Professional Documents
Culture Documents
JH-D-19-00010
Original Article
H
ypertension is a cardiovascular risk factor of out- J Hypertens 37:000–000 Copyright ß 2019 Wolters Kluwer Health, Inc. All rights
standing importance and accounts for up to 50% of reserved.
vascular risk [1]. Blood pressure (BP) has been DOI:10.1097/HJH.0000000000002237
Gotzmann et al.
Copyright © 2019 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
CE: Tripti; JH-D-19-00010; Total nos of Pages: 8;
JH-D-19-00010
(Schwarz GmbH & Co. KG, Kamen, Germany). The pres- TABLE 1. Clinical characteristics of study patients
sure transducer was attached to the patient’s heart by a fixed All patients (n ¼ 502)
suspension and remained there throughout the examina-
Age (years) 67.9 11.6
tion. Blood was then aspirated to remove any air bubbles
Female sex (n, %) 228 (45)
from the catheter. The catheter was then flushed with 10 ml BMI (kg/m2) 29 5.9
physiological saline solution and the measuring system was SBP (mmHg) 145 22
zeroed and calibrated. The natural frequency was con- DBP (mmHg) 80.2 14.6
firmed to be greater than 20 Hz, and the damping coeffi- Left ventricular ejection fraction (%) 54.3 12.4
cient was not less than 0.3. The procedure was repeated Hypertension (n, %) 454 (91)
Hypercholesterolemia (n, %) 356 (71)
when there was evidence of a damped curve. The mea-
Active smoker (n, %) 145 (29)
surement of systolic and diastolic central blood pressure Previous smoker (n, %) 103 (21)
values was digitally recorded. The data of at least 10 heart- Diabetes mellitus (n, %) 133 (26)
beats were averaged. Coronary artery disease (n, %) 287 (57)
Atrial fibrillation (n, %) 135 (27)
Noninvasive assessment of central aortic blood
pressure
The noninvasive blood pressure measurements were per- two-tailed t tests (for normally distributed variables) or a
formed under hemodynamically stable conditions by the Mann–Whitney U test (for nonnormally distributed varia-
SphygmoCor XCEL device and the Mobil-O-Graph NG bles), whenever appropriate. Dichotomic parameters were
device simultaneously and with the invasive central blood compared by chi-square test (Fisher’s exact test for inde-
pressure measurements. The sequence of the noninvasive pendent samples). P < 0.05 was regarded significant.
devices alternated randomly. In the noninvasive and inva-
sive examinations runs of bigeminy, trigeminy, or isolated
premature beats and the following compensatory beats
RESULTS
have been removed from analysis. In some cases, the Table 1 presents the epidemiological and medical charac-
devices reported error messages so that a sufficiently accu- teristics of the study population. Mean age was 67.9 11.6
rate central blood pressure measurement was not possible. ranging from 34 to 92 years. Three hundred and sixty-five
In these cases, the measurement was repeated a second patients (71%) had sinus rhythm, 135 (27%) suffered from
time with the device and, if necessary, a third time. If the atrial fibrillation at the time of the measurement procedure.
third measurement was also unsuccessful, no further Concomitant diseases constituted coronary artery disease
attempts were made. These measurements were defined (n ¼ 287, 57%), peripheral occlusive disease (n ¼ 45, 16%),
as ‘unsuccessful measurement’. and hypertension (n ¼ 454, 91%). The median number of
The recommendations of the ARTERY Society regarding antihypertensive drugs was ranging from 0 to 6, and con-
the sample characteristics and the statistical requirements tained diuretics, calcium-channel blockers, b-blockers, ACE
were fulfilled in this study [9]. inhibitors, AT1 blockers, alpha blockers, mineralocorticoid-
antagonists, and hydrazinophthalazine (Supplement Table
Statistics 1, http://links.lww.com/HJH/B145).
Results are presented as mean standard deviation (SD) in The invasive measurement of aortic blood pressure was
case of normal distribution. Continuous variables without performed in all patients as part of the cardiac catheter
normal distribution are presented as median (first quartile, examination. The simultaneous measurement of the non-
third quartile). Pearson correlation analyses were per- invasive central blood pressure using the SphygmoCor
formed for noninvasive vs. invasive values. Fisher’s z XCEL device was successful in 498 of 502 patients (99%).
was calculated to compare the correlation coefficients. Measurements using Mobil-O-Graph NG device were suc-
Moreover, comparison of systolic and diastolic central cessful in 441 of 502 patients (88%; P ¼ 0.451). In the
blood pressure values of the two test devices and the remaining patients, no values could be determined even
simultaneous invasively assessed reference cBP was per- after the third measurement.
formed by Tukey mean-difference plots (Bland–Altman Measurement failure (unsuccessful measurement or
plots). In accordance with Krouwer [10], the x-axis presents extreme outliers in the Bland–Altman plot) occurred in
the results of the invasive measurement as the gold standard 10 patients examined with the SphygmoCor XCEL device
method. Bias and the limits of agreement (bias and in 94 patients examined with Mobil-O-Graph NG
2 standard deviation) are reported. For the calculation device. Patients with measurement failure and successful
of the extreme outliers on Bland–Altman plot, the upper examination were compared in clinical characteristics. The
and lower quartiles (Q1 and Q3) of the systolic and diastolic results are presented in Supplement Tables 2 (http://link-
bias were calculated. A point beyond an outer fence (lower s.lww.com/HJH/B145) and 3 (http://links.lww.com/HJH/
outer fence: Q1 3 interquartile range, upper outer B145). In SphygmoCor XCEL device the invasive diastolic
fence: Q3 þ 3 interquartile range) was considered an central blood pressure was significantly lower in the group
extreme outlier. Extreme outliers in the Bland–Altman plot with measurement failure than in the group of patients with
(systolic or diastolic bias) and unsuccessful measurement successful measurement. In Mobil-O-Graph NG device, the
were considered as measurement failure. SBP was significantly higher and atrial fibrillation occurred
Comparison of each method’s findings in patients with more frequently in the group with measurement failure
and without atrial fibrillation was performed by unpaired compared with the patients with successful measurement.
Gotzmann et al.
TABLE 2. Results of simultaneous measurement of invasive and noninvasive central blood pressure with the SphygmoCor XCEL and
Mobil-O-Graph NG device
SphygmoCor measurements (n ¼ 498) Mobil-O-Graph measurements (n ¼ 441)
Invasive systolic cBP (mmHg) 131 21 Invasive systolic cBP (mmHg) 132 21
Invasive diastolic cBP (mmHg) 76.4 11.9 Invasive diastolic cBP (mmHg) 77.8 11.8
Noninvasive systolic cBP (mmHg) 127 20 Noninvasive systolic cBP (mmHg) 126 20
Noninvasive diastolic cBP (mmHg) 77 13 Noninvasive diastolic cBP (mmHg) 81.4 13.8
only the data from successful measurements of noninvasive devices are presented. cBP, central blood pressure.
Table 2 provides mean invasively and noninvasively the correlation coefficients in systolic and diastolic cBP
assessed cBP values. The performance of the different were significantly higher for the SphygmoCor XCEL device
noninvasively cBP measurement techniques was analyzed compared with the Mobil-O-Graph NG device (P < 0.001
using two different approaches: correlations between inva- each).
sively and noninvasively measured values, assessment of In the overall study population, the mean systolic bias of
bias and limits of agreement in a Bland–Altman analysis. SphygmoCor XCEL device to invasively assessed cBP was
Figure 1 illustrates the correlations between invasively 5.0 7.7 mmHg. The systolic bias of the Mobil-O-Graph
and noninvasively measured values. Correlations were NG device was slightly but significantly larger
highly significant for both SBP and DBP with both devices (6.0 10.4 mmHg, P ¼ 0.011). The diastolic bias of Sphyg-
(P < 0.001 each). With each device, correlation coefficients moCor XCEL device was significantly lower than the Mobil-
were higher for systolic than for diastolic cBP. Furthermore, O-Graph NG device bias (0.5 6.2 vs. 3.6 8.3 mmHg;
FIGURE 1 Relationship between systolic and diastolic central blood pressure comparing measurements of SphygmoCor device and Mobil-O-Graph device and invasively
measured central blood pressure in the overall study population (patients with and without atrial fibrillation).
Copyright © 2019 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
CE: Tripti; JH-D-19-00010; Total nos of Pages: 8;
JH-D-19-00010
FIGURE 2 Tukey mean-difference plots (Bland–Altmann plots) of systolic and diastolic central blood pressure comparing SphygmoCor device and Mobil-O-Graph device to
the goldstandard of invasively measured central blood pressure in the overall study population (patients with and without atrial fibrillation). Dotted lines provide mean bias
and 95% limits of agreement (2 standard deviation).
P < 0.001). Figure 2 provides the corresponding Tukey did not signficantly change the systolic bias of the Mobil-O-
mean-difference plots including bias and limits of agree- Graph NG device (6.2 12.3 mmHg, P ¼ 0.862) or the
ment (95% confidence interval, 2 standard deviation). diastolic bias (4.7 9.6 mmHg; P ¼ 0.095). Figure 4 pro-
In a second approach, we investigated the impact of vides the corresponding Tukey mean-difference plots.
atrial fibrillation on the accuracy of noninvasive cBP mea-
surement. The corresponding correlation analyses are DISCUSSION
shown in Fig. 3. Systolic and diastolic R2 values did not
significantly differ from those of the overall study popula- Until recently, noninvasive cBP measurement was not
tion with the SphygmoCor device (P ¼ 0.220 and P ¼ 0.750, possible in an automated manner and necessitated an
respectively), and with the Mobil-O-Graph device observer trained in applanation tonometry. The present
(P ¼ 0.062 and 0.224, respectively). In analogy to the overall study shows that the cBP values obtained by automated
study population, systolic and diastolic correlation coeffi- oscillometric cuff-based pulse contour analysis highly cor-
cients were higher in SphygmoCor device compared with relate to invasive measurements. The noninvasive devices
Mobil-O-Graph device in patients with atrial fibrillation thereby slightly underestimate systolic and slightly overes-
(P ¼ 0.003 and 0.002, respectively). The mean systolic bias timate diastolic cBP. This inaccuracy (systematic underesti-
of SphygmoCor XCEL device (5.7 8.9 mmHg) was mation of SBP but overestimation of DBP) has been
almost identical to that of the overall study population described before and may be the consequence of the
(P ¼ 0.191), whereas it increased by 1 mmHg diastolic calibration method [11]. In the present study, both devices
(1.5 6.9 mmHg, P ¼ 0.042). In analogy, atrial fibrillation made use of a traditional calibration based on SBP and DBP.
Gotzmann et al.
FIGURE 3 Relationship between systolic and diastolic central blood pressure comparing measurements of SphygmoCor device and Mobil-O-Graph device and invasively
measured central blood pressure in patients with atrial fibrillation.
This method, however, may underestimate true systolic With a number of 502 participants, the current study is
brachial and thereby true systolic cBP. In the meantime, the largest validation study of noninvasive cBP measure-
an alternative calibration using mean instead of SBP has ment techniques so far [6]. There are some validation
become available for the Mobil-O-Graph. There are some studies on the prior SphygmoCor system using applanation
hints that this calibration may indeed lead to a further tonometry [14–16]. To the best of our knowledge, there are
increase of the device’s accuracy [12]. only one small invasive validation study on the current
The mean differences of 5–6 mmHg in systolic and 1– oscillometric SphygmoCor device (36 patients) [17] and one
4 mmHg in diastolic cBP in our study appear acceptable small study on the Mobil-O-Graph (30 individuals) [18]. In
from a clinical point of view. Nevertheless, our findings the latter study, the device underestimated invasively mea-
should encourage manufacturers to adapt their general sured central BP by 3 mmHg. Noteworthy, the manufacturer
transfer functions. Mean biases in systolic and diastolic modified the transfer function since the publication of this
cBP were slightly but significantly lower for SphygmoCor report. The present study thereby constitutes the first vali-
XCEL compared with Mobil-O-Graph NG (P ¼ 0.011 and dation study of the device in its current form. Noteworthy,
P < 0.001, respectively). The separate analysis of those our study reveals still an underestimation of systolic cBP.
patients in the upper and lower quartiles (outliers) identi- The SphygmoCor device showed a slightly but significantly
fied a low diastolic (SphygmoCor XCEL) and a high systolic higher accuracy in the estimation of systolic and diastolic
cBP (Mobil-O-Graph) as risk factors for an inaccurate cBP compared with the Mobil-O-graph device regarding
measurement. A high SBP and a low DBP are indicators both correlation and biases.
of increased aortic stiffness and isolated systolic hyperten- A recent review on the accuracy of noninvasive mea-
sion. We have previously demonstrated that the accuracy of surement techniques identified 22 eligible studies, which
BP measurement is reduced in this population [13]. validated 11 different commercial devices in 808 study
Copyright © 2019 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
CE: Tripti; JH-D-19-00010; Total nos of Pages: 8;
JH-D-19-00010
FIGURE 4 Tukey mean-difference plots (Bland–Altmann plots) of systolic and diastolic central blood pressure comparing SphygmoCor device and Mobil-O-Graph device to
the goldstandard of invasively measured central blood pressure in patients with atrial fibrillation. Dotted lines provide mean bias and 95% limits of agreement (2 standard
deviation).
participants. [6]. The number of 502 in the present study cardiovascular endpoints made use of brachial and not cBP
constitutes the largest study population investigated so far. so far. The present oscillometric technology could be easily
The vast majority of validation studies in this review used implemented in cardiovascular large-scale studies. These
applanation tonometry for pulse wave analysis. The mean studies could investigate, whether cBP is indeed superior to
error in the estimation of systolic cBP was 4.5 mmHg. brachial BP in the prediction of cardiovascular endpoints.
Thus, the diagnostic accuracy of the present two oscillo- The most crucial limitation of the present study is a
metric devices was almost exactly as accurate than the potential selection bias. Comparison of noninvasively vs.
traditional tonometric devices. A measurement procedure invasively measured aortic BP was performed in patients
takes about 1–2 min including both peripheral BP mea- undergoing cardiac catheterization and not in healthy indi-
surement and pulse wave analysis. This time is only slightly viduals. The same is true for prior validation studies and the
longer than a standard BP measurement, and may therefore majority of the manufacturers’ approaches in the develop-
facilitate an integration in daily clinical practice. The auto- ment of the devices. This limitation is inevitable to avoid
mated procedure does not necessitate an intense training of aortic catheterization of healthy probands. Noteworthy, 159
staff anymore and the results are far less observer-depen- patients (32%) of the present study population with pre-
dent than previous tonometric approaches, which required served left ventricular ejection fraction and no severe heart
placement of the tip of a hand-held high-fidelity tonometer valve disease turned out to have an unremarkable coronary
on the patient’s artery. status. The mean systolic/diastolic biases in this subgroup
Despite the increasing data on the prognostic value of were 4.1 6 mmHg/0.1 6.2 mmHg for the SphygmoCor
cBP, its measurement is not recommended by current device and -4.7 7.6 mmHg/2.4 6.7 mmHg for the Mobil-
hypertension guidelines [19,20]. This decision is mainly O-Graph device. Thus, the present selection bias may be of
based on the fact that randomized controlled trials with subordinate clinical relevance.
Gotzmann et al.
The present study shows that two current automated 7. Cheng HM, Lang D, Tufanaru C, Pearson A. Measurement accuracy of
noninvasively obtained central blood pressure by applanation tonom-
oscillometric BP monitors are able to assess cBP as accurate etry: a systematic review and meta-analysis. Int J Cardiol 2013;
as former tonometric techniques. The automated measure- 167:1867–1876.
ment procedure shortens the time demand and minimizes 8. Millasseau S, Agnoletti D. Noninvasive estimation of aortic blood
observer errors. Manufacturers should feel encouraged to pressures: a close look at current devices and methods. Curr Pharm
adapt the transfer function to avoid the slight underestima- Des 2015; 21:709–718.
9. Sharman JE, Avolio AP, Baulmann J, Benetos A, Blacher J, Blizzard CL,
tion of systolic cBP in the future. This technique is able to et al. Validation of noninvasive central blood pressure devices:
substantially facilitate the use of cBP measurement in future ARTERY Society task force consensus statement on protocol standard-
studies and daily clinical practice. ization. Eur Heart J 2017; 38:2805–2812.
10. Krouwer JS. Why Bland-Altman plots should use X, not (Y þ X)/2 when
X is a reference method. Stat Med 2008; 27:778–780.
ACKNOWLEDGEMENTS 11. Picone DS, Schultz MG, Otahal P, Aakhus S, Al-Jumaily AM, Black JA,
et al. Accuracy of cuff-measured blood pressure: systematic reviews
We would like to thank the cardiac catheter team for their and meta-analyses. J Am Coll Cardiol 2017; 70:572–586.
exceptional support. 12. Wassertheurer S, Baumann M. Assessment of systolic aortic pressure
and its association to all cause mortality critically depends on waveform
Conflicts of interest calibration. J Hypertens 2015; 33:1884–1888; discussion 9.
There are no conflicts of interest. 13. Westhoff TH, Schmidt S, Meissner R, Zidek W, van der Giet M. The
impact of pulse pressure on the accuracy of wrist blood pressure
measurement. J Hum Hypertens 2009; 23:391–395.
REFERENCES 14. Davies JI, Band MM, Pringle S, Ogston S, Struthers AD. Peripheral
1. Lawes CM, Vander Hoorn S, Rodgers A, International Society of blood pressure measurement is as good as applanation tonometry
Hypertension. Global burden of blood-pressure-related disease, at predicting ascending aortic blood pressure. J Hypertens 2003;
2001. Lancet 2008; 371:1513–1518. 21:571–576.
2. Vlachopoulos C, Aznaouridis K, O’Rourke MF, Safar ME, Baou K, 15. Cloud GC, Rajkumar C, Kooner J, Cooke J, Bulpitt CJ. Estimation of
Stefanadis C. Prediction of cardiovascular events and all-cause mortal- central aortic pressure by SphygmoCor requires intra-arterial periph-
ity with central haemodynamics: a systematic review and meta-analy- eral pressures. Clin Sci (Lond) 2003; 105:219–225.
sis. Eur Heart J 2010; 31:1865–1871. 16. Laugesen E, Rossen NB, Peters CD, Mæng M, Ebbehøj E, Knudsen ST,
3. Cheng HM, Chuang SY, Sung SH, Yu WC, Pearson A, Lakatta EG, et al. et al. Assessment of central blood pressure in patients with type 2
Derivation and validation of diagnostic thresholds for central blood diabetes: a comparison between SphygmoCor and invasively mea-
pressure measurements based on long-term cardiovascular risks. J Am sured values. Am J Hypertens 2014; 27:169–176.
Coll Cardiol 2013; 62:1780–1787. 17. Shoji T, Nakagomi A, Okada S, Ohno Y, Kobayashi Y. Invasive
4. Dahlof B, Sever PS, Poulter NR, Wedel H, Beevers DG, Caulfield M, validation of a novel brachial cuff-based oscillometric device (Sphyg-
et al., ASCOT Investigators. Prevention of cardiovascular events with moCor XCEL) for measuring central blood pressure. J Hypertens 2017;
an antihypertensive regimen of amlodipine adding perindopril as 35:69–75.
required versus atenolol adding bendroflumethiazide as required, in 18. Weber T, Wassertheurer S, Rammer M, Maurer E, Hametner B, Mayer
the Anglo-Scandinavian Cardiac Outcomes Trial-Blood Pressure Low- CC, et al. Validation of a brachial cuff-based method for estimating
ering Arm (ASCOT-BPLA): a multicentre randomised controlled trial. central systolic blood pressure. Hypertension 2011; 58:825–832.
Lancet 2005; 366:895–906. 19. Williams B, Mancia G, Spiering W, Agabiti Rosei E, Azizi M, Burnier M,
5. Williams B, Lacy PS, Thom SM, Cruickshank K, Stanton A, Collier D, et al., ESC Scientific Document Group. 2018 ESC/ESH Guidelines for
et al., CAFE Investigators; Anglo-Scandinavian Cardiac Outcomes Trial the management of arterial hypertension. The Task Force for the
Investigators; CAFE Steering Committee and Writing Committee. Dif- Management of Arterial Hypertension of the European Society of
ferential impact of blood pressure-lowering drugs on central aortic Cardiology and the European Society of Hypertension. J Hypertens
pressure and clinical outcomes: principal results of the Conduit Artery 2018; 36:1953–2041.
Function Evaluation (CAFE) study. Circulation 2006; 113:1213–1225. 20. Whelton PK, Carey RM, Aronow WS, Casey DE Jr, Collins KJ, Dennison
6. Papaioannou TG, Karageorgopoulou TD, Sergentanis TN, Protogerou Himmelfarb C, et al. 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/
AD, Psaltopoulou T, Sharman JE, et al. Accuracy of commercial devices ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evalu-
and methods for noninvasive estimation of aortic systolic blood pres- ation, and Management of High Blood Pressure in Adults: a report of the
sure a systematic review and meta-analysis of invasive validation American College of Cardiology/American Heart Association Task Force
studies. J Hypertens 2016; 34:1237–1248. on Clinical Practice Guidelines. J Am Coll Cardiol 2018; 71:e127–e248.
Copyright © 2019 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.