Statistical Base Value of 24-Hour Blood Pressure Distribution in Patients With Essential Hypertension
Abstract—The purpose of this study was to calculate statistically the minimum (base) blood pressure (BP) of nighttime (sleep-time) BP values obtained by ambulatory BP monitoring (ABPM) and to investigate its clinical significance. Twenty-four-hour recording of ECG with ABPM was performed directly (n=89) or indirectly (n=117) in 206 patients with essential hypertension. A telemeter was used for the direct method and a multi-biomedical recorder (TM2425) was used for indirect measurement. First, minimum heart rate (HR0=60/RR0) was determined from sleep-time ECG. The mean product of sleep-time diastolic BP (DBP) and pulse interval (RR) was divided by RR0 to obtain DBP0 [DBP0=(DBP×RR)s/RR0]. The correlation between systolic BP (SBP) and DBP was used to determine SBP0 corresponding to DBP0. Statistical base mean BP (MBP0) was calculated from these values, and its reproducibility and relation to hypertension severity were investigated. MBP0 values were similar to true base values of sleep-time MBP obtained by the direct method (mean±SD difference, 2.0±4.2 mm Hg). Direct MBP0 criteria predicted hypertension severity (mild, moderate, or severe target organ damage) more accurately (predictive accuracy, 89%) than daytime MBP criteria (53%, P<0.01). Almost the same results were obtained using indirect MBP0 criteria. Day-to-day indirect MBP0 variation (mean absolute difference) was smaller (2.4±1.8 mm Hg) than day-to-day daytime and nighttime MBP variation (6.3±5.3 and 5.4±3.4 mm Hg, respectively; n=61, P<0.01), and the correlation coefficient between day-to-day variations of daytime MBP and physical activity (measured by an acceleration sensor) was 0.38 (P<0.05). In conclusion, statistical base BP was almost equal to true base (minimum) BP of sleep-time BP distribution. It was closely related to the severity of hypertensive organ damage, was highly reproducible, and is considered likely to serve stochastically and physiologically as a representative BP value in an individual subject.
The enormous fluctuations occurring in blood pressure (BP) over a 24-hour period can involve as many as 100 000 directly measured intra-arterial BP values and ≈100 measurements made indirectly by ambulatory BP monitoring (ABPM). A rational statistical way of treating these large numbers of values is fundamental to BP evaluation. Stochastically, each office BP or indirect ABPM value is only a sample of the 24-hour direct BP values occurring in an individual subject. Consequently, determining which of the measurement values to adopt as a representative BP value for an individual subject requires examination of the extent to which each value fulfills the following essential conditions.
First, the value must be logically sound, stochastically as well as physiologically. This means it must be a statistical parameter with a physiological basis. Second, to be clinically useful and applicable, it must be closely related to hypertension severity (organ damage). Third, it must be able to be measured noninvasively in clinical practice. Fourth, its measurement values must be highly reproducible and must demonstrate only slight day-to-day variation. Fifth, ideally, it should evaluate the prognosis and risk of hypertensive vascular complications.
Sir F. Horace Smirk1 reported that basal BP is more intimately related to hypertension severity than casual BP. Applying the same line of thought to 24-hour intra-arterial BP values, we reported base BP (that is, minimum BP occurring during nighttime sleep) to be a physiologically and stochastically important BP value.2 3 4 5 6 7 However, because base BP can be measured only with a large number of samples by the direct method, we found it necessary to devise a method of calculating statistical base BP with a small number of samples obtained indirectly by means of ABPM, and we have investigated the extent to which values obtained in this way satisfy the first 4 of the conditions set forth above.
This investigation was carried out in 206 patients (107 men and 99 women between 30 and 74 years of age) with hypertension (office systolic BP [SBP] >140 mm Hg and diastolic BP [DBP] >90 mm Hg on at least 3 visits) who were taking no medication or for whom medication had been withdrawn at least 2 weeks previously. These patients satisfied the following inclusion criteria: no valvular defect, myocardial infarction, or secondary cause of hypertension. They were subdivided into 3 groups (Table 1⇓). Group 1 (mild group with World Health Organization [WHO] stage I) consisted of subjects with no objective signs of hypertensive organic change.8 Their chest x-rays showed a cardiothoracic ratio (CTR) of <50%. The sum of SV1 and RV5 on their ECG was <4.0 mV, the echocardiographic left ventricular mass (LVM) index was <134 g/m2 for male subjects or 110 g/m2 for female subjects,9 and their Keith-Wagner-Barker (KW) ophthalmic category was lower than grade 1. They demonstrated no cerebral or renal hypertensive complications (plasma creatinine level was <91.5 μmol/L). Group 3 (severe group) consisted of patients with accelerated hypertension whose KW ophthalmic category was grade 3 or whose plasma creatinine level was >152 μmol/L. Group 2 (WHO stage II) was intermediate between these 2 (moderate group with LV hypertrophy and without retinopathy, plasma creatinine <152 μmol/L). Table 1⇓ shows numbers of subjects and gender, age, mean office BP, and main clinical characteristics for each group.
Methods consisted of direct and indirect ABPM measurements. Direct measurements were made in 89 patients (47 men and 42 women; 37 from group 1, 38 from group 2, and 14 from group 3) after hospitalization. Indirect BP measurements were made in 117 patients (60 men and 57 women; 52 from group 1, 52 from group 2, and 13 from group 3) at the outpatient clinic. All subjects gave informed consent to participation in the study, and the study protocol was approved by the ethics committee of the Yokohama City University Department of Internal Medicine.
Direct BP Measurement
Although it has been reported previously,6 10 we here briefly outline the procedure whereby direct BP measurements were made. Using a system developed in our laboratory, we used the telemetric method to perform continuous 24-hour monitoring of intra-arterial pressure and ECG under almost unrestricted conditions. With patients under local anesthesia, a stiff-walled Teflon catheter was inserted into the left brachial artery and connected to a strain-gauge transducer (Statham P50, Gould Statham Instruments Inc) attached to the chest wall at the level of the heart. BP calibrations were performed with a sphygmomanometer at 0, 100, and 200 mm Hg both before and after measurement. To prevent coagulation, a portable microinfusion pump (NEC Sanei Instruments Ltd) continuously flushed the catheter with heparinized saline solution. A V5 lead was used to record ECG waves, which, together with BP waveform, were telemetrically transmitted by means of a portable transmitter (model 1429, NEC Sanei Instruments Ltd). Both paper and tape recordings were collected (CR-31 portable tape recorder, TEAC Co). An analog-to-digital converter was used to input BP and ECG waves into a computer at a sampling rate of 1 kHz for subsequent computer processing to determine beat-to-beat heart rate (HR), SBP, and DBP.
In accordance with the hospital schedule, sleep-time (nighttime) was defined as the hours between 9 pm and 6 am. The remainder of the 24-hour period was regarded as waking time (daytime). The 24-hour trendgrams and sleep-time frequency histograms of BP and HR were made by computer (Figure 1A⇓ and 1B⇓).
Measurement of True Base BP Values From Direct Recordings
The minimum values (values corresponding to the 0.5% lower probability integral during 1 hour) of intra-arterial SBP and DBP were almost the same as baseline values during sleep-time2 (Figure 1A⇑). Sleep-time frequency distributions for intra-arterial SBP, DBP, and HR were compiled; values corresponding to the 0.5% lower probability integral were taken as true base SBP and DBP2 3 (Figure 1B⇑).
Measurement of Estimated Base BP Values From Direct Recordings
Because true base BP values can be calculated only on the basis of large numbers of directly obtained samples (about 30 000 samples), we devised a new method for determining statistical (estimated) base BP values from a small number of either directly or indirectly obtained samples. HR can be measured noninvasively by ECG, which makes it possible to derive the minimum value (lowest 1% value; HR0) from the HR frequency histogram obtained during sleep-time. This determines the corresponding pulse interval (RR) or RR0 (RR0=60/HR0 is the lowest 1% value of HR histogram during sleep-time) (Figure 2⇓). Base value DBP×RR (product of DBP and RR) was almost equal to mean sleep-time DBP×RR and mean waking-time DBP×RR.4 Using this value, we calculated the mean of DBP×RR during sleep-time: (DBP×RR)s. Then, dividing this by RR0, we obtained the statistical base DBP (DBP0): Because of the high positive coefficient of correlation (r) between direct SBP and DBP (r=0.70 to 0.98),11 we used the linear-regression formula SBP=a×DBP+b (where a is the regression coefficient and b is constant value) with sleep-time values to obtain the statistical base SBP (SBP0) corresponding to DBP0: Next, statistical base mean BP (MBP0) was calculated from DBP0 and SBP0 [MBP0=(SBP0−DBP0)/3+DBP0]. This method for obtaining statistical base BP values is applicable to BP values obtained both directly and indirectly (Figure 2A⇓ and 2B⇓). In this study, its accuracy was evaluated first using directly obtained BP values and then with indirect BP measurements.
Indirect BP Measurement
The portable multi-biomedical recorder (TM2425, A&D Co Ltd) can simultaneously record 24-hour indirect BP and ECG and measure body position, motion (acceleration), and temperature. The BP accuracy of TM2425 was validated according to and satisfied the criteria of the Association for the Advancement of Medical Instrumentation.12 13 Body motion (activity) was measured by an acceleration pickup sensor, and cumulative values for 1 minute were recorded at 18-millisecond intervals at frequencies ranging from 1 to 10 Hz in the vertical direction, with a sensitivity of 4.1×10−3 m/s2.12 SBP and DBP were recorded at 30-minute intervals, and HRs were recorded beat-to-beat on ECG by TM2425. The process of obtaining indirect statistical base BP by computer is as follows (Figure 2B⇑). First, a sleep-time HR distribution was compiled from the ECG obtained with TM2425. Arrhythmic beats were automatically excluded.12 HR frequency distribution, mean, SD, skewness (sk), and minimum (lowest 1%) value were calculated by computer, and HR0 and RR0 (60/HR0) were derived. After indirect DBP0 was obtained by means of Equation 1, Equation 2 was used to arrive at indirect SBP0 (Figure 2B⇑). The coefficient of correlation (r) between sleep-time SBP and DBP was often smaller than the corresponding value obtained with the direct method, but Equation 2 was applied only when r≥0.6 (n=82). In instances in which r<0.6 (n=35), the following equations (arrived at stochastically from least-squares linear regression analysis performed on 82 patients in whom correlation between SBP and DBP was r>0.6 were used for estimating SBP0): SBP0=DBP0×[(SBP/DBP)s+0.05] and (SBP/DBP)s=mean SBP/DBP ratio during sleep-time. The difference between SBP0 derived from these equations and SBP0 derived from Equation 2 was 1.5±4.9 mm Hg in the 82 patients. The indirect measurements were performed during the subjects’ daily routines, and sleep-times were determined from their diaries.
With the aim of investigating the reproducibility of statistical base BP, 24-hour measurements were performed twice (Figure 3⇓) using TM2425 for 61 patients (31 men and 30 women). The second measurement was made more than 1 week after the first (median interval of 3 weeks; range, 1 week to 2 months).
Because LV hypertrophy is related to hypertension severity, echocardiography was used to examine LVM. All subjects underwent standard M-mode and 2-dimensional echocardiography performed by a cardiologist using an echocardiograph equipped with a 2.5-MHz and a 3.6-MHz imaging transducer (SONOS 2500 ultrasound system, Hewlett-Packard Co). As recommended by the American Society of Echocardiography,14 LV dimensions were derived from 2-dimensionally guided M-mode tracings. LVM was calculated using the Penn convention15 and adjusted for body surface area (LVM index). Echocardiograms were read in a blinded fashion without knowledge of other findings.
Standard statistical methods, including paired 2-sample t test, nonpaired t test, F test, χ2 test, sensitivity test, specificity test,16 and ANOVA were used. Least-squares linear regression analysis and Bland-Altman plotting method were performed for 2 variables. MBP criteria (cutoff point) were determined by discriminant analysis17 to discriminate well between the groups [discriminant functions: Z12=a12MBP+b12, Z23=a23MBP+b23; Z12 (a12 and b12) and Z23 (a23 and b23) were determined to allocate a patient to group 2 if Z12 >0 and to group 3 if Z23 >0]. Predictive accuracy was calculated as follows: (n1+n2+n3)/total number; n1, n2, and n3 were numbers of patients classified accurately into each group (1, 2, and 3) by MBP criteria. Values were expressed as mean±SD, and values of P<0.05 were considered significant. The Multiple Statistical Analysis Program (Social Survey Research Information Co, Ltd) was used for calculations.
Comparison Between True Base BP and Statistical Base BP From Direct Recordings
The correlation between the true base DBP and statistical base DBP (direct DBP0) by the direct method was high (r=0.93). The mean difference between them was 2.6±4.8 mm Hg. A high coefficient of correlation (r=0.93) was also found between the true base SBP and statistical base SBP (direct SBP0). The difference between them was 1.0±6.0 mm Hg. MBP calculated from direct DBP0 and SBP0 was taken as direct MBP0. The difference between true base MBP and direct MBP0 assessed by paired t test and Bland-Altman plotting was not significant, although MBP0 tended to be higher than true base MBP (mean difference, 2.0±4.2 mm Hg) (Figure 4⇓).
Statistical Base BP and Hypertensive Target Organ Damage
Waking-time mean MBP was obtained and designated MBPW. Investigation was then performed to determine whether statistical base MBP or MBPW is more closely related to hypertension (organ damage) severity (Figure 5⇓, Table 2⇓). MBP criteria (MBPW=110 to 129 mm Hg; MBP0=89 to 109 mm Hg) were established to discriminate well between the groups by discriminant analysis (Table 2⇓).
Sensitivity and specificity for direct MBP0 criteria were almost equal to or higher than those for direct MBPW criteria (Figure 5⇑, Table 2⇑). The predictive accuracy (sum of accurately classified number/total number×100) of direct MBP0 criteria (79/89×100=89%) was significantly higher than that of direct MBPW criteria (47/89×100=53%). These results suggest that direct MBP0 criteria predict hypertension (target organ) severity more precisely than direct MBPW criteria.
In this case, MBPW is expressed as a function of φ×MBP0 (φ=MBPW/MBP0),5 and it becomes essential to know whether φ is independently related to hypertension severity. The absence of a significant difference in φ values among the 3 groups (Figure 5⇑) indicates that φ contains no information related to hypertension severity. A negative coefficient of correlation (r=−0.43, P<0.01) was observed between direct MBP0 and φ (Figure 5⇑, bottom).
Almost the same results were obtained from indirect measurements (Figure 6⇓, Table 2⇑). There was a high correlation between indirect MBP0 and echocardiographic LVM index (r=0.75, P<0.01; Figure 6⇓). The correlation coefficient (r) between indirect MBP0 and indirect MBPW was 0.72, and r between indirect MBPW and LVM index was 0.61.
Mean physical activity (acceleration) during waking hours (ACTw) in group 3 patients was less than in groups 2 and group 1 (8.4±2.4: 14.2±6.8: 18.7±6.4×10−3 m/s2 per minute; P<0.05).
Day-to-Day Variation of Statistical Base BP
To investigate the reproducibility, we used a multi-biomedical recorder (TM2425) to observe daily (day-to-day) variation (mean absolute difference between values obtained in 2 measurements) in indirect MBP0 and MBPW (Figure 7⇓). Day-to-day indirect MBP0 variation was significantly less (2.4±1.8 mm Hg) than daily indirect MBPW variation (6.3±5.3 mm Hg, P<0.001). Furthermore, day-to-day indirect MBP0 variation was significantly less than daily nighttime indirect MBP variation (5.4±3.4 mm Hg, P<0.01).
There was a significant correlation (r=0.38, P<0.05) between day-to-day MBPW variation (first-day MBPW− second-day MBPW) and physical activity variation (first-day ACTw−second-day ACTw, where ACTw is mean acceleration during waking hours).
Because of the great influence of daily activity and duration of sleep, the mean of nighttime BP values and the day-night differences demonstrate low reproducibility.18 19 20 To compensate for this fault, we propose statistical base BP as a new index representative of nocturnal sleep-time BP. This study investigates the extent to which this new index serves as an individual representative BP value satisfying the conditions set forth in the introduction.
Sir F.H. Smirk first stated that for purposes such as measuring basal metabolism, basal BP was more closely related to hypertensive severity and mortality than casual BP, but supplemental BP (casual BP minus basal BP) was not related to them.1 Applying this concept to ABPM measured values, we have specified minimum sleep-time BP (base BP) and have repeatedly studied its significance.2 3 4 5 6 7 Instead of manifesting a gaussian distribution, intra-arterial BP and HR values obtained during nighttime and daytime are distributed asymmetrically with a levopositional mode (mode<mean) in a manner approximating a gamma distribution that has a minimum value (location parameter).8 9 We found that base BP was almost the same as the location (L) parameter of the BP frequency distribution, which approximated a gamma distribution.2 3 The L parameters took approximately the same values for both nocturnal sleep-time and waking-time BP distributions and seemed to be the starting point for their gamma distributions.3 True base BP corresponds to an actual measured value of the L parameter,2 which is an important parameter of these BP distributions.
BP reaches a minimum value during the electroencephalographic slow-wave phase.3 6 21 The base BP waveform manifesting itself at this time takes the most efficient pulse interval (RR) for producing a maximum per-minute waveform area, which indicates a maximum per-minute blood flow in relation to BP value.5 Moreover, at this time, base BP demonstrates a higher correlation with total peripheral vascular resistance and arterial elastic modulus than MBPW.5 DBP0 and SBP0, which are statistical base BP values, are the estimated values for the true base BP values of each of their frequency distributions and are thought to satisfy the stochastical and logical conditions set forth in the introduction.
Every BPi (i=1 … n) during daytime and nighttime may be expressed by the formula BPi=φi×base BP (φi=incre-mental ratio). The φi value, which can be analyzed by means of the formula φi≒FI+BI (FI indicates cardiovascular function index; BI, baroreflex function index),5 7 may vary because of the changes of cardiovascular and baroreflex responses to physical and mental activity. Because φ (MBPW/MBP0) is unrelated to hypertension (target organ) severity and φ correlates with physical activity (group 3 patients had lower φ and lower physical activity), using MBP0 is considered more logical than using MBPW (φ×MBP0) for evaluation of hypertension severity. Figure 3⇑ shows typical variations of φ value and indirect MBPW caused by changes in physical activity (acceleration) monitored in 1 patient with TM2425. Although the means of waking-time BP and nighttime BP are also statistical parameters, because they are greatly influenced by daily activities, their φ values may be changed by alterations in such factors as external temperature, insufficient sleep, duration of sleep, physical activity (acceleration), emotional stress, and autonomic nervous activity.12 18 19
Indirect MBP0 values demonstrated less day-to-day variation than the indirect MBPW (φ×MBP0) and nighttime mean MBP values (Figure 7⇑). This means that daily activities and sleep duration may exert no influence on statistical base BP. It is therefore more logical to use it instead of mean nighttime BP as a value representative of sleep-time BP.
The circadian change in BP in human subjects is well recognized, with BP levels being lower at night, and this observation has led to the perception of hypertensive subjects as dippers (with a nocturnal BP fall or dip) and nondippers.22 23 There are many reports that nighttime BP is closely related to end-organ damage, which is more severe in nondippers than dippers.24 25 However, the reproducibility of BP dipping (nighttime BP/daytime BP) is poor.19 We suggest that the statistical base BP, instead of mean nighttime BP, may be useful for evaluating the extent of the dipping phenomenon (dipping corresponds to φ value).
It is true that we should compare indirectly measured statistical base BP with true base BP obtained by means of the direct method, but such an investigation would be very difficult because of the great burden imposed on subjects by the need for simultaneous 24-hour measurement of both direct and indirect BP. In addition, there are ethical reasons for rejecting such a practice. Our inability to perform this kind of study prevents our being able to conclude yet whether statistical indirectly measured base BP is consistent with true base BP value obtained with the direct method.
Nonetheless, statistical base BP is considered significant as a BP value that can serve as a normalized BP value free of the effects of environmental conditions. It reflects true basal BP, and because it approximately fulfills the first 4 of the 5 conditions set forth above (logical, clinical, noninvasive, and highly reproducible conditions), statistical base BP is believed to be suitable as an individual representative BP value. Determination of its relation to the second and fifth conditions (indicating severity and prognosis of hypertension) requires further study involving noninvasive BP measurement in a much larger number of subjects. The present study is only the first step toward achieving this ultimate goal.
The authors wish to thank Shigehiro Ishizuka and Kenichirou Yasaka, Research and Development Section, A&D Co, Ltd, Tokyo, Japan, for providing the computer program used for statistical base BP calculations of TM2425 recordings. This program is commercially available.
- Received March 5, 1998.
- Revision received March 19, 1998.
- Accepted April 13, 1998.
Tochikubo O, Kawano Y, Miyajima E, Minamisawa K, Ishii M. Relationship between blood pressure and heart rate while awake and asleep in patients with mild essential hypertension. J Circ J. 1993:57:1128–1136.
Minamisawa K, Tochikubo O, Ishii M. Systemic hemodynamics during sleep in young or middle-aged and elderly patients with essential hypertension. Hypertension. 1994;23:167–173.
Devereux RB, Casale PN, Kligfield P, Eisenberg RR, Miller D, Campo E, Alonso DR: Performance of primary and derived M-mode echocardiographic measurements for detection of left ventricular hypertrophy in necropsied subjects and in patients with systemic hypertension, mitral regurgitation and dilated cardiomyopathy. Am J Cardiol. 1986;57:1388–1393.
Tochikubo O, Ikeda A, Miyajima E, Ishii M. Effects of insufficient sleep on blood pressure monitored by a new multi-biomedical recorder. Hypertension. 1996;27:1318–1324.
Tochikubo O, Shigemasa T, Miyajima E, Ishii M, Nagashima T, Ishizuka S. The development of a new portable multi-biomedical recorder and its accuracy of blood pressure measurement. Ther Res. 1996;17:88–94.
Sahn DJ, DeMaria A, Kisslo J, Weyman A, for the Committee on M-Mode Standardization of American Society of Echocardiography. Recommendations regarding quantitation in M-mode echocardiography: results of a survey of echocardiographic measurements. Circulation. 1978;58:1072–1083.
Vecchio TJ. Predictive value of a single diagnostic test in unselected populations. N Engl J Med. 1966;274:1171–1173.
Armitage P, Berry G. Discriminant analysis. In: Armitage P, Berry G, eds. Statistical Method in Medical Research. Cambridge, UK: Blackwell Science Inc; 1994:358–372.
Gretler DD, Carlson GF, Montano AV, Murphy MB. Diurnal blood pressure variability and physical activity measured electronically and by diary. Am J Hypertens. 1993;6:127–133.
Pickering TG. The clinical significance of diurnal blood pressure variations: dippers and non-dippers. Circulation. 1990;81:700–702.