Abstract

Many surveys and cohort studies have used a random-zero sphygmomanometer blood pressure device (RZS) to measure subjects' blood pressure and to assess the value of blood pressure in predicting cardiovascular events. Recent studies used automated oscillometric blood pressure devices (AODs) that systematically measure higher blood pressure values than RZSs do, hampering comparability of values between these studies. In 2000–2003, the authors randomly used both an RZS and an AOD in an ongoing cohort study in Germany. This analysis aimed to compare blood pressure values by device and to develop an algorithm to convert estimates of blood pressure values from one device to the other. In a randomized subset of 2,365 subjects aged 45–75 years, each subject was measured three times with each device in a randomized order. The mean difference (AOD-RZS) between the devices was 3.9 mmHg for systolic blood pressure and 2.6 mmHg for diastolic blood pressure. The authors found that linear regression models including age, sex, and blood pressure level can be used to convert RZS blood pressure values to AOD blood pressure values, and vice versa. Results may help to better compare blood pressure values in epidemiologic studies that used different blood pressure devices.

The recording of blood pressure with a standard mercury sphygmomanometer is prone to errors induced by the observer's bias toward particular digits or findings. The random-zero sphygmomanometer blood pressure device (RZS) was introduced as a modified instrument with the main purpose of eliminating bias from preconceived expectations regarding blood pressure (1, 2).

In the past, many large epidemiologic studies used a standard mercury sphygmomanometer or RZS to assess the value of blood pressure in predicting cardiovascular events (38). However, several studies have repeatedly cast doubt on the accuracy of the RZS (914); compared with standard mercury sphygmomanometers, this instrument underrecords blood pressure (15). Brown et al. (16) found that random-zero values are not randomly distributed when the cuff inflation time is too short, the mercury sticks in the manometer, the air bleed screw leaks, or the interval between opening the reservoir tap and spinning the thumbwheel is too short.

Recently, the European Society of Hypertension recommended against using random-zero blood pressure devices because they failed the quality standards of the British Hypertension Society and the Association for the Advancement of Medical Instrumentation (14). Currently, Europe and the United States are working to phase out the use of mercury sphygmomanometer devices because of mercury's potential threat to human health and the environment (17). The use of mercury in Sweden and the Netherlands is no longer permitted in hospitals (18), the Mayo Clinic (Rochester, Minnesota) has replaced all mercury manometers with commercially available aneroid devices (19), and the Veterans Administration hospitals in the United States have already banned mercury sphygmomanometers (20).

More recent epidemiologic studies used automated blood pressure devices (2126). Automated oscillometric blood pressure devices (AODs) use different blood pressure measuring algorithms that are generally not disclosed by the manufacturer. The measurement principle of these devices involves detecting pulse wave oscillations and estimating systolic blood pressure (SBP) and diastolic blood pressure (DBP) values via numeric algorithms. Several factors may limit oscillometric instruments: 1) motion of the arm can induce noise that is interpreted as pressure oscillation, 2) arrhythmias can distort the oscillometric envelope, 3) reliability and validity studies have been conducted in only selected populations (by age and disease), 4) studies tested the manufacturer's cuff only (19), and 5) epidemiologic data for cardiovascular risk prediction are lacking (20).

Because of systematic measurement differences between blood pressure devices, comparison of blood pressure values between different populations or within populations over time is hampered when different blood pressure devices have been used. For example, the Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) project (4) conducted repeated surveys. German MONICA surveys 1–3 used a Hawksley random-zero device, and the most recent survey used an automated device. Time trend analyses of blood pressure are hampered because the measurement device was changed.

Furthermore, the cardiovascular predictive value of blood pressure measured by automated devices may not necessarily be identical, and independent predictive validation remains a critical issue when automated devices are used in epidemiologic studies. To our knowledge, the predictive value of these two different types of blood pressure recording devices has not been compared in one large epidemiologic cohort study until now.

In our ongoing population-based cohort study, we used two blood pressure recording devices for the same subjects (RZS and AOD) to investigate agreement between the two recording methods with regard to blood pressure level and the power of these devices to predict coronary events. The aim of this analysis was to compare blood pressure values by device and to develop an algorithm to convert estimates of blood pressure values from one device to the other. This algorithm could be used for comparison between studies that used either RZSs or AODs.

MATERIALS AND METHODS

The rationale for the ongoing Heinz Nixdorf Recall Study has been described in detail previously (27, 28). Briefly, it is a population-based, prospective cohort study of the comparative predictive value of modern risk stratification techniques for coronary events (acute myocardial infarction and/or sudden cardiac death). The baseline assessment included measures of subclinical atherosclerosis (ankle-brachial index, coronary calcium by electron-beam computed tomography, carotid intima-media thickness), comprehensive laboratory tests, questionnaires for comprehensive risk assessment, blood pressure measurements, resting and exercise electrocardiograms, and other items. From December 2000 through August 2003, we recruited 4,814 subjects aged 45–75 years who lived in one of three industrial cities of the Ruhr area (Essen, Bochum, Mülheim) of Germany. Subjects were randomly selected from mandatory lists of residences that are regarded as the most complete population-based sampling frames in Germany (29). The response proportion, expressed as the recruitment efficacy proportion, was 55.8 percent (30). Currently, participants are followed until 5 years after baseline examination. The end of the follow-up period is scheduled to be July 2008.

Baseline examination of our study subjects included a questionnaire-based risk assessment, anthropometric measurements, comprehensive laboratory tests of blood and urine, resting and exercise electrocardiograms, ankle-brachial index, ultrasound scanning of carotid intima-media thickness, and noncontrast-enhanced electron-beam computed tomography to assess coronary calcium quantities. The study organization was certified according to Deutsches Institut für Normung (DIN), Euronorm (EN) ISO 9001:2000 (31).

The protocol for blood pressure recording included two different measurement methods applied one after another on the same subject. We used an AOD (Omron HEM-705CP; OMRON Corporation, Hoofddorp, the Netherlands) that displayed blood pressure values to the nearest 1 mmHg. In addition, we used a random-zero blood pressure device (Mark II; Hawksley, Lancing, United Kingdom) that displayed blood pressure values to the nearest 2 mmHg. All devices were regularly calibrated by the Bureau of Standards (Board of Weights and Measures). Study personnel were certified and regularly trained in measuring blood pressure according to the standards of the World Health Organization (WHO) MONICA blood pressure recording protocol (32) and were monitored during the field phase by an external quality control group that organized regular audits.

Blood pressure was recorded three times for each participant (with a 3-minute interval in between) by using each blood pressure recording device on the right arm. Before the first measurement, the circumference of the upper arm was measured. For the AOD, we used a cuff whose size (width × length) was either 14 cm × 48 cm (circumferences of ≤31.9 cm) or 16 cm × 65 cm (circumferences of ≥32.0 cm); for the RZS, we used a cuff whose size was 12 cm × 23 cm (circumferences of ≤20 cm), 12 cm × 28 cm (circumferences of ≤31.9 cm), or 14 cm × 40 cm (circumferences of ≥32.0 cm). All were original cuffs from the manufacturers of the blood pressure devices. Study personnel followed several instructions: 1) the rubber bladder of the cuff had to lie over the brachial artery, 2) auscultation had to be at heart level, 3) the peak inflation had to be about 30 mmHg above radial pulse disappearance, 4) the cuff deflation rate had to be 2 mmHg/second or less (RZS), and 5) the blood pressure values had to be recorded to the nearest even digit (RZS only).

Blood pressure was recorded during a computer-assisted personal interview that was automatically interrupted for the blood pressure recording after the first 53 questions and after a maximum of 100 additional questions, depending on the length of a subject's occupational history. This course of action guaranteed that subjects rested at least 5 minutes before their first blood pressure measurement. The order of the use of the blood pressure devices was pseudo-randomized beginning in July 7, 2001, and was based on a subject's personal identification number assigned when the first invitational letter was sent to subjects. Subjects with odd study numbers were assigned the order RZS-AOD, and subjects with even study numbers were assigned the order AOD-RZS.

Because of the long duration of the examination process and in case of a temporary shortage of study personnel, a reduced program of blood pressure recordings (two AOD and one RZS reading) was applied. Figure 1 presents the sequence of blood pressure recording.

FIGURE 1.

Sequence of blood pressure (BP) measurements taken with two different blood pressure recording devices (either an automated oscillometric device or a random-zero device) in the Heinz Nixdorf Recall Study, Germany, 2000–2003. CAPI, computer-assisted personal interview.

FIGURE 1.

Sequence of blood pressure (BP) measurements taken with two different blood pressure recording devices (either an automated oscillometric device or a random-zero device) in the Heinz Nixdorf Recall Study, Germany, 2000–2003. CAPI, computer-assisted personal interview.

For the analyses presented here, we excluded participants for the following reasons: 1) examination conducted before randomization of the order of the blood pressure devices was applied (before July 7, 2001; n = 617); 2) fewer than three blood pressure recordings with each device (n = 760; for 528 of these subjects, the reason was that a reduced blood pressure program was used); 3) information missing on cuff size, or wrong cuff size used (n = 145); 4) pseudo-randomization not followed (n = 179); and 5) time difference more than 30 minutes between blood pressure recordings with the first and second devices (n = 748), leaving information for 2,365 participants in the final data set. The mean time difference between the start of the first and the second blood pressure recordings was 22 minutes (standard deviation, 4.2 minutes).

For each blood pressure recording device, we disregarded the first measurement from each because it is typically systematically higher than further serial measurements (20). We calculated mean SBP and DBP from the second and third recordings. On the basis of these mean values, we calculated pulse pressure (SBP – DBP) and mean arterial pressure (1/3 SBP + 2/3 DBP). In addition, we categorized the blood pressure values according to the JNC-7 classification (33).

To graphically illustrate the individual differences in the blood pressure values by recording device, we plotted the difference between the two devices (AOD – RZS) against the average for the two devices according to the method of Bland and Altman (34). To measure agreement of the blood pressure classes according to JNC-7 (33) between the two devices, we calculated observed and chance-corrected agreement as measured by the kappa statistic (35).

In addition, we used multiple linear regression models to estimate the AOD values from RZS values (and vice versa), adjusting for age, sex, mean blood pressure level, and other variables. We checked the model fit by using residual plots. All analyses were performed with SAS 9.1 software (36). To study the robustness of our findings, we reran our regression models by also including data for subjects for whom the time difference between measurement with the first and the second blood pressure devices was more than 30 minutes.

RESULTS

The baseline comparison of the 2,365 randomized subjects is presented in table 1. The time differences between measurements by device were independent of device order. Compared with subjects for whom the order was RZS-AOD, those for whom the order was AOD-RZS were on average slightly heavier (1.2 kg) and consequently had a slightly higher average body mass index.

TABLE 1.

Baseline comparison of the 2,365 randomized participants whose blood pressure was measured, by device order, in the Heinz Nixdorf Recall Study, Germany, 2000–2003


 

First AOD*, then RZS* (n = 1,238)
 
 
First RZS, then AOD (n = 1,127)
 
 

 
Mean or no.
 
SD* or %
 
Mean or no.
 
SD or %
 
Age (years) 60.8 7.8 61.2 0.22 
Weight (kg) 79.2 14.7 78.0 0.05 
Height (cm) 167.9 8.8 167.6 0.56 
Body mass index (weight (kg)/height (m)228.1 4.6 27.7 0.04 
Time difference between device testing (minutes) 22.5 4.1 22.4 0.32 
Waist-to-hip ratio 0.91 0.09 0.91 0.31 
Heart rate (bpm) at first blood pressure measurement with the first device 76.8 12.0 76.1 11.4 
Sex: women 657 53.1 603 53.5 
History of coronary artery disease
 
101
 
8.2
 
84
 
7.5
 

 

First AOD*, then RZS* (n = 1,238)
 
 
First RZS, then AOD (n = 1,127)
 
 

 
Mean or no.
 
SD* or %
 
Mean or no.
 
SD or %
 
Age (years) 60.8 7.8 61.2 0.22 
Weight (kg) 79.2 14.7 78.0 0.05 
Height (cm) 167.9 8.8 167.6 0.56 
Body mass index (weight (kg)/height (m)228.1 4.6 27.7 0.04 
Time difference between device testing (minutes) 22.5 4.1 22.4 0.32 
Waist-to-hip ratio 0.91 0.09 0.91 0.31 
Heart rate (bpm) at first blood pressure measurement with the first device 76.8 12.0 76.1 11.4 
Sex: women 657 53.1 603 53.5 
History of coronary artery disease
 
101
 
8.2
 
84
 
7.5
 
*

AOD, automated oscillometric blood pressure device; RZS, random-zero sphygmomanometer blood pressure device; SD, standard deviation.

History of an acute myocardial infarction and/or a coronary intervention including coronary angioplasty, stent, or artery bypass.

Average SBP and DBP values were lower when measured with the RZS compared with the AOD. Accordingly, use of the RZS resulted in a lower prevalence of hypertensive blood pressure (29.1 percent) than use of the AOD (37.5 percent). The mean difference between the two devices (AOD – RZS) was 3.9 mmHg for SBP and 2.6 mmHg for DBP. The similar difference in the SBP and DBP values between the devices resulted in a small difference of 1.2 mmHg in pulse pressure. Although these differences appear small, the variation underlying these mean differences is large, as indicated by the large standard deviations (table 2). The observed agreement of the classification of blood pressure according to JNC-7 (33) between the devices was 0.63, the unweighted kappa was 0.48 (95 percent confidence interval: 0.45, 0.50), and the weighted kappa was 0.62 (95 percent confidence interval: 0.60, 0.64) (data not shown).

TABLE 2.

Comparison of the two blood pressure recording devices among 2,365 randomized participants in the Heinz Nixdorf Recall Study, Germany, 2000–2003*



 

RZS,
 

AOD
 

Mean difference
 
Mean blood pressure value (mmHg)§    
    Systolic blood pressure 128.5 (18.5) 132.4 (20.1) 3.9 (9.9) 
    Diastolic blood pressure 78.1 (10.0) 80.7 (10.3) 2.6 (6.6) 
    Pulse pressure 50.4 (14.2) 51.7 (14.5) 1.2 (9.1) 
    Mean arterial pressure 94.9 (11.6) 97.9 (12.7) 3.0 (6.6) 
Blood pressure class (no. (%))    
    Normal 708 (29.9) 607 (25.7)  
    Prehypertensive 970 (41.0) 871 (36.8)  
    Hypertension stage 1 520 (22.0) 653 (27.6)  
    Hypertension stage 2 167 (7.1) 234 (9.9)  
Mean difference between AOD and RZS (AOD − RZS) by blood pressure category#    
    Systolic difference (mmHg)    
        Normal   1.2 (8.2) 
        Prehypertensive   4.4 (9.9) 
        Hypertension stage 1   5.3 (10.3) 
        Hypertension stage 2   6.4 (12.2) 
    Diastolic difference (mmHg)    
        Normal   1.1 (6.2) 
        Prehypertensive   2.7 (6.2) 
        Hypertension stage 1   3.7 (6.7) 
        Hypertension stage 2
 

 

 
4.3 (7.8)
 


 

RZS,
 

AOD
 

Mean difference
 
Mean blood pressure value (mmHg)§    
    Systolic blood pressure 128.5 (18.5) 132.4 (20.1) 3.9 (9.9) 
    Diastolic blood pressure 78.1 (10.0) 80.7 (10.3) 2.6 (6.6) 
    Pulse pressure 50.4 (14.2) 51.7 (14.5) 1.2 (9.1) 
    Mean arterial pressure 94.9 (11.6) 97.9 (12.7) 3.0 (6.6) 
Blood pressure class (no. (%))    
    Normal 708 (29.9) 607 (25.7)  
    Prehypertensive 970 (41.0) 871 (36.8)  
    Hypertension stage 1 520 (22.0) 653 (27.6)  
    Hypertension stage 2 167 (7.1) 234 (9.9)  
Mean difference between AOD and RZS (AOD − RZS) by blood pressure category#    
    Systolic difference (mmHg)    
        Normal   1.2 (8.2) 
        Prehypertensive   4.4 (9.9) 
        Hypertension stage 1   5.3 (10.3) 
        Hypertension stage 2   6.4 (12.2) 
    Diastolic difference (mmHg)    
        Normal   1.1 (6.2) 
        Prehypertensive   2.7 (6.2) 
        Hypertension stage 1   3.7 (6.7) 
        Hypertension stage 2
 

 

 
4.3 (7.8)
 
*

Except for blood pressure class, all values in parentheses are standard deviations.

RZS, random-zero sphygmomanometer blood pressure device; AOD, automated oscillometric blood pressure device.

Zero values ranged from 0 mmHg to 36 mmHg.

§

Average (standard deviation) of the mean blood pressures based on the second and third measurements.

According to the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (33), normal: systolic blood pressure (SBP) <140 and diastolic blood pressure (DBP) <80 mmHg; prehypertension: SBP 120–139 or DBP 80–89 mmHg; hypertension stage 1: SBP 140–159 or DBP 90–99 mmHg; and hypertension stage 2: SBP ≥160 or DBP ≥100 mmHg.

#

Blood pressure classes are based on the mean pressures obtained by using the two devices.

Figure 2 presents Bland and Altman (34) plots of the individual differences between the blood pressure devices against blood pressure level (mean blood pressure measured by the two devices). SBP differed by more than 10 mmHg for 788 subjects (33.3 percent) and by more than 20 mmHg for 140 subjects (5.9 percent). In contrast, DBP differed by more than 10 mmHg for 318 subjects (13.5 percent) and by more than 20 mmHg for 14 subjects (0.6 percent). Based on visual inspection and the results of the linear regression models—with the difference as the dependent variable and the mean pressure level as the independent variable—the difference between the blood pressure devices depended on blood pressure level: the higher the level, the larger the difference between AOD and RZS.

FIGURE 2.

Individual differences between systolic blood pressure (SBP; top) and diastolic blood pressure (DBP; bottom) values obtained with two different blood pressure recording devices among 2,365 study subjects in the Heinz Nixdorf Recall Study, Germany, 2000–2003. Solid lines, overall mean difference; dashed lines, 95% confidence interval. Linear regression models—SBP difference: Y = –8.17 + 0.09 × level (mmHg) (p < 0.0001); DBP difference: Y = –0.74 + 0.04 × level (mmHg) (p = 0.003).

FIGURE 2.

Individual differences between systolic blood pressure (SBP; top) and diastolic blood pressure (DBP; bottom) values obtained with two different blood pressure recording devices among 2,365 study subjects in the Heinz Nixdorf Recall Study, Germany, 2000–2003. Solid lines, overall mean difference; dashed lines, 95% confidence interval. Linear regression models—SBP difference: Y = –8.17 + 0.09 × level (mmHg) (p < 0.0001); DBP difference: Y = –0.74 + 0.04 × level (mmHg) (p = 0.003).

The differences between the devices appeared to be larger among men and to increase by age for both men and women. With increasing SBP and DBP, the difference between the devices became larger. Subjects first measured by RZS showed larger SBP and DBP differences between the devices than subjects who were first measured by AOD (table 3).

TABLE 3.

Potential determinants of blood pressure differences (mmHg) between the AOD* and the RZS,* Heinz Nixdorf Recall Study, Germany, 2000–2003


 

No.
 

Systolic blood pressure difference
 
 
Diastolic blood pressure difference
 
 

 
 Mean
 
SD*
 
Mean
 
SD
 
Overall 2,365 3.9 9.9 2.6 6.6 
Sex      
    Men 1,105 7.0 9.2 3.4 6.5 
    Women 1,260 1.1 9.7 1.9 6.6 
Age group (years)      
    45–54 585 3.0 9.4 2.0 6.6 
    55–64 945 3.8 9.5 2.6 6.3 
    65–75 835 4.6 10.6 3.0 6.8 
Sex and age group (years)      
    Men      
        45–54 272 6.7 8.4 2.6 6.7 
        55–64 426 6.5 9.1 3.4 6.4 
        65–75 407 7.6 9.8 3.9 6.3 
    Women      
        45–54 313 –0.2 9.0 1.6 6.5 
        55–64 519 1.6 9.3 2.0 6.2 
        65–75 428 1.7 10.5 2.1 7.0 
Blood pressure level      
    1st tertile 790 1.5 8.2 2.2 6.2 
    2nd tertile 784 4.7 10.1 2.5 6.4 
    3rd tertile 791 5.4 10.8 3.1 7.0 
Sex and blood pressure level      
    Men      
        1st tertile 274 4.8 8.5 3.2 6.8 
        2nd tertile 392 8.2 8.5 3.2 6.1 
        3rd tertile 439 7.1 10.0 3.7 6.6 
    Women      
        1st tertile 516 –0.3 7.5 1.6 5.9 
        2nd tertile 392 1.2 10.4 1.9 6.7 
        3rd tertile 352 3.1 11.2 2.4 7.3 
Device order      
    AOD-RZS 1,238 3.1 9.7 1.8 6.6 
    RZS-AOD
 
1,127
 
4.6
 
10.1
 
3.5
 
6.5
 

 

No.
 

Systolic blood pressure difference
 
 
Diastolic blood pressure difference
 
 

 
 Mean
 
SD*
 
Mean
 
SD
 
Overall 2,365 3.9 9.9 2.6 6.6 
Sex      
    Men 1,105 7.0 9.2 3.4 6.5 
    Women 1,260 1.1 9.7 1.9 6.6 
Age group (years)      
    45–54 585 3.0 9.4 2.0 6.6 
    55–64 945 3.8 9.5 2.6 6.3 
    65–75 835 4.6 10.6 3.0 6.8 
Sex and age group (years)      
    Men      
        45–54 272 6.7 8.4 2.6 6.7 
        55–64 426 6.5 9.1 3.4 6.4 
        65–75 407 7.6 9.8 3.9 6.3 
    Women      
        45–54 313 –0.2 9.0 1.6 6.5 
        55–64 519 1.6 9.3 2.0 6.2 
        65–75 428 1.7 10.5 2.1 7.0 
Blood pressure level      
    1st tertile 790 1.5 8.2 2.2 6.2 
    2nd tertile 784 4.7 10.1 2.5 6.4 
    3rd tertile 791 5.4 10.8 3.1 7.0 
Sex and blood pressure level      
    Men      
        1st tertile 274 4.8 8.5 3.2 6.8 
        2nd tertile 392 8.2 8.5 3.2 6.1 
        3rd tertile 439 7.1 10.0 3.7 6.6 
    Women      
        1st tertile 516 –0.3 7.5 1.6 5.9 
        2nd tertile 392 1.2 10.4 1.9 6.7 
        3rd tertile 352 3.1 11.2 2.4 7.3 
Device order      
    AOD-RZS 1,238 3.1 9.7 1.8 6.6 
    RZS-AOD
 
1,127
 
4.6
 
10.1
 
3.5
 
6.5
 
*

AOD, automated oscillometric blood pressure device; RZS, random-zero sphygmomanometer blood pressure device; SD, standard deviation.

Tertiles are based on the mean pressures obtained by using the two devices. Systolic blood pressure: 1st: 80–121 mmHg, 2nd: 122–137 mmHg, 3rd: 138–214 mmHg; diastolic blood pressure: 1st: 48–75 mmHg, 2nd: 76–83 mmHg, 3rd: 84–132 mmHg.

The first blood pressure value obtained from each device was, as expected, higher than the following two values. If the measurement order was AOD-RZS, the RZS values were higher than expected based on RZS measurements 1–3 among the group first measured by RZS. The dashed lines in figure 3 indicate the expected values assuming that the two subgroups are exchangeable. The order effect did not markedly change when we included subjects for whom the time difference between measurement with the first and the second blood pressure devices was more than 30 minutes.

FIGURE 3.

Mean systolic blood pressure (SBP; top) and diastolic blood pressure (DBP; bottom) values obtained during six blood pressure measurements taken with two different recording devices (either an automated oscillometric blood pressure device (AOD) or a random-zero sphygmomanometer blood pressure device (RZS)), by order of device in the Heinz Nixdorf Recall Study, Germany, 2000–2003. 1–3 AOD: measurements 1–3 taken with an AOD; 4–6 RZS: measurements 4–6 taken with an RZS; 1–3 RZS: measurements 1–3 taken with an RZS; 4–6 AOD: measurements 4–6 taken with an AOD. The dashed lines show the expected random-zero blood pressure values based on the subgroup first measured by RZS and then by AOD.

FIGURE 3.

Mean systolic blood pressure (SBP; top) and diastolic blood pressure (DBP; bottom) values obtained during six blood pressure measurements taken with two different recording devices (either an automated oscillometric blood pressure device (AOD) or a random-zero sphygmomanometer blood pressure device (RZS)), by order of device in the Heinz Nixdorf Recall Study, Germany, 2000–2003. 1–3 AOD: measurements 1–3 taken with an AOD; 4–6 RZS: measurements 4–6 taken with an RZS; 1–3 RZS: measurements 1–3 taken with an RZS; 4–6 AOD: measurements 4–6 taken with an AOD. The dashed lines show the expected random-zero blood pressure values based on the subgroup first measured by RZS and then by AOD.

Univariate comparisons of the mean blood pressure differences between the randomized groups are complicated because of confounding between blood pressure level, age, and sex. Tables 4 and 5 present the results of multiple linear regression models with the AOD-based SBP and DBP values as the dependent variables, and vice versa. These equations may be used to convert random-zero blood pressure values to automated oscillometric blood pressure values, and vice versa. For example, model M1 predicts that a man aged 55 years with an RZS-based SBP of 160 mmHg has an AOD-based SBP of 3.3171 + 0.9201 × 160 + 0.1311 × 55 + 6.0246 × 1 = 163.77, that is, 164 mmHg. The predicted mean values based on the different models were very similar and did not show a consistent pattern by device or DBP and SBP. For investigators who want to convert their RZS-based values to automated oscillometric blood pressure values (or vice versa), model M1 may be appropriate. Adjustment for weight, body mass index, measurement time difference between the devices, and observer did not markedly change the results. First-order interaction terms including age × sex and age × blood pressure resulted in virtually no change in the explained variance of the models presented in tables 4 and 5 (data not shown).

TABLE 4.

Results (beta coefficients (standard errors)) of the multiple linear regression model used to convert blood pressure values from the RZS* to the AOD,* Heinz Nixdorf Recall Study, Germany, 2000–2003


Model
 

Intercept
 

RZS value (mmHg)
 

Device order
 

Age (years)
 

Sex
 

Adjusted R2
 

Predicted mean pressure (mmHg) (AOD)
 

Mean squared difference
 
Systolic blood pressure         
Complete data set (N = 2,365)         
    M1 3.3171 (1.7984) 0.9201 (0.0109)  0.1311 (0.0261) 6.0246 (0.3869) 0.78 132.40 (17.84) 87.28 
    M2 2.6248 (1.8054) 0.9224 (0.0109) 1.3265 (0.3849) 0.1273 (0.0261) 6.0242 (0.3860) 0.79 132.40 (17.84) 86.84 
Subgroup AOD-RZS§ (n = 1,238)         
    M3 4.4215 (2.3741) 0.9203 (0.0143)  0.1010 (0.0351) 6.1764 (0.5230) 0.80 132.56 (18.45) 82.99 
Subgroup RZS-AOD (n = 1,127)         
    M4 1.8674 (2.7339) 0.9253 (0.0168)  0.1563 (0.0389) 5.8877 (0.5720) 0.76 132.21 (17.17) 90.95 
Diastolic blood pressure         
Complete data set (N = 2,365)         
    M1 14.5647 (1.4735) 0.8092 (0.0130)  0.0329 (0.0167) 2.0108 (0.2599) 0.64 80.71 (8.25) 38.88 
    M2 13.6049 (1.4768) 0.8148 (0.0130) 1.3420 (0.2565) 0.0311 (0.0166) 2.0016 (0.2585) 0.64 80.71 (8.28) 38.44 
Subgroup AOD-RZS (n = 1,238)         
    M3 14.7689 (1.9970) 0.8008 (0.0178)  0.0303 (0.0228) 1.9884 (0.3590) 0.64 80.72 (8.29) 38.53 
Subgroup RZS-AOD (n = 1,127)         
    M4
 
13.5418 (2.1658)
 
0.8314 (0.0191)
 

 
0.0330 (0.0243)
 
2.0263 (0.3730)
 
0.64
 
80.72 (8.28)
 
38.29
 

Model
 

Intercept
 

RZS value (mmHg)
 

Device order
 

Age (years)
 

Sex
 

Adjusted R2
 

Predicted mean pressure (mmHg) (AOD)
 

Mean squared difference
 
Systolic blood pressure         
Complete data set (N = 2,365)         
    M1 3.3171 (1.7984) 0.9201 (0.0109)  0.1311 (0.0261) 6.0246 (0.3869) 0.78 132.40 (17.84) 87.28 
    M2 2.6248 (1.8054) 0.9224 (0.0109) 1.3265 (0.3849) 0.1273 (0.0261) 6.0242 (0.3860) 0.79 132.40 (17.84) 86.84 
Subgroup AOD-RZS§ (n = 1,238)         
    M3 4.4215 (2.3741) 0.9203 (0.0143)  0.1010 (0.0351) 6.1764 (0.5230) 0.80 132.56 (18.45) 82.99 
Subgroup RZS-AOD (n = 1,127)         
    M4 1.8674 (2.7339) 0.9253 (0.0168)  0.1563 (0.0389) 5.8877 (0.5720) 0.76 132.21 (17.17) 90.95 
Diastolic blood pressure         
Complete data set (N = 2,365)         
    M1 14.5647 (1.4735) 0.8092 (0.0130)  0.0329 (0.0167) 2.0108 (0.2599) 0.64 80.71 (8.25) 38.88 
    M2 13.6049 (1.4768) 0.8148 (0.0130) 1.3420 (0.2565) 0.0311 (0.0166) 2.0016 (0.2585) 0.64 80.71 (8.28) 38.44 
Subgroup AOD-RZS (n = 1,238)         
    M3 14.7689 (1.9970) 0.8008 (0.0178)  0.0303 (0.0228) 1.9884 (0.3590) 0.64 80.72 (8.29) 38.53 
Subgroup RZS-AOD (n = 1,127)         
    M4
 
13.5418 (2.1658)
 
0.8314 (0.0191)
 

 
0.0330 (0.0243)
 
2.0263 (0.3730)
 
0.64
 
80.72 (8.28)
 
38.29
 
*

RZS, random-zero sphygmomanometer blood pressure device; AOD, automated oscillometric blood pressure device.

AOD-RZS: 0; RZS-AOD: 1.

1 = male, 0 = female.

§

Subjects were first measured by AOD and then by RZS.

Subjects were first measured by RZS and then by AOD.

TABLE 5.

Results (beta coefficients (standard errors)) of the multiple linear regression model used to convert blood pressure values from the AOD* to the RZS,* Heinz Nixdorf Recall Study, Germany, 2000–2003


Model
 

Intercept
 

AOD value (mmHg)
 

Device order
 

Age (years)
 

Sex
 

Adjusted R2
 

Predicted mean pressure (mmHg) (RZS)
 

Mean squared difference
 
Systolic blood pressure         
Complete data set (N = 2,365)         
    M1 18.4523 (1.6510) 0.8153 (0.0097)  0.0668 (0.0247) –4.1439 (0.3728) 0.77 128.65 (16.24) 77.35 
    M2 19.1164 (1.6512) 0.8146 (0.0097) –1.6039 (0.3611) 0.0700 (0.0246) –4.1455 (0.3714) 0.77 128.53 (16.26) 76.69 
Subgroup AOD-RZS§ (n = 1,238)         
    M3 15.3955 (2.2246) 0.8370 (0.0130)  0.0838 (0.0335) –4.3224 (0.5116) 0.79 129.42 (17.04) 75.48 
Subgroup RZS-AOD (n = 1,127)         
    M4 21.8881 (2.4374) 0.7881 (0.0143)  0.0549 (0.0361) –4.0056 (0.5391) 0.75 127.58 (15.34) 77.45 
Diastolic blood pressure         
Complete data set (N = 2,365)         
    M1 19.5679 (1.4068) 0.7664 (0.0123)  –0.0509 (0.0162) –0.4674 (0.2559) 0.63 78.11 (7.90) 36.82 
    M2 20.1798 (1.3974) 0.7665 (0.0122) –1.6365 (0.2479) –0.0482 (0.0161) –0.4758 (0.2537) 0.63 78.11 (7.94) 36.16 
Subgroup AOD-RZS (n = 1,238)         
    M3 18.8158 (1.9376) 0.7770 (0.0172)  –0.0408 (0.0225) –0.3374 (0.3579) 0.63 78.90 (8.01) 37.38 
Subgroup RZS-AOD (n = 1,127)         
    M4
 
20.1722 (2.0109)
 
0.7546 (0.0174)
 

 
–0.0580 (0.0231)
 
–0.6407 (0.3595)
 
0.63
 
77.23 (7.77)
 
34.75
 

Model
 

Intercept
 

AOD value (mmHg)
 

Device order
 

Age (years)
 

Sex
 

Adjusted R2
 

Predicted mean pressure (mmHg) (RZS)
 

Mean squared difference
 
Systolic blood pressure         
Complete data set (N = 2,365)         
    M1 18.4523 (1.6510) 0.8153 (0.0097)  0.0668 (0.0247) –4.1439 (0.3728) 0.77 128.65 (16.24) 77.35 
    M2 19.1164 (1.6512) 0.8146 (0.0097) –1.6039 (0.3611) 0.0700 (0.0246) –4.1455 (0.3714) 0.77 128.53 (16.26) 76.69 
Subgroup AOD-RZS§ (n = 1,238)         
    M3 15.3955 (2.2246) 0.8370 (0.0130)  0.0838 (0.0335) –4.3224 (0.5116) 0.79 129.42 (17.04) 75.48 
Subgroup RZS-AOD (n = 1,127)         
    M4 21.8881 (2.4374) 0.7881 (0.0143)  0.0549 (0.0361) –4.0056 (0.5391) 0.75 127.58 (15.34) 77.45 
Diastolic blood pressure         
Complete data set (N = 2,365)         
    M1 19.5679 (1.4068) 0.7664 (0.0123)  –0.0509 (0.0162) –0.4674 (0.2559) 0.63 78.11 (7.90) 36.82 
    M2 20.1798 (1.3974) 0.7665 (0.0122) –1.6365 (0.2479) –0.0482 (0.0161) –0.4758 (0.2537) 0.63 78.11 (7.94) 36.16 
Subgroup AOD-RZS (n = 1,238)         
    M3 18.8158 (1.9376) 0.7770 (0.0172)  –0.0408 (0.0225) –0.3374 (0.3579) 0.63 78.90 (8.01) 37.38 
Subgroup RZS-AOD (n = 1,127)         
    M4
 
20.1722 (2.0109)
 
0.7546 (0.0174)
 

 
–0.0580 (0.0231)
 
–0.6407 (0.3595)
 
0.63
 
77.23 (7.77)
 
34.75
 
*

AOD, automated oscillometric blood pressure device; RZS, random-zero sphygmomanometer blood pressure device.

AOD-RZS: 0; RZS-AOD: 1.

1 = male, 0 = female.

§

Subjects were first measured by AOD and then by RZS.

Subjects were first measured by RZS and then by AOD.

DISCUSSION

Because prediction of cardiovascular disease risk has traditionally been based on mercury blood pressure measurements (e.g., Framingham risk score), and AODs use a different method of measurement, there is some uncertainty about whether traditional risk prediction based on random-zero mercury devices can be applied to subjects measured with automated oscillometric devices. In addition, comparability of survey data based on different methods of blood pressure measurement is hampered because of systematic differences between the methods. The observation of systematically higher SBP and DBP values with AODs compared with RZSs implies that cross-sectional studies that determine the prevalence of hypertensive blood pressure values produce higher prevalence estimates when blood pressure measurement is based on automated oscillometric devices compared with random-zero devices. In our study, the overall prevalence of hypertensive blood pressure differed by 8.4 percentage points between the devices. As we have shown, the difference in prevalence depended on age, sex, and blood pressure level, complicating comparison of studies that used either a random-zero or an automated oscillometric device.

Mackie et al.'s study results (15) and these authors' evaluation of six other studies that compared RZS and standard mercury sphygmomanometers consistently showed that random-zero sphygmomanometry underestimates SBP by 1.35 mmHg and DBP by 1.97 mmHg when compared with the standard mercury sphygmomanometer. It is interesting to note that the pulse pressures obtained by the two devices, which may be another important predictor of cardiovascular events (37, 38), are quite comparable.

The blood pressure differences between the devices were greater among men than women, increased by age independently of sex, and depended on blood pressure level. The increasing difference between blood pressure values by increasing blood pressure levels has also been described by others, including O'Brien et al. (39) in 1993 and Whincup et al. (40), who compared the Hawksley random sphygmomanometer with the Dinamap 1846SX (GE Healthcare, Chalfont St. Giles, United Kingdom). The age dependence of the blood pressure differences may be explained by the particular vulnerability of the oscillometric technique in elderly people with stiff arteries due to atherosclerosis (41). The difference between the devices may also have been aggravated by different degrees of recording precision. The automated blood pressure device recorded blood pressure values to the nearest 1 mmHg, whereas blood pressure values from the random-zero device were displayed to the nearest 2 mmHg.

Our data may enable investigators to convert blood pressure values derived from random-zero devices to virtual blood pressure values obtained with automated oscillometric devices, and vice versa. This conversion is important for comparisons between populations, where blood pressure values were derived by using either device, and for comparisons within populations, where the blood pressure recording device has been changed over time (usually from random-zero to automated oscillometric devices). This conversion would improve comparability of the blood pressure values across studies.

However, the conversion appears to be complicated because of an order effect in our data. One potential explanation may be observer bias. Once the observer knows the blood pressure level obtained from the automated device, he or she will be able to more accurately detect phase I and phase V of the auscultatory measurement of the random-zero measurement. Because of the random change in the zero level of the RZS, this explanation is not very plausible. If that explanation were correct, the subgroup first measured by using the random-zero device would provide the most appropriate conversion algorithm. However, it was disturbing to find that the difference between the blood pressure devices—if any—increased over time, which would argue against observer bias that usually diminishes over time. In addition, adjustment of the order effect by observer did not markedly change the order effect. It is even more difficult to speculate about potential physiologic mechanisms within the tissues (e.g., connective tissue, arterial walls) of the upper arm that could explain the order effect. We are not aware of any publication that deals with this question. We believe that the linear regression models of our complete data set (model M1) that take age, sex, and measured blood pressure into account are appropriate as long as we cannot interpret the order effect in our data.

Although our study has strengths, several factors limit our results. First, we were not able to simultaneously measure blood pressure with the two devices. The median time difference between measurements with the two devices was 22 minutes in our study and may have introduced some extra variation that cannot be attributed to the different measurement devices. Because of the pseudo-randomization of the devices, we expect that this extra variation was random and did not result in systematic differences between the blood pressure devices.

Second, although study personnel were certified to use both blood pressure devices at the beginning of the study and underwent regular monitoring throughout the field phase, we cannot guarantee that handling errors did not influence our results, although this influence is difficult to interpret in light of the device order effect in our data. Several handling errors regarding the Hawksley RZS have been described in the literature, including too-short time intervals of cuff inflation and too-short time intervals between opening the reservoir tap and spinning the thumbwheel (16), last digit preferences (12), and inaccurate recording of the final zero (13).

In conclusion, the blood pressure differences between the Hawksley random-zero and the Omron automated oscillometric devices depend on blood pressure level, age, and sex. Linear regression models taking these variables into account enable conversion of blood pressure values obtained from one device to those from the other, and vice versa. The conversion may help to better compare epidemiologic and clinical blood pressure values between studies that used different blood pressure measurement devices.

The authors thank the Heinz Nixdorf Stiftung (Chairman: Dr. jur G. Schmidt) for sponsoring this study. They gratefully acknowledge collaboration with Dr. Dietich Grönemeyer (Bochum); Dr. Raimund Seibel (Mülheim); and Dr. Klaus Mann, Dr. Lothar Volbracht, Dr. Simone Münkel, Tico Ebralitze, Jana Kondratieva, Astrid Feuersenger, Herbert Hirche (Essen), and Dr. Richard Peter (Ulm).

Conflict of interest: none declared.

References

1.
Garrow JS. Zero-muddler for unprejudiced sphygmomanometry. (Letter).
Lancet
 
1963
;
41
:
1205
.
2.
Wright BM, Dore CF. A random zero sphygmomanometer.
Lancet
 
1970
;
1
:
337
–8.
3.
Vasan RS, Larson MG, Leip EP, et al. Impact of high-normal blood pressure on the risk of cardiovascular disease.
N Engl J Med
 
2001
;
345
:
1291
–7.
4.
Kuulasmaa K, Tunstall-Pedoe H, Dobson A, et al. Estimation of contribution of changes in classic risk factors to trends in coronary-event rates across the WHO MONICA Project populations.
Lancet
 
2000
;
355
:
675
–87.
5.
Vargas CM, Ingram DD, Gillum RF. Incidence of hypertension and educational attainment: the NHANES I Epidemiologic Followup Study.
Am J Epidemiol
 
2000
;
152
:
272
–8.
6.
Wilsgaard T, Jacobsen BK, Schirmer H, et al. Tracking cardiovascular risk factors: the Tromsø Study, 1979–1995.
Am J Epidemiol
 
2001
;
154
:
418
–26.
7.
Hofman A, Grobboe DE, de Jong PT, et al. Determinants of disease and disability in the elderly: the Rotterdam Elderly Study.
Eur J Epidemiol
 
1991
;
7
:
403
–22.
8.
Maher JE, Raz JA, Bielak LF, et al. Potential of quantity of coronary artery calcification to identify new risk factors for asymptomatic atherosclerosis.
Am J Epidemiol
 
1996
;
144
:
943
–53.
9.
Evans JG, Prior IAM. Experience with the random-zero sphygmomanometer.
Br J Prev Soc Med
 
1970
;
24
:
10
–15.
10.
O'Brien E, Mee F, Atkins N, et al. Inaccuracy of Hawksley random zero sphygmomanometer.
Lancet
 
1990
;
336
:
1465
–8.
11.
Conroy RM, O'Brien E, O'Malley K, et al. Measurement error in the Hawksley random zero sphygmomanometer: what damage has been done and what can we learn?
BMJ
 
1993
;
306
:
319
–22.
12.
Lawson M, Johnston A. The Hawksley random zero sphygmomanometer. Should be abandoned. (Letter).
BMJ
 
1993
;
307
:
123
.
13.
McGurk C, Nugent A, McAuley D, et al. Sources of inaccuracy in the use of the Hawksley random-zero sphygmomanometer.
J Hypertens
 
1997
;
15
:
1379
–84.
14.
O'Brien E, Waeber B, Parati G, et al. Blood pressure measuring devices: recommendations of the European Society of Hypertension.
BMJ
 
2001
;
322
:
531
–6.
15.
Mackie A, Whincup P, McKinnon M. Does the Hawksley random zero sphygmomanometer underestimate blood pressure, and by how much?
J Hum Hypertens
 
1995
;
9
:
337
–43.
16.
Brown WC, Kennedy S, Inglis GC, et al. Mechanisms by which the Hawksley random zero sphygmomanometer underestimates blood pressure and produces a non-random distribution of RZ values.
J Hum Hypertens
 
1997
;
11
:
75
–93.
17.
Sheps G. Automated blood pressure devices.
WHL [World Hypertension League] Newsletter
 
2001
;
78
:
1
–2.
18.
Beevers G, Lip GYH, O'Brien E. Blood pressure measurement. Part II. Conventional sphygmomanometer: technique of auscultatory blood pressure measurement.
BMJ
 
2001
;
322
:
1043
–7.
19.
Summary report. National High Blood Pressure Education Program (NHBPEP), National Heart, Lung, and Blood Institute (NHLBI), and American Heart Association (AHA) Working Meeting on Blood Pressure Measurement. National Institutes of Health, Bethesda, Maryland, April 19,
2002
. (http://www.nhlbi.nih.gov/health/prof/heart/hbp/bpmeasu.pdf).
20.
Pickering TG, Hall JE, Appel LJ, et al. Recommendations for blood pressure measurements in humans and experimental animals. Part 1: blood pressure measurement in humans.
Hypertension
 
2005
;
45
:
142
–61.
21.
The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. The ARIC investigators.
Am J Epidemiol
 
1989
;
129
:
687
–702.
22.
Goldstein RZ, Hurwitz BE, Llabre MM, et al. Modeling preclinical cardiovascular risk for use in epidemiologic studies. Miami Community Health Study.
Am J Epidemiol
 
2001
;
154
:
765
–76.
23.
Bild DE, Bluemke DA, Burke GL, et al. Multi-Ethnic Study of Atherosclerosis: objectives and design.
Am J Epidemiol
 
2002
;
156
:
871
–81.
24.
Hunt SC, Ellison RC, Atwood LD, et al. Genome scans for blood pressure and hypertension: the National Heart, Lung, and Blood Institute Family Heart Study.
Hypertension
 
2002
;
40
:
1
–6.
25.
Rathmann W, Haastert B, Icks A, et al. High prevalence of undiagnosed diabetes mellitus in southern Germany: target populations for efficient screening. The KORA survey 2000.
Diabetologia
 
2003
;
46
:
182
–9.
26.
Völzke H, Schwahn C, Wolff B, et al. Hepatitis B and C virus infection and the risk of atherosclerosis in a general population.
Atherosclerosis
 
2004
;
174
:
99
–103.
27.
Schmermund A, Möhlenkamp S, Stang A, et al. Assessment of clinically silent atherosclerotic disease and established and novel risk factors for predicting myocardial infarction and cardiac death in healthy middle-aged subjects: rationale and design of the Heinz Nixdorf RECALL Study. Risk Factors, Evaluation of Coronary Calcium and Lifestyle.
Am Heart J
 
2002
;
144
:
212
–18.
28.
Schmermund A, Möhlenkamp S, Berenbein S, et al. Population-based assessment of subclinical coronary atherosclerosis using electron-beam computed tomography.
Atherosclerosis
 
2006
;
185
:
177
–82.
29.
Stang A, Ahrens W, Jöckel KH. Control response proportions in population-based case-control studies in Germany.
Epidemiology
 
1999
;
10
:
181
–3.
30.
Stang A, Moebus S, Dragano N, et al. Baseline recruitment and analyses of nonresponse of the Heinz Nixdorf Study: identifiability of phone numbers as the major determinant of response.
Eur J Epidemiol
 
2005
;
20
:
489
–96.
31.
International Organization for Standardization. ISO 9001:2000. (http://www.iso.org). Accessed January
2001
.
32.
Hense HW, Koivisto AM, Kuulasmaa K, et al. Assessment of blood pressure measurement quality in the baseline surveys of the WHO MONICA project.
J Hum Hypertens
 
1995
;
9
:
935
–46.
33.
Chobanian AV, Bakris GL, Black HR, et al. The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report.
JAMA
 
2003
;
289
:
2560
–72.
34.
Bland JM, Altman DG. Comparing methods of measurement: why plotting difference against standard method is misleading.
Lancet
 
1995
;
346
:
1085
–7.
35.
Fleiss JL. Statistical methods for rates and proportions. 2nd ed. New York, NY: John Wiley & Sons, Inc,
1981
.
36.
SAS Institute, Inc. Software release 9.1. Cary, NC: SAS Institute, Inc,
2002
.
37.
Franklin SS, Larson MG, Khan SA, et al. Does the relation of blood pressure to coronary heart disease risk change with aging? The Framingham Heart Study.
Circulation
 
2001
;
103
:
1245
–9.
38.
Sesso HD, Stampfer MJ, Rosner B, et al. Systolic and diastolic blood pressure, pulse pressure, and mean arterial pressure as predictors of cardiovascular risk in men.
Hypertension
 
2000
;
36
:
801
–7.
39.
O'Brien E, Atkins N, Conroy R, et al. The Hawksley random zero sphygmomanometer. Comparison with mercury instrument is illogical. (Letter; author's reply).
BMJ
 
1993
;
307
:
562
–3.
40.
Whincup PH, Bruce NG, Cook DG, et al. The Dinamap 1846SX automated blood pressure recorder: comparison with Hawksley random zero sphygmomanometer under field conditions.
J Epidemiol Community Health
 
1992
;
46
:
164
–9.
41.
Jones DW, Appel LJ, Sheps SG, et al. Measuring blood pressure accurately. New and persistent challenges. (Editorial).
JAMA
 
2003
;
289
:
1027
–30.