practical by reducing expense and intrusiveness in patients' daily lives. Previous studies have suggested that a limited 6‐hour ABPM session may accurately replicate the daytime mean BP. ⁴ , ⁵ , ⁶ The purpose of our study was to determine whether a shortened time period of ABPM could accurately reflect the overall mean systolic BP as determined from a full 24‐hour ABPM session in a large heterogeneous sample of ABPM studies.

METHODS

Data Source

ABPM data for this study were obtained from a database of all ABPM studies performed by the ABPM referral clinic in the Family Care Center of the University of Iowa Hospitals and Clinics. Sessions included for the analysis were conducted from the inception of the clinic in January 2001 through June 2007 (N=569). The ABPM referral clinic is codirected by a board‐certified family physician and board‐certified pharmacotherapist and processes referrals ordered for the usual clinical indications for ABPM that are encountered during routine clinical practice. In addition, ABPM studies performed as part of 2 clinical hypertension management study protocols conducted by our research group were also available for analysis. ⁷ In both studies, patients received 24‐hour ABPM at baseline.

A total of 1348 ABPM sessions constituted the original datasets available for this analysis. From this number, we excluded any repeat ABPM studies in individual patients and from the research studies (39/569) and 1 additional entry that contained no ambulatory monitoring data. The Final merged dataset included 1004 ABPM sessions for analysis.

All ABPM studies were performed using the SpaceLabs 90217‐A Ultralite monitors (SpaceLabs Medical, Inc, Redmond, WA) and processed using SpaceLabs ABP Report Management System software (v 1.03.11 and v 2.00.06). Although no formal validation protocol was used within individual patients undergoing ABPM, a standard office‐based sphygmomanometer reading was taken at the time of ABPM placement in more than two‐thirds of the sessions.

Monitors were programmed to record BP every 20 minutes during daytime hours (0600 to either 1800, 2000, or 2200 hours) and every 30 minutes during nighttime hours (either 1800, 2000, or 2200 hours to 0600 hours), yielding approximately 3 readings/h during the daytime and 2 readings/h during nighttime periods. All patients undergoing ABPM were educated about the procedure in a standardized manner and fitted with the monitor during a regular workday. Four different cuff sizes were available for use, and the appropriate cuff size was selected based on arm circumference as recommended in BP measurement guidelines. ⁸ Start and stop times were not uniformly standardized for the patients and were scheduled according to monitor availability and patient convenience.

Statistical Analysis

Random effects models using maximum likelihood estimation techniques were used to study the effects of monitor duration adjusting for time of day (and vice versa), while accommodating the within‐patient correlation of the repeated measures. The first hour of each ABPM session (the “white‐coat window”) ⁹ was excluded from analysis, as the mean systolic BP difference between the first hour and the remaining hours was 2.79 mm Hg higher (P<.001). After exclusion of the first hour, shortened intervals of the subsequent 3, 5, and 7 hours (which correspond to 4‐, 6‐, and 8‐hour total sessions) were considered in the analysis. Partial Pearson correlations, adjusted for the time of day when the patients started their sessions, were calculated to compare the mean systolic BP obtained from a shortened time interval with the mean systolic BP obtained from the entire ABPM session period (eg, 2nd–4th vs 2nd–24th hour) and from the remainder of the ABPM session (eg, 2nd–4th vs 5th–24th hour). In addition, concordance limits between the shortened time interval and the entire time period were calculated using the Bland‐Altman method. ¹⁰

All data were analyzed at the individual patient level before describing means for the sample as a whole. To describe the mean systolic BP levels of the daytime and nighttime periods, all of the observations were first stratified by daytime and nighttime, and then the mean for each patient for these periods was determined. The mean of those means was then calculated to represent the entire sample. Daytime was defined as 0600 to 1800 hours and nighttime was defined as 1800 to 0600 hours. Means for the 24‐hour and the shortened time periods were calculated in a similar manner.

RESULTS

The data consisted of 22,008 hours of BP monitoring with an average of 21.9 h/participant. A mean ± SD of 51.5±11.6 total readings per participant were obtained, with an average of 85%±16.6 of the attempted readings resulting in a successful BP assessment. The average age of patients undergoing ABPM was 55.7±15.3 years; 54% were female. In the patients who had ABPM performed in the ABPM referral clinic, the clinical indication for ordering the ABPM session were as follows: borderline hypertension, 22.3%; evaluation of BP control on current antihypertensive medication, 27.9%; suspected white‐coat hypertension, 25.8%; treatment resistance, 12.1%; suspected hypotension, 2.1%; masked hypertension, 1.5%; and unknown reasons, 8.3%. All patients who had ABPM performed at the beginning of the 2 hypertension clinical studies had uncontrolled hypertension per standardized BP measurement in the clinic.

Figure 1 portrays the systolic BP means plotted against time worn, while adjusting for the time of day. As mentioned previously, the first hour (white‐coat window) of each ABPM session was excluded from analysis; the mean systolic BP between the first hour and the remaining hours was 2.79 mm Hg higher, even after controlling for time of day (P<.001). Substantial time‐of‐day trends were noted for systolic BP, with average levels staying <125 mm Hg from midnight to 0400 hours, then increasing to mean levels that approached 140 mm Hg for the time period of 0700 to 1800 hours, then decreasing toward 125 mm Hg (Figure 2). The means ± SD of the daytime, nighttime, and 24‐hour periods were 138.78±14.57 mm Hg, 127.83±16.76 mm Hg, and 133.38±14.74 mm Hg, respectively.

Figure 1.

Systolic blood pressure (SBP) means plotted against time worn and adjusted by time of day.

Figure 2.

Means of systolic blood pressure (SBP) plotted against time of day, adjusted by time worn (first hour excluded).

The correlations for the mean systolic BP from the shortened time periods and the mean systolic BP derived from the entire time period and from the remaining time periods are shown in Table I. Incremental improvement in the correlations appears greatest when increasing the monitoring duration from 4 to 6 hours, achieving a correlation coefficient of 0.828 for the entire time period and of 0.710 for the remainder time period. Minimal differences were observed between unadjusted correlations and the correlations adjusted for the time of day when monitoring began.

Table I.

Correlation of Mean Systolic BP Values From Shortened Time Period Compared With Entire Time Period and Remaining Time Period

Shortened Time Period	Mean±SD Systolic BP, mm Hg	Entire Time Period a (Mean±SD Systolic BP = 133.38±14.74 mm Hg)		Remaining Time Period b
		Unadjusted Correlation	Adjusted Correlation c	Unadjusted Correlation	Adjusted Correlation c
2nd–4th Hour	139.32±15.55	0.764	0.765	0.684	0.685
2nd–6th Hour	138.80±14.74	0.828	0.83	0.710	0.711
2nd–8th Hour	138.58±15.38	0.859	0.868	0.695	0.712
Abbreviation: BP, blood pressure. P<.001 for all correlations. a2nd–24th hour. beg, 5th–24th hour for row 1, 7th–24th hour for row 2, and 9th–24th hour for row 3. cAdjustment is for time of day when monitoring began.

Although results of the correlation analysis suggest substantial correlation between the shortened time period and full session, a Bland‐Altman plot of the 6‐hour window (Figure 3) indicated much less precision in the agreement between the mean systolic BP of the shortened time period and the full 24‐hour session. Differences in the means ranging from 5.91 to 5.21 mm Hg were observed with the shortened time periods compared with the full 24‐hour session (Table II). Consistent with the correlation analysis, prolonging the window for an additional 2 hours did not appear to result in substantial improvement in concordance with the full 24‐hour session.

Figure 3.

Bland‐Altman plot for the agreement of the 6‐hour shortened‐window systolic blood pressure with the 24‐hour mean systolic blood pressure.

Table II.

Statistics for Bland‐Altman Analysis

	Shortened Time Period
	2–4 hours	2–6 hours	2–8 hours
Mean±SD of difference (shortened vs entire)	5.91±10.43	5.41±8.82	5.21±8.02
95% Confidence interval for mean	5.25–6.57	4.75–6.07	4.70–5.72
Precision (mean±1SD)	−4.52±16.34	−3.41±14.22	−2.81±13.23
Limits of agreement (mean±2SD)	−14.95±26.77	−12.22±23.04	−10.84±21.26

DISCUSSION

The ability to accurately predict the average BP of a patient with the fewest number of BP readings is advantageous to clinicians and patients alike. Historically, this has been the fundamental principle behind office‐based sphygmomanometry, in which it is assumed that the BP reading obtained in the office reflects the patient's overall BP status. Office BP is, after all, the gold standard on which interventions to lower BP have been based and on which reductions in cardiovascular outcomes are observed. The higher the office BP, the greater the risk; the lower the treatment BP, the lower the risk. However, use of office‐based BP has numerous limitations, including observer bias, terminal digit preference, and inability to detect the white‐coat response; 24‐hour ABPM and frequently repeated home BP recordings may avoid these problems. ⁸ Unfortunately, ABPM is the more costly alternative, and it is time‐consuming and potentially uncomfortable and intrusive to patients' daily lives because it must be performed during usual daily activities and while sleeping.

Interest in short‐term alternatives to 24‐hour monitoring to represent the whole‐day BP was first reported in 1982. ¹¹ In this study of 6 patients, average BP values recorded every 7.5 minutes from 0800 to 1000 hours correlated closely with the full 24‐hour average, although they were higher than the overall 24‐hour average. In a follow‐up study in 50 patients, 11 of 50 with office hypertension were normotensive by the 0800 to 1000 average, which was confirmed in the full 24‐hour average, suggesting that a limited duration of automatic readings can help avoid misclassification of patients' hypertensive status. ¹²

Additional studies have suggested that shorter time periods of ABPM can be predictive of the daytime mean BP. Chanudet and associates ⁴ evaluated different time spans within a 24‐hour period in 354 men with normal to slightly elevated BP and found that a mean of 4 readings/h during any 4‐hour period between 1000 and 2200 hours correlated with mean daytime systolic BP from 24‐hour ABPM. In a retrospective study by Sheps and colleagues, ⁵ mathematic modeling was used to determine that most of the predicted ability of 24‐hour ABPM was obtained with 6 hours of monitoring. This finding was replicated in a prospective follow‐up to their study. ⁶

Our study found that a shortened time interval of 6 to 8 hours' duration can approximate the mean systolic BP obtained from a whole‐day ABPM session; there is, however, some difference. The Bland‐Altman analysis indicates that if a shortened time period is selected, the mean difference compared with the 24‐hour average systolic BP is approximately 5 to 6 mm Hg. This finding is not unexpected considering that most patients begin ABPM in the morning and the shortened time periods evaluated occurred during daytime hours; thus, the mean of these periods are not influenced by the circadian variation in BP observed throughout a 24‐hour period and does not include the nocturnal decrease in BP. Based on the calculated means for the shortened time periods and the daytime period, our results confirm that the shortened time period may reflect the daytime BP. A downward adjustment factor may be necessary if this mean is to be extrapolated to a 24‐hour average. Depending on what information the clinician is seeking from the test, the differences observed between the shortened time period and full 24‐hour period may or may not be acceptable for clinical decision making.

To our knowledge, our study is the largest sample from which the question of utility of abbreviated ABPM sessions has been evaluated. Unlike other studies, it is also heterogeneous with regard to whether patients were treated or untreated while undergoing ABPM. Also unique is that we examined the relationship between the shortened time period and the 24‐hour mean as opposed to the daytime mean. Future studies can develop more specific predictive models for the shortened time period that take into account additional factors such as the indication for ordering the test, the frequency and duration of readings, and exploration of when the optimal shortened time period windows with greatest predictability for the 24‐hour mean occur. Information such as this could be used to make a recommendation for beginning the recording at a specific time or an indication‐specific recommendation for duration and frequency of monitoring.

Our study findings should be interpreted within the context of an important limitation. That is, collapsing of the 24‐hour data into a shortened interval results in a loss of ability to evaluate specific circadian variation; nighttime dipping for example, can be important for evaluating individual patient cardiovascular risks. ¹³ If a model of short‐term ABPM is developed as an acceptable substitute for 24‐hour ABPM, additional studies will be necessary to evaluate whether data obtained from such a shortened time interval correlate with target organ damage and clinical outcomes in the same manner as the full 24‐hour recording. Second, we chose to examine duration of the recording session as opposed to number of readings. Duration was chosen because there is no standard protocol for the frequency of readings with ABPM, and we felt analysis of session duration would result in more generalizable findings than would frequency of readings. It is possible that by reprogramming the frequency of readings, the predictive ability of a shortened time with the entire session period could be improved. However, a recent prospective study using Bland‐Altman analysis suggests the reproducibility of the mean BP depends more on measurement duration than frequency. ¹⁴

It has been argued by some that evaluation of BP using ABPM should be performed in all patients with hypertension newly diagnosed by means of office BP, both from an individual risk prediction standpoint and from a cost‐effectiveness standpoint to the population. ¹ , ¹⁵ However, the test may be expensive (as much as $350–$500 per session) for the individual patient, as it is not routinely covered by insurance for many indications. Our results suggest that in situations in which the overall mean BP is the primary question, limited ABPM has the potential to approximate 24‐hour monitoring, but it must also be recognized that recurrent home BP recordings over time may be just as effective in predicting risk and progress—at less cost and inconvenience. Home BP values, however, will not determine nocturnal BP changes. Additional benefits of home BP readings include obtaining a series of BP levels over time rather than one 24‐ or even 48‐hour recording. Future data‐driven studies can seek to develop better predictive models and will require testing and validation in prospectively designed studies. The result of such research on shortened ABPM could be a less costly alternative that also minimizes patient discomfort and inconvenience and perhaps leads to increased use of ABPM in routine clinical practice.

CONCLUSIONS

A shortened ABPM time period of 6 to 8 hours may potentially approximate the patient's overall BP obtained from a full 24‐hour session, but it may result in overestimation of the true 24‐hour average. It is important to note that shortened sessions will not characterize circadian dipping patterns, and a full 24‐hour session may be preferred to evaluate an overall individual patient's cardiovascular risk; specific advantages of a shortened ABPM time over repeated home BP monitoring remain to be determined. Further research should be undertaken to evaluate the predictive value of shorter ABPM sessions with target organ damage and clinical outcomes.

Disclosure:

This work was funded in part by the National Heart, Lung, and Blood Institute (1 RO1 HL069801‐01A1). This research was presented as a scientific poster at the 22nd Annual Meeting of the American Society of Hypertension, Chicago, IL, May 2007.