eabpS

PImr

1731JACC Vol. 58, No. 16, 2011 CorrespondenceOctober 11, 2011:1729–32

Letters to the Editor

R

Invasive Validation of theN-Point Moving Average MethodWe read with great interest the recent paper by Williams et al. (1)introducing and validating a new method to derive central aorticsystolic blood pressure (cSBP). We fully agree that simplermethods to derive cSBP will facilitate the distribution of thisimportant measure into large clinical trials and, eventually, intoclinical routine. However, some questions about the invasivevalidation need to be addressed.

Typically, in studies of this kind, mean differences betweenmeasured and calculated cSBPs are small (1 to 2 mm Hg), and SDsof differences are in the range of 7 to 11 mm Hg (2,3). Williamst al. (1) present the data as mean � SE, which is unusual. SE ispproximately a factor square root(n) smaller than SD, which muste kept in mind when interpreting the results. In the firstaragraph of the section on invasive validation, the authors use theE on the basis of n � 20 (invasive cSBP 139.6 � 4.3 mm Hg)

and then proceed to use the SE on the basis of n � 200 (invasivecSBP 139.6 � 1.4 mm Hg) and stay unclear when presenting thedifferences between calculated cSBP and invasive cSBP (�0.41 �2.5 mm Hg). Supposing again n � 20 for SE, the usualpresentation of these data on the basis of mean � SD leads tomean difference of �0.41 mm Hg and an SD of 11.2 mm Hg,which would be in line with the published literature.

The presentation of data stays unusual for Figure 5 of theirpaper (1). The assumption that data based on multiple samplingwindows of 10 s, using the same calibration, are independent isquestionable. Such data may not be suitable for regressionanalysis and provide misleading coefficients and p values. Thisis even visually unveiled by the vertical data clustering along theregression line (Fig. 5A of their paper [1]). It would beinformative to see the corresponding Bland-Altman plot on aper-patient basis.

To summarize, we have significant questions about the presen-tation of the results of the invasive part of the validation study.

Siegfried Wassertheurer, MDBernhard Hametner, DI*Thomas Weber, MDBernd Eber, MD

*Cardiology DepartmentKlinikum Wels-GrieskirchenGrieskirchnerstrasse 42Wels, Upper Austria A-4600AustriaE-mail: thomas.weber3@liwest.at

doi:10.1016/j.jacc.2011.05.055

lease note: Drs. Wassertheurer and Hametner are employees of the Austriannstitute of Technology, developing new methods for noninvasive hemodynamiconitoring. Drs. Weber and Eber have reported that they have no relationships

elevant to the contents of this letter to disclose.

EFERENCES

1. Williams B, Lacy PS, Yan P, et al. Development and validation of anovel method to derive central aortic systolic pressure from the radialpressure waveform using an N-point moving average method. J AmColl Cardiol 2011;57:951–61.

2. Smulyan H, Siddiqui DS, Carlson RJ, et al. Clinical utility of aorticpulses and pressures calculated from applanated radial-artery pulses.Hypertension 2003;42:150–5.

3. Cheng HM, Wang KL, Chen YH, et al. Estimation of central systolicblood pressure using an oscillometric blood pressure monitor. Hyper-tens Res 2010;33:592–9.

Reply

We thank Dr. Wassertheurer and colleagues for their interest inour study (1) and note their agreement about the importance ofdeveloping simpler noninvasive methods for deriving central aorticsystolic blood pressure (CASP) in man. They comment on thelevel of agreement between invasive and noninvasive measure-

-20

-15

-10

-5

0

5

10

15

20

100 120 140 160 180 200

lartneC evisavnI sunim PSAC FT

GSB

P (m

mH

g)

Average CASP mmHg

2SD = 10.0 mmHg

- 1.1 mmHg

A

-20

-15

-10

-5

0

5

10

15

20

100 120 140 160 180 200

NPM

A-C

ASP

min

us In

vasi

ve C

entr

al

SBP

(mm

Hg)

Average CASP mmHg

2SD = 5.0 mmHg

- 0.4 mmHg

B

Figure 1 Bland-Altman Plots Comparing Invasive Aortic RootSystolic Pressure With Noninvasively Derived CASP

(A) All radial pressure waveform blocks calibrated to the initial brachial bloodpressure and processed using the AtCor processing algorithm (GTF CASP); (B)initial waveform block calibrated to brachial blood pressure with auto-updatingof subsequent waveform blocks using the A-pulse device and processed usingan N-point moving average (NPMA-CASP). Data show mean difference (dashedline) together with 2 SDs of the mean difference (dotted lines). CASP � cen-tral aortic systolic blood pressure.

Ab

tttmn

*PPNCC

*

UCGGLUE

R

1732 Correspondence JACC Vol. 58, No. 16, 2011October 11, 2011:1729–32

ments and suggest that analyzing multiple 10-s sampling windowsusing the same noninvasive initial calibration values is inappropri-ate and may have influenced our results.

Calibration of radial pressure waves to a single brachial bloodpressure (BP) measurement in the noninvasive measurement ofCASP is standard practice for many devices. However, the A-pulsetonometer (HealthSTATS International, Singapore) used in ourstudy operates in a different way. After calibration of the initial10-s radial waveform block to brachial BP, pulse wave height issubsequently scaled automatically to generate updated brachialBPs. This method allows the device to update brachial BP on ablock-by-block basis across the sampling period, accounting fornatural fluctuations in brachial BP over time (2). This techniquepotentially reduces scatter in the agreement between invasive andnoninvasive measurements, largely accounting for the improvedaccuracy reported. Indeed, it is remarkable that a single brachialBP calibration of initial radial pressure waves yielded such goodresults in previous studies over extended sampling periods.

Mindful of the above, we analyzed the influence of calibratingall 10-s data blocks acquired during sampling to the initial brachialBP, which constitutes standard practice for all other devices. Inthis, we also processed radial waveforms using the algorithm of anestablished method (SphygmoCor, AtCor Medical, West Ryde,Australia). Bland-Altman comparison of such processed datarevealed a wider degree of scatter (p � 0.01) compared with thatreported in our study (1) (Fig. 1). This finding suggests thatwaveform calibration drift across the sampling period is likely tocontribute to greater inaccuracy in deriving CASP in other studies.

With regard to data presentation for t test comparison in ourstudy (1), we presented mean � SE values for invasive andnoninvasive data from our 20 subjects, both as mean per subjectand for each individual data block. By contrast, data for Bland-

Altman comparisons are frequently presented as mean � SD.

ccordingly, this format was followed in presenting the differenceetween invasive and noninvasive data.

In summary, we thank Dr. Wassertheurer and colleagues forheir comments, which prompted us to better define the impor-ance of automated waveform calibration updating in radialonometry measurements. This, together with use of an N-pointoving average, seems to provide improved accuracy for the

oninvasive measurement of CASP in man.

Bryan Williams, MDeter Lacy, PhDeter Yan, MBBSgak Hwee Chuahen Liang, PhDhoon Meng Ting, MBBS

Department of Cardiovascular Sciences and Leicester NIHRBiomedical Research Unit in Cardiovascular Disease

niversity of Leicesterlinical Sciences Winglenfield Hospitalroby Roadeicester LE3 9QPnited Kingdom-mail: bw17@le.ac.uk

doi:10.1016/j.jacc.2011.07.014

EFERENCES

1. Williams B, Lacy PS, Yan P, Hwee CN, Liang C, Ting CM.Development and validation of a novel method to derive central aorticsystolic pressure from the radial pressure waveform using an N-pointmoving average method. J Am Coll Cardiol 2011;57:951–61.

2. Ng K-M, Ting C-M, Yeo J-H, et al. Progress on the development of

the MediWatch ambulatory blood pressure monitor and related devices.Blood Press Monit 2004;9:149–65.