Introduction

Clinic (office) blood pressure (BP) has traditionally been measured manually by healthcare professionals. However, observed BP measurements have the well-known limitation of the white coat effect, where values recorded in the clinic are higher than at home [1], meaning that ‘usual’ BP is not always indicated by clinic-based readings. Twenty-four hour ambulatory BP monitoring (ABPM) is now the gold standard in BP measurement and is a valuable indicator of cardiovascular risk [2]. Indeed, ABPM, which provides a more reliable measure of ‘usual’ BP than clinic-based readings, has superior prognostic value compared with clinic BP [3] and is recommended by current guidelines as part of the routine assessment for diagnosing and managing hypertension [4,5,6]. However, ABPM can cause disruption to daily activities and relies on patient compliance. Thus, ABPM remains impractical for repeated or regular BP monitoring. Clearly, the ideal sphygmomanometer for recording BP routinely would combine the practicality of manual measurements with the prognostic value and lack of white coat effect associated with ambulatory readings.

Automated office BP (AOBP) is gaining recognition as an alternative way of measuring BP in the clinic. Sphygmomanometers that measure AOBP allow multiple BP readings to be recorded with the patient at rest and the operator absent from the examination room [7]. This is thought to minimise the white coat effect, and thus yield BP values that are closer to usual BP than observed office readings [8]. In addition, evidence suggests that AOBP is a good indicator of target organ damage, producing a similar correlation with left ventricular mass as ambulatory readings [9]. More recently, AOBP has been used in major hypertension trials such as the Systolic Blood Pressure Intervention Trial (SPRINT) [10], although recent data confirm that both attended and unattended BP readings were included in the analyses [11], and AOBP was recently included in the Canadian Hypertension Education Program guidelines as the preferred method of measuring in-office BP [5].

Several devices that can record AOBP have been validated for accuracy against mercury sphygmomanometry, including BpTru (BpTru Medical Devices, Coquitlam, BC, Canada) [12], Omron HEM-907 (Omron Healthcare, Kyoto, Japan) [13] and Microlife WatchBP (Microlife AG, Widnau, Switzerland) [14]. AOBP readings taken by both BpTru and Microlife WatchBP devices correlate well with ABPM [15, 16] and the Omron HEM-907 records similar BP readings to BpTru [17]. A recent sub-study from SPRINT compared ambulatory BP and clinic-based readings [18]. However, ambulatory and clinic readings may have been taken up to 3 weeks apart, and it is not clear whether the clinic readings were attended (observer present) or not. Therefore, the aim of our study was to evaluate the Omron HEM-907 device further, by comparing, directly, automated BP with observed office and ambulatory BP readings.

Methods

The study sample consisted of 108 participants who were recruited from hypertension clinics and a volunteer database held by the Vascular Research Clinic at Addenbrooke’s Hospital, Cambridge. Informed consent was given by all participants. Basic demographic data including age, sex, height, weight and current medications were recorded for each participant. After five minutes of seated rest in a quiet examination room, both OOBP and AOBP recordings were taken. The study operator was present during this period of rest. The order of OOBP and AOBP was randomised between participants, and there was no additional rest period between measurement modes. Following the office recordings, an ABPM was fitted to take daytime BP readings. Participants were followed up in the clinic exactly 1 week later, where OOBP and AOBP readings were repeated in the same order, after 5 minutes of seated rest.

Omron HEM-907 measurements

All office BP measurements were taken with the Omron HEM-907 device. Participants were seated comfortably and their non-dominant arm was used. Different cuff sizes were available, depending on arm circumference. For observed BP measurements, three single observed recordings separated by 1-minute intervals were made with the study operator present in the examination room. The first reading was discarded and the mean of the two remaining readings formed the OOBP. For automated measurements, three automated readings separated by two minute intervals were taken using the Omron HEM-907 device, whereas the study operator was out of the examination room. The average of these readings formed the AOBP.

Ambulatory monitoring

ABPM readings were recorded using the Mobil-O-Graph device (I.E.M., Stolberg, Germany) [19]. Six different Mobil-O-Graph devices were used, all of which were factory calibrated on a regular basis (at least once per year). The non-dominant arm was again used and different cuff sizes were available depending on arm circumference. The device was set to record BP at 30 min intervals during daytime hours and participants were instructed to wear the device for a minimum of 8 hours during the day and engage in routine daily activities. The mean AABP was then calculated.

Comparing Mobil-O-Graph and Omron HEM-907 devices

In order to compare, directly, the performance of the Omron HEM-907 and Mobil-O-Graph devices under standardised conditions, BP readings recorded by both devices were investigated in a further 30 volunteers. Age and sex were recorded for each participant. After 5 minutes of seated rest, BP was recorded in the non-dominant arm with both the Mobil-O-Graph and Omron HEM-907 devices, with the study operator present in the examination room. The same Mobil-O-Graph device was used for all recordings. The order in which the devices were used was randomised. Two recordings were made with each device and the average for each device was calculated.

Data analysis

Data were analysed using SPSS (Version 22) and are presented as means ± standard deviations. Paired-samples t tests were used for comparison of AOBP with both OOBP and AABP readings and Bland–Altman analyses were conducted to assess agreement between measurement modes [20]. Pearson correlation coefficients (r) were also calculated to examine continuous associations between AOBP, OOBP and AABP.

Results

Subject characteristics

Subject characteristics are presented in Table 1. There were 108 participants with an average age of 53 ± 21 years (range 20–85 years). Fifty-five participants (51%) were male and the average BMI was 26.7 ± 4.9 kg/m2. Fifty-six participants (52%) were hypertensive and 30 (28%) were taking antihypertensive medication. Two participants were lost to follow-up at the second visit. An average of 19 daytime ambulatory BP readings were recorded per participant.

Table 1 Subject characteristics
Full size table

Comparison of mean BP readings

Average BP readings for each of the measurement modalities are shown in Table 2. On both visits AOBP was significantly lower than OOBP for both systolic BP (SBP) and diastolic BP (DBP), with a mean difference of −3 ± 7/−2 ± 4 mmHg for visit 1 and −4 ± 7/−3 ± 8 mmHg for visit 2 (P < 0.001 for all comparisons). AOBP was also significantly lower than AABP for both SBP and DBP, with a mean difference of −2 ± 11/−7 ± 10 mmHg (P = 0.02 and P < 0.001 for SBP and DBP, respectively). There was a high degree of correlation between AOBP and AABP (r = 0.75 and r = 0.64 for SBP and DBP, respectively, P < 0.001 for both) and between OOBP and AABP (r = − 0.66 for both SBP and DBP, P < 0.001 for both). There was no evidence of bias in the measurements modes as demonstrated by Bland–Altman plots for AOBP versus AABP (Fig. 1) and OOBP versus AABP (Fig. 2). Repeat readings for AOBP and OOBP were compared between visits 1 and 2. Mean differences were 1 ± 13/2 ± 8 mmHg for OOBP (P = 0.25 and P = 0.2 for SBP and DBP, respectively) and 2 ± 12/3 ± 10 mmHg for AOBP (P = 0.07 and P = 0.007 for SBP and DBP, respectively).

Table 2 Mean differences and correlations between measurement modes
Full size table
Fig. 1
figure 1

Bland–Altman plot showing agreement between AOBP and AABP readings (visit 1) for systolic blood pressure

Full size image
Fig. 2
figure 2

Bland–Altman plot showing agreement between OOBP and AABP readings (visit 1) for systolic blood pressure

Full size image

The influence of randomising the order of observed and automated BP readings is presented in Table 3. Irrespective of the order of measurements, AOBP was lower than OOBP for both SBP and DBP, although the differences were not statistically significant when OOBP was measured first. Similar results were obtained at the repeat visit (data not shown). In addition, averaging all three OOBP readings rather than the second and third readings did not meaningfully influence the results (average of three OOBP readings: 134 ± 18 and 77 ± 12 mmHg for SBP and DBP, visit 1; 132 ± 17 and 75 ± 12 mmHg for SBP and DBP, visit 2).

Table 3 Mean BP values for OOBP and AOBP at visit 1, with respect to the order taken
Full size table

Comparing Mobil-O-Graph with Omron HEM-907

Mobil-O-Graph and Omron HEM-907 readings were compared under standardised conditions in the clinic, in 30 volunteers. The average age was 58 ± 20 years (range 25—86 years) and there were 21 females (70%). Average values for SBP and DBP are shown in Table 4. The Mobil-O-Graph device recorded significantly lower SBP values (mean difference − 4 ± 9 mmHg, P = 0.016) but significantly higher DBP values (mean difference of 6 ± 6 mmHg, P < 0.001) compared with the Omron device. However, both SBP and DBP readings were highly correlated between devices (r = 0.92 and r = 0.86, respectively, P < 0.001 for both).

Table 4 Comparison of Mobil-O-Graph and Omron HEM-907 devices under standardised conditions
Full size table

Discussion

The current study aimed to compare automated BP measured using the Omron HEM-907 device with both observed office and ambulatory measurements. Our main findings were that automated BP was significantly lower than observed BP, and that automated BP was also significantly lower than ambulatory BP, albeit with a high correlation between measurement modalities. These data suggest that automated measurement of BP with the Omron HEM-907 is not associated with a white coat effect in the clinic setting and provides a suitable and convenient alternative for assessing usual BP.

White coat effect

In the current study, the same device (Omron HEM-907) was used to record BP in both automated and manual BP modes, and the absence of a white coat effect with automated measurements was consistent between study visits. This suggests that any differences were more likely to result from the measurement modality, i.e., the presence or absence of an operator in the examination room, rather than any inherent variability in the device itself, although the use of different measurement intervals for observed (one minute) and automated (two minutes) readings may have introduced additional variability. We deliberately randomised the order in which observed or automated readings were taken, so as to reduce the influence of any potential ‘order effect’. The lower SBP seen with automated versus observed measurements in our study was smaller in magnitude than that reported with the BpTru device [21]. However, all participants in the study using BpTru were known hypertensives and the average SBP on manual measurement was higher than in our study (151 ± 10mHg [21]). Moreover, the study utilising the BpTru device measured automated BP following the manual measurement, which may have resulted in an order effect [22]. Nevertheless, the reduction in manual office DBP with BpTru [21] was similar to the difference between observed and automated measurements seen in our study.

Automated versus ambulatory BP

We observed that automated BP was also significantly lower than daytime ambulatory BP, for both SBP and DBP, although this difference was small. Although this is in agreement with a previously published finding using the BpTru device [21], it has been suggested that allowing a period of rest prior to automated measurements, as done in the current study, may result in BP values, which are lower than ambulatory values [22]. Our randomisation procedure also meant that in half of the participants, automated measurements followed observed measurements and, therefore, an even longer rest period. However, as demonstrated in Table 3, the ‘order effect’ was much greater for observed measurements and so is unlikely to explain the overall lower automated versus ambulatory BP values obtained in our study. The study participants were relatively young, with an average age of 53 ± 21 years, and it is possible that greater activity levels during the daytime ambulatory monitoring period may have contributed to the higher BP values yielded by ambulatory, compared with automated BP monitoring.

The difference between automated and ambulatory BP readings in this study could also result from inherent differences between the measurement devices (Omron HEM-907 and Mobil-O-Graph). Indeed, a simple head-to-head comparison between the devices, under carefully standardised conditions, showed that the Mobil-O-Graph yielded lower values of SBP (by ~ 4 mmHg), but higher values of DBP (by ~ 6 mmHg). Assuming consistency and applying this offset to our observed AABP values would yield mean AABP values of 137/76 mmHg. The differences between devices could relate to several factors but most likely involves the measurement algorithm employed by each device. Our data are not supported by previously published studies with the Mobil-O-Graph, which demonstrate reasonable equivalence with the Spacelabs 90207 ambulatory BP monitor, particularly for DBP [23, 24]. However, previous studies [13, 25, 26] and a recent review [27] highlight a tendency for the Omron 907 device to record lower values of DBP. Regardless of the source of the discrepancy between devices, our data suggest that we may have under-estimated the true difference between automated and ambulatory readings in our study, at least for SBP, and support the notion [22] that a period of rest is probably not required prior to undertaking automated BP readings.

BP measurements from three different sphygmomanometers have now been compared against ambulatory BP [15, 16, 18], with our study being the first to compare readings between the Omron HEM-907 and ambulatory BP in a well-controlled setting, where both observed and automated readings were considered. In our study, automated BP measured with the Omron HEM-907 showed high correlations with ambulatory measurements for SBP and DBP, which compares favourably with results obtained with the Microlife WatchBP and BpTru devices (r = 0.819/0.801 [16] and r = 0.571/0.610[15], respectively). Although our sample size was considerably lower than the BpTru and SPRINT ABPM studies (n = 108 versus n = 481 and n = 897), it was similar to that used with Microlife WatchBP (n = 100). However, the ambulatory monitor used in each study has differed by manufacturer (A&D[15], Space Labs [16] and, in the current study, Mobil-O-Graph). For a more accurate comparison between the different measurement modalities, a single ambulatory device should be used. Ideally such a device would be able to measure BP in both ambulatory and automated modes so as to definitively test the hypothesis that automated and ambulatory measurement modes yield similar BP readings.

Limitations

Our study had a number of limitations. First, we included a relatively low number of hypertensive participants in comparison with other studies. However, we wanted to investigate the Omron HEM-907 device across a range of BP values, rather than just in the hypertensive range. In addition, the observed BP was the average of the second and third readings, with the first reading discarded, whereas automated BP was the average of all three readings. However, re-analysis of the data based on all three readings made no material difference to the results. Of note, observed readings were taken at 1-minute intervals, whereas automated readings were taken at 2-minute intervals, which may have introduced additional variability. Finally, we performed sequential readings of observed and automated BP during the same clinic visit, and allowed a rest period prior to the automated measurement. It has previously been suggested that allowing a rest period may result in automated BP values that are significantly lower than ambulatory values [22]. Therefore, we randomised the order of readings to reduce any potential order effects associated with sequential BP measurement. Performing automated or observed office BP readings in a randomised order before or after the ambulatory measurements would have provided an alternative approach.

Observed BP measurements in the clinic setting are subject to the white coat effect, yielding higher values than ambulatory readings, which are currently considered the gold standard for BP measurement. However, the utility of automated BP readings is becoming increasingly recognised. We have demonstrated that automated SBP and DBP readings measured with the Omron HEM-907 device were somewhat lower than observed BP measurements, an effect that was reproducible over two clinic visits. Moreover, automated BP readings correlated well with daytime ambulatory BP readings. Overall, our findings suggest that automated BP measured using the Omron HEM-907 provides a useful alternative to observed clinic measurements, which would likely be higher in routine office practice.