- Email: [email protected]
Carotid Tonometry Versus Synthesized Aorta Pressure Waves for the Estimation of Central Systolic Blood Pressure and Augmentation Index
AJH
2005; 18:1168 –1173
Carotid Tonometry Versus Synthesized Aorta Pressure Waves for the Estimation of Central Systolic Blood Pressure and Augmentation Index Patrick Segers, Ernst Rietzschel, Steven Heireman, Marc De Buyzere, Thierry Gillebert, Pascal Verdonck, and Luc Van Bortel Objective: To assess the interchangeability of carotid tonometry and synthesized aorta pressure waveforms for estimating central systolic blood pressure (SBP) and augmentation index (AIx). Methods: Tonometry waveforms were acquired with a custom built hardware and software platform in 276 subjects (179 men/97 women; aged 45.5 ⫾ 5.7 years; mean ⫾ standard deviation) at the radial (Pwf,ra), brachial (Pwf,ba), and carotid artery (Pwf,ca). The Pwf,ba was calibrated using systolic (SBPba) and diastolic (DBPba) sphygmomanometer pressure. The DBPba and calculated mean (MAPba) brachial pressure were subsequently used for calibration of Pwf,ra and Pwf,ca. A central pressure waveform (Pwf,sao) was synthesized from Pwf,ra using a generalized pressure transfer function (TFF). The AIx and SBP were measured on Pwf,ra, Pwf,ca, and Pwf,sao. Results: The SBPra, SBPca, and SBPsao were 138.5 ⫾ 16.8, 130.0 ⫾ 16.2, and 131.1 ⫾ 16.6 mm Hg, respectively. The SBPra correlated well with the SBPca (r ⫽ 0.93) and the SBPsao (r ⫽ 0.94), as did the SBPca and the
SBPsao (r ⫽ 0.97) with a mean bias of 1.35 ⫾ 3.90 mm Hg. The AIx derived from Pwf,ra, Pwf,ca, and Pwf,sao were ⫺20.8% ⫾ 14.5%, 12.4% ⫾ 13.9%, and 20.0% ⫾ 11.7%, respectively. The correlation between radial and carotid, and radial and central AIx was 0.72 and 0.94, respectively. The correlation between AIx derived from Pwf,ca and Pwf,sao was 0.75 with a bias of 11.0% ⫾ 14% (all correlations P ⬍ .001). Conclusions: The use of a generalized TFF in combination with well-calibrated radial pressure curves yields estimates of SBP in good agreement with carotid tonometry. Although AIx derived from a measured radial pressure curve correlates surprisingly closely with AIx measured on a synthesized aortic pressure curve, the correlation with a directly measured AIx on carotid signals is relatively poor. Am J Hypertens 2005;18:1168 –1173 © 2005 American Journal of Hypertension, Ltd. Key Words: Applanation tonometry, arterial function, transfer function, wave reflection, Sphygmocor.
he analysis of aortic pressure wave morphology and concepts of arterial wave reflection have become common knowledge,1– 4 as well as the notion that early return of reflected waves, boosting systolic arterial pressure, is associated with arterial stiffening in hypertension and cardiovascular disease.2,5,6 In the past decade, the field of “pulse wave analysis”—in a broader context— has evolved into a subdomain of arterial function assessment. In the available arsenal of indices reflecting arterial
T
function, the augmentation index (AIx)—which quantifies the contribution of reflected waves on the arterial pressure wave— has taken a prominent place. The AIx was first defined in the late 1980s by Kelly et al,7 who simplified and formalized the wave contour classification scheme of Murgo et al.3 In the past few years, AIx has gained widespread application, mainly due to the development of noninvasive pressure waveform measuring tools (arterial applanation tonometry) and algorithms for the automated detection of landmark points on the pressure wave contour
Received January 10, 2005. First decision March 25, 2005. Accepted April 14, 2005. From the Cardiovascular Mechanics and Biofluid Dynamics, Hydraulics Laboratory (PS, PV), Department of Cardiovascular Diseases (ER, MDB, TG), and Department of Pharmacology (SH, LVB), Ghent University Hospital, Gent, Belgium. This research was funded by the Fund for Scientific Research
Flanders, Belgium (FWO-Vlaanderen) research grants G.0192.03 and G.0427.03 (the Asklepios Study). Address correspondence and reprint requests to Dr. Patrick Segers, Cardiovascular Mechanics and Biofluid Dynamics, Hydraulics Laboratory, Ghent University, Sint-Pietersnieuwstraat 41, B-9000 Gent, Belgium; e-mail: [email protected]
0895-7061/05/$30.00 doi:10.1016/j.amjhyper.2005.04.005
© 2005 by the American Journal of Hypertension, Ltd. Published by Elsevier Inc.
AJH–September 2005–VOL. 18, NO. 9
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
allowing calculation of the AIx. Although different formulations are used, AIx is nowadays most commonly defined as the ratio of the difference between the “secondary” and “primary peak,” and pulse pressure, expressed as a percentage (see Methods section for more precise definitions). The largest amount of literature data on AIx comes from the commercially available Sphygmocor device (AtCor Medical Pty, Ltd., West Ryde, NSW, Australia), where central aortic pressure is reconstructed, making use of radial artery tonometry and a generalized radial-to-aorta transfer function. This synthesized waveform is subsequently used for waveform analysis and calculation of AIx. In recent literature, however, the use of synthesized aortic pressure waveforms has been the subject of debate.8 –16 It has been claimed that AIx can be derived from a pressure contour assessed at the radial artery, as radial AIx highly correlates with central AIx, and thus, should contain the same diagnostic information.9 On the other hand, the characteristic marks associated with wave reflection are not always clearly visible on the radial artery wave contour. The present study had multiple aims. First, we wanted to develop dedicated hardware and software, independent of the Sphygmocor platform, for the acquisition and processing of applanation tonometry waveforms and validate algorithms for the automated detection of the primary and secondary peak on measured radial and carotid, as well as synthesized central aorta waveforms. Second, the relation between AIx at the different vascular territories was assessed. Third, the calibrated waveforms were analyzed in terms of systolic and pulse pressure to assess the relation between noninvasively measured carotid pressure and synthesized central pressure.
Methods Subjects A total of 276 subjects free from overt cardiovascular disease (179 men, aged 45.5 ⫾ 5.7 years (mean ⫾ standard deviation)) were included in the study. All subjects were examined in supine position in a quiet room (temperature 21° to 23°C). After a resting period of 10 min, brachial systolic and diastolic blood pressure (BP) was measured three times using a validated BP monitor (Omron HEM-907, Omron Matsuzaka Co. Ltd., Matsuzuka, Japan) on the right arm. The average of minimally three successive readings yielded systolic (SBPba) and diastolic (DBPba) BP. Measurement Set-Up Applanation tonometry was performed using a Millar pentype tonometer (SPT 301, Millar Instruments, Houston, TX). For data acquisition, we used a dedicated hardware and software platform (National Instruments SC-2345 signal acquisition and conditioning hardware, Austin, TX). To acquire and process the data, a graphic user interface
1169
was developed in Matlab (version 6.5, The Mathworks, Natick, MA). All tonometry data were recorded at 200 samples/sec. Data were recorded in continuous sequences of 20 sec. Postprocessing included signal filtering (Savitsky-Golay filter; Matlab, The Mathworks Inc.). identification of individual cycles, detrending (linearly smoothing out eventual differences in the numerical value of the start and end of the cycle), and averaging. Cycles with a cycle length shorter or longer than 20% of the mean cycle duration were automatically deselected, as well as cycles with a shape surpassing the “envelope” curves, which were constructed from the average ⫾ two times the standard deviation. This was done in an iterative way, until all cycles were within the envelope curves. As an arbitrary quality criterion, we accepted data only when minimally 10 cycles were retained. The average of these remaining cycles was considered as the tonometry recording for that measuring location. Measuring Protocol and Assessment of BP We used a systematic nomenclature whereby Pwf,xx denotes a complete pressure waveform at site xx (ba for brachial artery; ca for carotid artery, sao for synthesized aortic waveform). The maximum, minimum, and mean pressure at site xx are, respectively, denoted as SBPxx, DBPxx, and MAPxx. Local BP was assessed by calibration of pressure waveforms as proposed by Kelly and Fitchett17 and validated by Van Bortel et al.18 Tonometry was first performed at the level of the left brachial artery, and the tonometric recording was calibrated by designating the peak and trough of the curve the value of SBPba and DBPba, respectively, yielding a scaled brachial artery pressure tracing, Pwf,ba. Numerical averaging of Pwf,ba yielded mean arterial pressure, MAPba. Brachial tonometry occupied the upper part of the graphic user interface. Next, tonometry was performed at the left radial artery, with the measurements being visualized in the lower part of the interface. For the calibration of the averaged radial artery tonometric recording, we used the validated assumption that diastolic pressure and mean pressure remain fairly constant in the large arteries. As such, the trough and mean of all tonometric recordings were assigned the values of DBPba and MAPba giving scaled radial artery pressure (Pwf,ra) and carotid artery pressure (Pwf,ca). The maximum of the radial and carotid pressure curves was, respectively, radial systolic pressure, SBPra, and carotid systolic pressure, SBPca. The difference between systolic and diastolic pressure at radial, brachial, and carotid artery sites are, respectively, indicated as PPba, PPra, and PPca. We also synthesized central aortic pressure waveforms, Pwf,sao, from Pra, using a generalized radial-to-aorta pressure transfer function (TFF) and a previously applied frequency domain approach.19,20 The MAPsao was as-
1170
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
sumed equal to MAPba. The PPsao and SBPsao were derived. The Augmentation Index The AIx was defined as 100 (P2 ⫺ P1)/(Ps ⫺ Pd). For waves with a shoulder preceding the systolic pressure (so-called A-type waves), P1 is a pressure value characteristic for the shoulder, and P2 is the systolic pressure (Fig. 1). The AIx values are positive. For C-type waves, the shoulder follows the systolic pressure, and P1 is the systolic pressure, whereas P2 is a pressure value characteristic for the shoulder, thus yielding negative values for AIx (Fig. 1). The algorithm used for automated detection of the shoulder point was earlier described by Takazawa et al21 and is based on the fourth derivative of the pressure signal. For A-type waves, the peak coincides with the first positive-to-negative zero crossing of the fourth derivative, whereas for C-type waves, the second negative-to-positive is indicative for the pressure shoulder (see also Fig. 1, lower panels). We refer to the work of Takazawa et al21 for more details on the algorithm. We aimed to automatically detect the shoulder points for all 276 subjects, and AIx was subsequently calculated for Pwf,ca, Pwf,ra, and Pwf,sao, indicated as AIxca, AIxra, and AIxsao, respectively. The pressure tracings and subsequent derivatives were filtered to avoid the introduction of derivation-induced noise. Again, a Savitsky-Golay filter was used. As a validation of the algorithm, the shoulder pressure was also visually identified with the mouse cursor with the operator blinded to the result of the automated detection and the derivatives. The AIx, derived from visually and automatically assessed points, were compared. All algorithms were programmed in Matlab (version 6.5, The Mathworks).
AJH–September 2005–VOL. 18, NO. 9
Statistical Analysis Data are reported as mean ⫾ standard deviations. For the validation of the automated detection algorithm, linear regression analysis was used, as well as Bland-Altman analysis.22 Relations between AIx and SBP derived from radial and central calibrated waveforms were assessed using linear regression analysis. The relation between SBP and AIx derived from Pwf,ca and Pwf,sao was assessed using linear regression and Bland-Altman plots. The analysis was done in SPSS for Windows 11.5.1 (SPSS Inc., Chicago, IL).
Results BP Brachial SBP and DBP were 132.4 ⫾ 15.3 mm Hg and 80.9 ⫾ 10.3 mm Hg, respectively. Mean arterial BP (MAPba), calculated from averaging Pwf,ba, was 102.5 ⫾ 12.0 mm Hg, whereas MAPest, estimated as (2Pd ⫹ Ps)/3, was 98.1 ⫾ 11.1 mm Hg. Regression analysis showed that the relation between both can be described by MAPba ⫽ 1.045 MAPest, r2 ⫽ 0.96. The values of SBP at the different locations, and the correlation coefficients between them, are given in Table 1 and in Fig. 2. Augmentation Index The fully automated detection of the shoulder point was successful in 810 of 828 analyzed waveforms (98%). In the remaining cases, the algorithm situated the shoulder point in the diastolic phase, or was unable to detect any shoulder point. As an indication of the accuracy of the detection of the shoulder point, we calculated AIx from the manually and automatically detected point. The correlation coefficient between AIx assessed from manual and automatically detected shoulder point was 0.991, 0.996, and 0.986 for Pwf,ra, Pwf,ca, and Pwf,sao, respectively (all P
FIG. 1 Principle behind the automated detection of the shoulder point (bottom row) based on 4th derivative of the pressure signal (bottom row). Top panels show a carotid recording with an A-type (A) and C-type wave profile (B), while (C) displays a radial artery recording.
AJH–September 2005–VOL. 18, NO. 9
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
1171
Table 1. Blood pressure and derived parameters Brachial artery DBP (mm Hg) MAP (mm Hg) SBP (mm Hg) Alx (%)
80.9 ⫾ 10.3 — 132.4 ⫾ 15.3 —
Pwf,ra 80.9 102.5 138.5 ⫺20.8
⫾ ⫾ ⫾ ⫾
10.3 12.0 16.8 14.5
Pwf,ca 80.9 102.5 130.0 12.4
⫾ ⫾ ⫾ ⫾
10.3 12.0 16.2 13.9
Pwf,sao 81.1 102.5 131.3 20.0
⫾ ⫾ ⫾ ⫾
10.6 12.0 16.6 11.7
Correlation between SBP at different vascular sites
SBPba SBPra SBPca SBPsao
SBPba
SBPra
SBPca
SBPsao
— 0.948 0.943 0.925
0.948 — 0.926 0.940
0.943 0.926 — 0.972
0.925 0.940 0.972 —
Correlation between AIx at different vascular sites
AIxra AIxca AIxsao
AIxra
AIxca
AIxsao
— 0.717 0.939
0.717 — 0.745
0.939 0.745 —
Top rows: overview of systolic (SBP), diastolic (DBP), and mean blood pressure (MAP) and augmentation index (AIx) derived from measured pressure waveforms at the radial (Pwf,ra) and carotid artery (Pwf,ca), and from the synthesized aorta pressure (Pwf,sao). Brachial artery blood pressure was measured with the Omron blood pressure monitor. Middle rows: corelation coefficients between SBP at the different locations. Bottom rows: correlation between AIx assessed from radial (AIxra), carotid (AIxca), and reconstructed aorta pressure (AIxsao). All correlations P ⬍ .001.
⬍ .001), with mean differences of ⫺0.93% ⫾ 2.21%, 0.21% ⫾ 1.38%, and 0.38% ⫾ 2.01%. The AIx at radial, carotid, and synthesized aorta wave forms were ⫺20.8% ⫾ 14.5%, 12.4% ⫾ 13.9%, and 20.0% ⫾ 11.7%, respectively. The correlation coefficients between AIx at the different locations are listed in Table 1 and in Fig. 3.
Discussion
FIG. 2 Top panels: relation between radial systolic blood pressure (SBP) and SBP obtained from carotid tonometry (A) or from synthesized aorta pressure wave (B). Bottom panels: relation between SBP obtained from carotid tonometry and synthesized aorta waveform, with linear regression analysis on the left (C) and BlandAltman analysis on the right (D). See text for other abbreviations.
FIG. 3 Top panels: relation between augmentation index (AIx) derived from the radial waveform and AIx obtained from carotid tonometry (A) or from the synthesized aorta pressure wave (B). Bottom panels: relation between AIx obtained from carotid tonometry and synthesized aorta waveform, with linear regression analysis on the left (C) and Bland-Altman analysis on the right (D).
The quantification of the contribution of reflected waves in the arterial system, through the AIx, has become the subject of debate. Although there is consensus that AIx probably carries prognostic information and that it is ideally calculated from a waveform nearby the heart, the
1172
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
methodology to assess this central waveform provokes controversy. The most direct approach is to measure waveforms as close to the heart as possible, with the common carotid artery as an appealing location.23 Chen et al23 demonstrated, from simultaneous invasive measurement of aorta pressure and carotid applanation tonometry, that AIx derived from both measurements correlate reasonably well (r ⫽ 0.78), although they found that AIxca was generally lower than AIx derived from a measured aorta pressure (AIxmao). An alternative methodology consists of measuring pressure waveforms at the radial artery and to synthesize central pressure by using a transfer function as is done with the Sphygmocor device. Although the application of generalized transfer functions has been the subject of discussion,10,19,24 there are indications that, at least for the estimation of central systolic and pulse pressure, synthesized central pressures adequately reflect aorta pressure.20,25 The question remains whether these synthesized central pressures also allow for the assessment of central AIx. Chen et al,20 in their transfer function validation study, reported that AIxsao underestimates AIxmao by about 7% (absolute value). No correlation coefficients were reported. Hope et al13 compared AIxmao and AIxsao and found rather poor correlation coefficients (approximately 0.20), results highly contested by O’Rourke and co-workers, who attributed the poor agreement to an inadequate fluid-filled system used to measure ascending aorta pressure.14,26 In a more recent article where they discuss the calibration of tonometry waveforms, Hope et al27 measured aorta pressure with high fidelity transducers and reported a correlation coefficient between AIxmao and AIxsao of 0.65, significantly higher than in their previous study, which seems to support the hypothesis of O’Rourke et al. It was also found that AIxsao was about 7% (absolute AIx values) lower than AIxmao. Using data reported in an article by Söderstöm et al,28 it can be calculated that AIxsao (using Sphygmocor) underestimated AIxmao by 3.6% ⫾ 9.4% (absolute AIx values), where AIxmao was derived from pressure waveforms measured with highfidelity tip transducers. The correlation coefficient between AIxmao and AIxsao was 0.53 (P ⬍ .01). Segers et al29 also reported a relatively good correlation coefficient 0.64 and an underestimation of AIx when based on synthesized waveforms. From the available literature data, it thus appears that the correlation between AIxca and AIxmao is (slightly) better than the correlation between AIxsao and AIxmao, and from this perspective, it might be better to assess AIx from a carotid tonometry waveform rather than from a synthesized aorta pressure. The question has also been posed whether a transfer function adds any value as such. Millasseau et al,9 using the Sphygmocor, reported a strong correlation (r ⫽ 0.96) between AIxra and AIxsao. Also, and perhaps even more important, changes in AIx induced by nitroglycerine and norepinephrine, induced parallel changes in AIxra and
AJH–September 2005–VOL. 18, NO. 9
AIxsao. A similarly high correlation of r ⫽ 0.89 has recently been reported by Hope et al27 using the transfer function proposed by Fetics et al.30 In this study, we applied a similar transfer function to generate the synthesized waveforms, and we found a correlation coefficient between AIxra and AIxsao of 0.94. These correlation coefficients are much higher than the correlation that is found between AIxra and AIxmao derived from simultaneous measurements. Takazawa et al,21 for instance, reported a correlation coefficient of 0.74 between AIxra and AIxmao, similar to what we found in this study (comparing AIxra to AIxca). This suggests that the mathematical transformation of the radial waveform into a central waveform takes away some of the “in situ” variability attributed to specificities of the aorta–radial pathway and the variability related to the measuring inaccuracies in a second measurement, which is inevitably present when one compares indices derived from two separate (eventually also nonsimultaneous) measurements. Given the high correlation between indices derived from Pwf,ra and Pwf,sao, it is a valid question whether the wave transform effectively yields additional information. Supporters of synthesized central pressure use two arguments: 1) Pwf,sao yields more accurate estimates of central systolic and pulse pressure than peripheral BP. Table 1 shows that the transfer function corrects for the aorta-to-radial pressure amplification. Also, the correlation between SBPsao and SBPca (r ⫽ 0.97) is higher than between SBPba and SBPca (r ⫽ 0.94), suggesting that synthesized central pressure is also a better estimate of central pressure (⬃SBPca) than brachial artery BP. 2) The characteristic points, used to calculate AIx, are generally more pronounced on the central pressure waveform, and thus easier to detect than on the radial artery waveform, especially in young, healthy subjects.16 Both of these arguments in favor of Pwf,sao are, however, also valid for carotid artery pressure waveforms. A contraindication often mentioned is that carotid artery tonometry is harder to perform than radial artery tonometry. In this study, with the tonometer in the hands of trained operators, there was no dropout in subjects due to inability to perform carotid tonometry. Also, there were no particular problems to (automatically) detect the characteristic points on the radial artery waveform. The final accuracy of the pressure assessment highly depends on the appropriate calibration of these waveforms. As it is common practice, we also calibrated all tonometry tracings assuming constant DBP and MAP. However, we used a somewhat more complex protocol to assess MAP (see Methods section),17,18 rather than using the approximation with the third to two-third rule of thumb, or relying on MAP estimated with BP monitors. According to our data, the rule of thumb yields values of MAP that are about 5% lower than our protocol, thus with the underestimation becoming more important (in absolute numbers) for higher BPs. The impact on a calibrated waveform with “true” SBP, MAP, and DBP of 120, 95, and 80 mm Hg, respectively, would be a rescaled wave-
AJH–September 2005–VOL. 18, NO. 9
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
form with SBP, MAP, and DBP of 109.1, 90.9, and 80 mm Hg, respectively, thus a 9% reduction in SBP, illustrating the tremendous impact of using correct values for mean BP. Because the calibration is equally dependent on DBP, accurate measurement of this pressure parameter is of an equally paramount importance, as also addressed by Hope et al.27 Note, however, that calibration is only important for the assessment of BP and not for AIx, which is a dimensionless number solely determined by the shape of the waveform. The major limitation of our study is the absence of invasive central pressure recordings, which would have allowed to make more decisive conclusions. Also, we only measured data during baseline steady state conditions, without inducing physical or pharmacologic hemodynamic changes. Finally, we calculated Pwf,sao using the TFF as published by Chen et al20 without, however, transforming MAP (amplitude TFF at 0 Hz was assumed 1), and thus neglecting viscous pressure losses along the radial–aorta pathway. This may be the reason that SBPsao was, on average, slightly higher than SBPca, while it was anticipated that, with peripheral amplification, SBPsao would be slightly lower. Also, we did not assess the effect of using different transfer functions on SPBsao or AIxsao. In conclusion, we have shown that applying a generalized transfer function to well-calibrated radial pressure curves yields estimates of SBP in good agreement with carotid tonometry. The correlation between AIxsao and a directly measured AIxca, however, is relatively poor. We recommend the use of directly measured carotid tonometry waveforms for the assessment of central BP and, especially, AIx.
References 1. 2.
3.
4. 5. 6.
7.
8.
9.
Milnor WR: Hemodynamics, 2nd ed. Baltimore, Williams& Wilkins, 1989. Westerhof N, Sipkema P, van den Bos CG, Elzinga G: Forward and backward waves in the arterial system. Cardiovasc Res 1972;6:648 – 656. Murgo JP, Westerhof N, Giolma JP, Altobelli SA: Aortic input impedance in normal man: relationship to pressure wave forms. Circulation 1980;62:105–116. Nichols W, O’Rourke M: McDonald’s Blood Flow in Arteries, 4th ed. London, Edward Arnold, 1998. O’Rourke MF: Mechanical principles. Arterial stiffness and wave reflection. Pathol Biol (Paris) 1999;47:623– 633. Westerhof N, O’Rourke MF: Haemodynamic basis for the development of left ventricular failure in systolic hypertension and for its logical therapy. J Hypertens 1995;13:943–952. Kelly R, Hayward C, Avolio A, O’Rourke M: Noninvasive determination of age-related changes in the human arterial pulse. Circulation 1989;80:1652–1659. Lehmann ED: Where is the evidence that radial artery tonometry can be used to accurately and noninvasively predict central aortic blood pressure in patients with diabetes? Diabetes Care 2000;23: 869 – 871. Millasseau SC, Patel SJ, Redwood SR, Ritter JM, Chowienczyk PJ: Pressure wave reflection assessed from the peripheral pulse: is a transfer function necessary? Hypertension 2003;41:1016 –1020.
1173
10. Cameron JD, McGrath BP, Dart AM: Use of radial artery applanation tonometry and a generalized transfer function to determine aortic pressure augmentation in subjects with treated hypertension. J Am Coll Cardiol 1998;32:1214 –1220. 11. Hoeks AP, Meinders JM, Dammers R: Applicability and benefit of arterial transfer functions. J Hypertens 2003;21:1241–1243. 12. Hope SA, Meredith IT, Cameron JD: Reliability of transfer functions in determining central pulse pressure and augmentation index. J Am Coll Cardiol 2002;40:1196 –7. 13. Hope SA, Tay DB, Meredith IT, Cameron JD: Use of arterial transfer functions for the derivation of aortic waveform characteristics. J Hypertens 2003;21:1299 –1305. 14. O’Rourke MF, Nichols WW: Use of arterial transfer function for the derivation of aortic waveform characteristics. J Hypertens 2003;21: 2195–9. 15. O’Rourke MF, Avolio A, Qasem A: Clinical assessment of wave reflection. Hypertension 2003;42:e15– 6. 16. Avolio AP, O’Rourke MF: Applicability and benefit of arterial transfer functions. J Hypertens 2003;21:2199 –2201. 17. Kelly R, Fitchett D: Noninvasive determination of aortic input impedance and external left ventricular power output: a validation and repeatability study of a new technique. J Am Coll Cardiol 1992;20:952–963. 18. Van Bortel LM, Balkestein EJ, van der Heijden-Spek JJ, Vanmolkot FH, Staessen JA, Kragten JA, Vredeveld JW, Safar ME, Struijker Boudier HA, Hoeks AP: Non-invasive assessment of local arterial pulse pressure: comparison of applanation tonometry and echotracking. J Hypertens 2001;19:1037–1044. 19. Segers P, Carlier S, Pasquet A, Rabben SI, Hellevik LR, Remme E, De Backer T, De Sutter J, Thomas JD, Verdonck P: Individualizing the aorto-radial pressure transfer function: feasibility of a model-based approach. Am J Physiol Heart Circ Physiol 2000;279:H542–H549. 20. Chen CH, Nevo E, Fetics B, Pak PH, Yin FC, Maughan WL, Kass DA: Estimation of central aortic pressure waveform by mathematical transformation of radial tonometry pressure. Validation of generalized transfer function. Circulation 1997;95:1827–1836. 21. Takazawa K, Tanaka N, Takeda K, Kurosu F, Ibukiyama C: Underestimation of vasodilator effects of nitroglycerin by upper limb blood pressure. Hypertension 1995;26:520 –523. 22. Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307– 310. 23. Chen C-H, Ting C-T, Nussbacher A, Nevo E, Kass DA, Pak P, Wang S-P, Chang M-S, Yin FCP: Validation of carotid artery tonometry as a means of estimating augmentation index of ascending aortic pressure. Hypertension 1996;27:168 –175. 24. Hope SA, Tay DB, Meredith IT, Cameron JD: Comparison of generalized and gender-specific transfer functions for the derivation of aortic waveforms. Am J Physiol Heart Circ Physiol 2002;283: H1150 –H1156. 25. Pauca AL, O’Rourke MF, Kon ND: Prospective evaluation of a method for estimating ascending aortic pressure from the radial artery pressure waveform. Hypertension 2001;38:932–937. 26. O’Rourke MF, Kim M, Adji A, Nichols WW, Avolio A: Use of arterial transfer function for the derivation of aortic waveform characteristics. J Hypertens 2004;22:431– 4. 27. Hope SA, Meredith IT, Cameron JD: Effect of non-invasive calibration of radial waveforms on error in transfer-function-derived central aortic waveform characteristics. Clin Sci (Lond) 2004;107:205–211. 28. Söderstöm S, Nyberg G, O’Rourke MF, Sellgren J, Ponten J: Can a clinically useful aortic pressure wave be derived from a radial pressure wave? Br J Anaesth 2002;88:481– 488. 29. Segers P, Qasem A, De Backer T, Carlier S, Verdonck P, Avolio A: Peripheral “oscillatory” compliance is associated with aortic augmentation index. Hypertension 2001;37:1434 –1439. 30. Fetics B, Nevo E, Chen CH, Kass DA: Parametric model derivation of transfer function for noninvasive estimation of aortic pressure by radial tonometry. IEEE Trans Biomed Eng 1999;46:698 –706.
2005; 18:1168 –1173
Carotid Tonometry Versus Synthesized Aorta Pressure Waves for the Estimation of Central Systolic Blood Pressure and Augmentation Index Patrick Segers, Ernst Rietzschel, Steven Heireman, Marc De Buyzere, Thierry Gillebert, Pascal Verdonck, and Luc Van Bortel Objective: To assess the interchangeability of carotid tonometry and synthesized aorta pressure waveforms for estimating central systolic blood pressure (SBP) and augmentation index (AIx). Methods: Tonometry waveforms were acquired with a custom built hardware and software platform in 276 subjects (179 men/97 women; aged 45.5 ⫾ 5.7 years; mean ⫾ standard deviation) at the radial (Pwf,ra), brachial (Pwf,ba), and carotid artery (Pwf,ca). The Pwf,ba was calibrated using systolic (SBPba) and diastolic (DBPba) sphygmomanometer pressure. The DBPba and calculated mean (MAPba) brachial pressure were subsequently used for calibration of Pwf,ra and Pwf,ca. A central pressure waveform (Pwf,sao) was synthesized from Pwf,ra using a generalized pressure transfer function (TFF). The AIx and SBP were measured on Pwf,ra, Pwf,ca, and Pwf,sao. Results: The SBPra, SBPca, and SBPsao were 138.5 ⫾ 16.8, 130.0 ⫾ 16.2, and 131.1 ⫾ 16.6 mm Hg, respectively. The SBPra correlated well with the SBPca (r ⫽ 0.93) and the SBPsao (r ⫽ 0.94), as did the SBPca and the
SBPsao (r ⫽ 0.97) with a mean bias of 1.35 ⫾ 3.90 mm Hg. The AIx derived from Pwf,ra, Pwf,ca, and Pwf,sao were ⫺20.8% ⫾ 14.5%, 12.4% ⫾ 13.9%, and 20.0% ⫾ 11.7%, respectively. The correlation between radial and carotid, and radial and central AIx was 0.72 and 0.94, respectively. The correlation between AIx derived from Pwf,ca and Pwf,sao was 0.75 with a bias of 11.0% ⫾ 14% (all correlations P ⬍ .001). Conclusions: The use of a generalized TFF in combination with well-calibrated radial pressure curves yields estimates of SBP in good agreement with carotid tonometry. Although AIx derived from a measured radial pressure curve correlates surprisingly closely with AIx measured on a synthesized aortic pressure curve, the correlation with a directly measured AIx on carotid signals is relatively poor. Am J Hypertens 2005;18:1168 –1173 © 2005 American Journal of Hypertension, Ltd. Key Words: Applanation tonometry, arterial function, transfer function, wave reflection, Sphygmocor.
he analysis of aortic pressure wave morphology and concepts of arterial wave reflection have become common knowledge,1– 4 as well as the notion that early return of reflected waves, boosting systolic arterial pressure, is associated with arterial stiffening in hypertension and cardiovascular disease.2,5,6 In the past decade, the field of “pulse wave analysis”—in a broader context— has evolved into a subdomain of arterial function assessment. In the available arsenal of indices reflecting arterial
T
function, the augmentation index (AIx)—which quantifies the contribution of reflected waves on the arterial pressure wave— has taken a prominent place. The AIx was first defined in the late 1980s by Kelly et al,7 who simplified and formalized the wave contour classification scheme of Murgo et al.3 In the past few years, AIx has gained widespread application, mainly due to the development of noninvasive pressure waveform measuring tools (arterial applanation tonometry) and algorithms for the automated detection of landmark points on the pressure wave contour
Received January 10, 2005. First decision March 25, 2005. Accepted April 14, 2005. From the Cardiovascular Mechanics and Biofluid Dynamics, Hydraulics Laboratory (PS, PV), Department of Cardiovascular Diseases (ER, MDB, TG), and Department of Pharmacology (SH, LVB), Ghent University Hospital, Gent, Belgium. This research was funded by the Fund for Scientific Research
Flanders, Belgium (FWO-Vlaanderen) research grants G.0192.03 and G.0427.03 (the Asklepios Study). Address correspondence and reprint requests to Dr. Patrick Segers, Cardiovascular Mechanics and Biofluid Dynamics, Hydraulics Laboratory, Ghent University, Sint-Pietersnieuwstraat 41, B-9000 Gent, Belgium; e-mail: [email protected]
0895-7061/05/$30.00 doi:10.1016/j.amjhyper.2005.04.005
© 2005 by the American Journal of Hypertension, Ltd. Published by Elsevier Inc.
AJH–September 2005–VOL. 18, NO. 9
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
allowing calculation of the AIx. Although different formulations are used, AIx is nowadays most commonly defined as the ratio of the difference between the “secondary” and “primary peak,” and pulse pressure, expressed as a percentage (see Methods section for more precise definitions). The largest amount of literature data on AIx comes from the commercially available Sphygmocor device (AtCor Medical Pty, Ltd., West Ryde, NSW, Australia), where central aortic pressure is reconstructed, making use of radial artery tonometry and a generalized radial-to-aorta transfer function. This synthesized waveform is subsequently used for waveform analysis and calculation of AIx. In recent literature, however, the use of synthesized aortic pressure waveforms has been the subject of debate.8 –16 It has been claimed that AIx can be derived from a pressure contour assessed at the radial artery, as radial AIx highly correlates with central AIx, and thus, should contain the same diagnostic information.9 On the other hand, the characteristic marks associated with wave reflection are not always clearly visible on the radial artery wave contour. The present study had multiple aims. First, we wanted to develop dedicated hardware and software, independent of the Sphygmocor platform, for the acquisition and processing of applanation tonometry waveforms and validate algorithms for the automated detection of the primary and secondary peak on measured radial and carotid, as well as synthesized central aorta waveforms. Second, the relation between AIx at the different vascular territories was assessed. Third, the calibrated waveforms were analyzed in terms of systolic and pulse pressure to assess the relation between noninvasively measured carotid pressure and synthesized central pressure.
Methods Subjects A total of 276 subjects free from overt cardiovascular disease (179 men, aged 45.5 ⫾ 5.7 years (mean ⫾ standard deviation)) were included in the study. All subjects were examined in supine position in a quiet room (temperature 21° to 23°C). After a resting period of 10 min, brachial systolic and diastolic blood pressure (BP) was measured three times using a validated BP monitor (Omron HEM-907, Omron Matsuzaka Co. Ltd., Matsuzuka, Japan) on the right arm. The average of minimally three successive readings yielded systolic (SBPba) and diastolic (DBPba) BP. Measurement Set-Up Applanation tonometry was performed using a Millar pentype tonometer (SPT 301, Millar Instruments, Houston, TX). For data acquisition, we used a dedicated hardware and software platform (National Instruments SC-2345 signal acquisition and conditioning hardware, Austin, TX). To acquire and process the data, a graphic user interface
1169
was developed in Matlab (version 6.5, The Mathworks, Natick, MA). All tonometry data were recorded at 200 samples/sec. Data were recorded in continuous sequences of 20 sec. Postprocessing included signal filtering (Savitsky-Golay filter; Matlab, The Mathworks Inc.). identification of individual cycles, detrending (linearly smoothing out eventual differences in the numerical value of the start and end of the cycle), and averaging. Cycles with a cycle length shorter or longer than 20% of the mean cycle duration were automatically deselected, as well as cycles with a shape surpassing the “envelope” curves, which were constructed from the average ⫾ two times the standard deviation. This was done in an iterative way, until all cycles were within the envelope curves. As an arbitrary quality criterion, we accepted data only when minimally 10 cycles were retained. The average of these remaining cycles was considered as the tonometry recording for that measuring location. Measuring Protocol and Assessment of BP We used a systematic nomenclature whereby Pwf,xx denotes a complete pressure waveform at site xx (ba for brachial artery; ca for carotid artery, sao for synthesized aortic waveform). The maximum, minimum, and mean pressure at site xx are, respectively, denoted as SBPxx, DBPxx, and MAPxx. Local BP was assessed by calibration of pressure waveforms as proposed by Kelly and Fitchett17 and validated by Van Bortel et al.18 Tonometry was first performed at the level of the left brachial artery, and the tonometric recording was calibrated by designating the peak and trough of the curve the value of SBPba and DBPba, respectively, yielding a scaled brachial artery pressure tracing, Pwf,ba. Numerical averaging of Pwf,ba yielded mean arterial pressure, MAPba. Brachial tonometry occupied the upper part of the graphic user interface. Next, tonometry was performed at the left radial artery, with the measurements being visualized in the lower part of the interface. For the calibration of the averaged radial artery tonometric recording, we used the validated assumption that diastolic pressure and mean pressure remain fairly constant in the large arteries. As such, the trough and mean of all tonometric recordings were assigned the values of DBPba and MAPba giving scaled radial artery pressure (Pwf,ra) and carotid artery pressure (Pwf,ca). The maximum of the radial and carotid pressure curves was, respectively, radial systolic pressure, SBPra, and carotid systolic pressure, SBPca. The difference between systolic and diastolic pressure at radial, brachial, and carotid artery sites are, respectively, indicated as PPba, PPra, and PPca. We also synthesized central aortic pressure waveforms, Pwf,sao, from Pra, using a generalized radial-to-aorta pressure transfer function (TFF) and a previously applied frequency domain approach.19,20 The MAPsao was as-
1170
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
sumed equal to MAPba. The PPsao and SBPsao were derived. The Augmentation Index The AIx was defined as 100 (P2 ⫺ P1)/(Ps ⫺ Pd). For waves with a shoulder preceding the systolic pressure (so-called A-type waves), P1 is a pressure value characteristic for the shoulder, and P2 is the systolic pressure (Fig. 1). The AIx values are positive. For C-type waves, the shoulder follows the systolic pressure, and P1 is the systolic pressure, whereas P2 is a pressure value characteristic for the shoulder, thus yielding negative values for AIx (Fig. 1). The algorithm used for automated detection of the shoulder point was earlier described by Takazawa et al21 and is based on the fourth derivative of the pressure signal. For A-type waves, the peak coincides with the first positive-to-negative zero crossing of the fourth derivative, whereas for C-type waves, the second negative-to-positive is indicative for the pressure shoulder (see also Fig. 1, lower panels). We refer to the work of Takazawa et al21 for more details on the algorithm. We aimed to automatically detect the shoulder points for all 276 subjects, and AIx was subsequently calculated for Pwf,ca, Pwf,ra, and Pwf,sao, indicated as AIxca, AIxra, and AIxsao, respectively. The pressure tracings and subsequent derivatives were filtered to avoid the introduction of derivation-induced noise. Again, a Savitsky-Golay filter was used. As a validation of the algorithm, the shoulder pressure was also visually identified with the mouse cursor with the operator blinded to the result of the automated detection and the derivatives. The AIx, derived from visually and automatically assessed points, were compared. All algorithms were programmed in Matlab (version 6.5, The Mathworks).
AJH–September 2005–VOL. 18, NO. 9
Statistical Analysis Data are reported as mean ⫾ standard deviations. For the validation of the automated detection algorithm, linear regression analysis was used, as well as Bland-Altman analysis.22 Relations between AIx and SBP derived from radial and central calibrated waveforms were assessed using linear regression analysis. The relation between SBP and AIx derived from Pwf,ca and Pwf,sao was assessed using linear regression and Bland-Altman plots. The analysis was done in SPSS for Windows 11.5.1 (SPSS Inc., Chicago, IL).
Results BP Brachial SBP and DBP were 132.4 ⫾ 15.3 mm Hg and 80.9 ⫾ 10.3 mm Hg, respectively. Mean arterial BP (MAPba), calculated from averaging Pwf,ba, was 102.5 ⫾ 12.0 mm Hg, whereas MAPest, estimated as (2Pd ⫹ Ps)/3, was 98.1 ⫾ 11.1 mm Hg. Regression analysis showed that the relation between both can be described by MAPba ⫽ 1.045 MAPest, r2 ⫽ 0.96. The values of SBP at the different locations, and the correlation coefficients between them, are given in Table 1 and in Fig. 2. Augmentation Index The fully automated detection of the shoulder point was successful in 810 of 828 analyzed waveforms (98%). In the remaining cases, the algorithm situated the shoulder point in the diastolic phase, or was unable to detect any shoulder point. As an indication of the accuracy of the detection of the shoulder point, we calculated AIx from the manually and automatically detected point. The correlation coefficient between AIx assessed from manual and automatically detected shoulder point was 0.991, 0.996, and 0.986 for Pwf,ra, Pwf,ca, and Pwf,sao, respectively (all P
FIG. 1 Principle behind the automated detection of the shoulder point (bottom row) based on 4th derivative of the pressure signal (bottom row). Top panels show a carotid recording with an A-type (A) and C-type wave profile (B), while (C) displays a radial artery recording.
AJH–September 2005–VOL. 18, NO. 9
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
1171
Table 1. Blood pressure and derived parameters Brachial artery DBP (mm Hg) MAP (mm Hg) SBP (mm Hg) Alx (%)
80.9 ⫾ 10.3 — 132.4 ⫾ 15.3 —
Pwf,ra 80.9 102.5 138.5 ⫺20.8
⫾ ⫾ ⫾ ⫾
10.3 12.0 16.8 14.5
Pwf,ca 80.9 102.5 130.0 12.4
⫾ ⫾ ⫾ ⫾
10.3 12.0 16.2 13.9
Pwf,sao 81.1 102.5 131.3 20.0
⫾ ⫾ ⫾ ⫾
10.6 12.0 16.6 11.7
Correlation between SBP at different vascular sites
SBPba SBPra SBPca SBPsao
SBPba
SBPra
SBPca
SBPsao
— 0.948 0.943 0.925
0.948 — 0.926 0.940
0.943 0.926 — 0.972
0.925 0.940 0.972 —
Correlation between AIx at different vascular sites
AIxra AIxca AIxsao
AIxra
AIxca
AIxsao
— 0.717 0.939
0.717 — 0.745
0.939 0.745 —
Top rows: overview of systolic (SBP), diastolic (DBP), and mean blood pressure (MAP) and augmentation index (AIx) derived from measured pressure waveforms at the radial (Pwf,ra) and carotid artery (Pwf,ca), and from the synthesized aorta pressure (Pwf,sao). Brachial artery blood pressure was measured with the Omron blood pressure monitor. Middle rows: corelation coefficients between SBP at the different locations. Bottom rows: correlation between AIx assessed from radial (AIxra), carotid (AIxca), and reconstructed aorta pressure (AIxsao). All correlations P ⬍ .001.
⬍ .001), with mean differences of ⫺0.93% ⫾ 2.21%, 0.21% ⫾ 1.38%, and 0.38% ⫾ 2.01%. The AIx at radial, carotid, and synthesized aorta wave forms were ⫺20.8% ⫾ 14.5%, 12.4% ⫾ 13.9%, and 20.0% ⫾ 11.7%, respectively. The correlation coefficients between AIx at the different locations are listed in Table 1 and in Fig. 3.
Discussion
FIG. 2 Top panels: relation between radial systolic blood pressure (SBP) and SBP obtained from carotid tonometry (A) or from synthesized aorta pressure wave (B). Bottom panels: relation between SBP obtained from carotid tonometry and synthesized aorta waveform, with linear regression analysis on the left (C) and BlandAltman analysis on the right (D). See text for other abbreviations.
FIG. 3 Top panels: relation between augmentation index (AIx) derived from the radial waveform and AIx obtained from carotid tonometry (A) or from the synthesized aorta pressure wave (B). Bottom panels: relation between AIx obtained from carotid tonometry and synthesized aorta waveform, with linear regression analysis on the left (C) and Bland-Altman analysis on the right (D).
The quantification of the contribution of reflected waves in the arterial system, through the AIx, has become the subject of debate. Although there is consensus that AIx probably carries prognostic information and that it is ideally calculated from a waveform nearby the heart, the
1172
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
methodology to assess this central waveform provokes controversy. The most direct approach is to measure waveforms as close to the heart as possible, with the common carotid artery as an appealing location.23 Chen et al23 demonstrated, from simultaneous invasive measurement of aorta pressure and carotid applanation tonometry, that AIx derived from both measurements correlate reasonably well (r ⫽ 0.78), although they found that AIxca was generally lower than AIx derived from a measured aorta pressure (AIxmao). An alternative methodology consists of measuring pressure waveforms at the radial artery and to synthesize central pressure by using a transfer function as is done with the Sphygmocor device. Although the application of generalized transfer functions has been the subject of discussion,10,19,24 there are indications that, at least for the estimation of central systolic and pulse pressure, synthesized central pressures adequately reflect aorta pressure.20,25 The question remains whether these synthesized central pressures also allow for the assessment of central AIx. Chen et al,20 in their transfer function validation study, reported that AIxsao underestimates AIxmao by about 7% (absolute value). No correlation coefficients were reported. Hope et al13 compared AIxmao and AIxsao and found rather poor correlation coefficients (approximately 0.20), results highly contested by O’Rourke and co-workers, who attributed the poor agreement to an inadequate fluid-filled system used to measure ascending aorta pressure.14,26 In a more recent article where they discuss the calibration of tonometry waveforms, Hope et al27 measured aorta pressure with high fidelity transducers and reported a correlation coefficient between AIxmao and AIxsao of 0.65, significantly higher than in their previous study, which seems to support the hypothesis of O’Rourke et al. It was also found that AIxsao was about 7% (absolute AIx values) lower than AIxmao. Using data reported in an article by Söderstöm et al,28 it can be calculated that AIxsao (using Sphygmocor) underestimated AIxmao by 3.6% ⫾ 9.4% (absolute AIx values), where AIxmao was derived from pressure waveforms measured with highfidelity tip transducers. The correlation coefficient between AIxmao and AIxsao was 0.53 (P ⬍ .01). Segers et al29 also reported a relatively good correlation coefficient 0.64 and an underestimation of AIx when based on synthesized waveforms. From the available literature data, it thus appears that the correlation between AIxca and AIxmao is (slightly) better than the correlation between AIxsao and AIxmao, and from this perspective, it might be better to assess AIx from a carotid tonometry waveform rather than from a synthesized aorta pressure. The question has also been posed whether a transfer function adds any value as such. Millasseau et al,9 using the Sphygmocor, reported a strong correlation (r ⫽ 0.96) between AIxra and AIxsao. Also, and perhaps even more important, changes in AIx induced by nitroglycerine and norepinephrine, induced parallel changes in AIxra and
AJH–September 2005–VOL. 18, NO. 9
AIxsao. A similarly high correlation of r ⫽ 0.89 has recently been reported by Hope et al27 using the transfer function proposed by Fetics et al.30 In this study, we applied a similar transfer function to generate the synthesized waveforms, and we found a correlation coefficient between AIxra and AIxsao of 0.94. These correlation coefficients are much higher than the correlation that is found between AIxra and AIxmao derived from simultaneous measurements. Takazawa et al,21 for instance, reported a correlation coefficient of 0.74 between AIxra and AIxmao, similar to what we found in this study (comparing AIxra to AIxca). This suggests that the mathematical transformation of the radial waveform into a central waveform takes away some of the “in situ” variability attributed to specificities of the aorta–radial pathway and the variability related to the measuring inaccuracies in a second measurement, which is inevitably present when one compares indices derived from two separate (eventually also nonsimultaneous) measurements. Given the high correlation between indices derived from Pwf,ra and Pwf,sao, it is a valid question whether the wave transform effectively yields additional information. Supporters of synthesized central pressure use two arguments: 1) Pwf,sao yields more accurate estimates of central systolic and pulse pressure than peripheral BP. Table 1 shows that the transfer function corrects for the aorta-to-radial pressure amplification. Also, the correlation between SBPsao and SBPca (r ⫽ 0.97) is higher than between SBPba and SBPca (r ⫽ 0.94), suggesting that synthesized central pressure is also a better estimate of central pressure (⬃SBPca) than brachial artery BP. 2) The characteristic points, used to calculate AIx, are generally more pronounced on the central pressure waveform, and thus easier to detect than on the radial artery waveform, especially in young, healthy subjects.16 Both of these arguments in favor of Pwf,sao are, however, also valid for carotid artery pressure waveforms. A contraindication often mentioned is that carotid artery tonometry is harder to perform than radial artery tonometry. In this study, with the tonometer in the hands of trained operators, there was no dropout in subjects due to inability to perform carotid tonometry. Also, there were no particular problems to (automatically) detect the characteristic points on the radial artery waveform. The final accuracy of the pressure assessment highly depends on the appropriate calibration of these waveforms. As it is common practice, we also calibrated all tonometry tracings assuming constant DBP and MAP. However, we used a somewhat more complex protocol to assess MAP (see Methods section),17,18 rather than using the approximation with the third to two-third rule of thumb, or relying on MAP estimated with BP monitors. According to our data, the rule of thumb yields values of MAP that are about 5% lower than our protocol, thus with the underestimation becoming more important (in absolute numbers) for higher BPs. The impact on a calibrated waveform with “true” SBP, MAP, and DBP of 120, 95, and 80 mm Hg, respectively, would be a rescaled wave-
AJH–September 2005–VOL. 18, NO. 9
CAROTID TONOMETRY VERSUS SYNTHESIZED WAVEFORMS
form with SBP, MAP, and DBP of 109.1, 90.9, and 80 mm Hg, respectively, thus a 9% reduction in SBP, illustrating the tremendous impact of using correct values for mean BP. Because the calibration is equally dependent on DBP, accurate measurement of this pressure parameter is of an equally paramount importance, as also addressed by Hope et al.27 Note, however, that calibration is only important for the assessment of BP and not for AIx, which is a dimensionless number solely determined by the shape of the waveform. The major limitation of our study is the absence of invasive central pressure recordings, which would have allowed to make more decisive conclusions. Also, we only measured data during baseline steady state conditions, without inducing physical or pharmacologic hemodynamic changes. Finally, we calculated Pwf,sao using the TFF as published by Chen et al20 without, however, transforming MAP (amplitude TFF at 0 Hz was assumed 1), and thus neglecting viscous pressure losses along the radial–aorta pathway. This may be the reason that SBPsao was, on average, slightly higher than SBPca, while it was anticipated that, with peripheral amplification, SBPsao would be slightly lower. Also, we did not assess the effect of using different transfer functions on SPBsao or AIxsao. In conclusion, we have shown that applying a generalized transfer function to well-calibrated radial pressure curves yields estimates of SBP in good agreement with carotid tonometry. The correlation between AIxsao and a directly measured AIxca, however, is relatively poor. We recommend the use of directly measured carotid tonometry waveforms for the assessment of central BP and, especially, AIx.
References 1. 2.
3.
4. 5. 6.
7.
8.
9.
Milnor WR: Hemodynamics, 2nd ed. Baltimore, Williams& Wilkins, 1989. Westerhof N, Sipkema P, van den Bos CG, Elzinga G: Forward and backward waves in the arterial system. Cardiovasc Res 1972;6:648 – 656. Murgo JP, Westerhof N, Giolma JP, Altobelli SA: Aortic input impedance in normal man: relationship to pressure wave forms. Circulation 1980;62:105–116. Nichols W, O’Rourke M: McDonald’s Blood Flow in Arteries, 4th ed. London, Edward Arnold, 1998. O’Rourke MF: Mechanical principles. Arterial stiffness and wave reflection. Pathol Biol (Paris) 1999;47:623– 633. Westerhof N, O’Rourke MF: Haemodynamic basis for the development of left ventricular failure in systolic hypertension and for its logical therapy. J Hypertens 1995;13:943–952. Kelly R, Hayward C, Avolio A, O’Rourke M: Noninvasive determination of age-related changes in the human arterial pulse. Circulation 1989;80:1652–1659. Lehmann ED: Where is the evidence that radial artery tonometry can be used to accurately and noninvasively predict central aortic blood pressure in patients with diabetes? Diabetes Care 2000;23: 869 – 871. Millasseau SC, Patel SJ, Redwood SR, Ritter JM, Chowienczyk PJ: Pressure wave reflection assessed from the peripheral pulse: is a transfer function necessary? Hypertension 2003;41:1016 –1020.
1173
10. Cameron JD, McGrath BP, Dart AM: Use of radial artery applanation tonometry and a generalized transfer function to determine aortic pressure augmentation in subjects with treated hypertension. J Am Coll Cardiol 1998;32:1214 –1220. 11. Hoeks AP, Meinders JM, Dammers R: Applicability and benefit of arterial transfer functions. J Hypertens 2003;21:1241–1243. 12. Hope SA, Meredith IT, Cameron JD: Reliability of transfer functions in determining central pulse pressure and augmentation index. J Am Coll Cardiol 2002;40:1196 –7. 13. Hope SA, Tay DB, Meredith IT, Cameron JD: Use of arterial transfer functions for the derivation of aortic waveform characteristics. J Hypertens 2003;21:1299 –1305. 14. O’Rourke MF, Nichols WW: Use of arterial transfer function for the derivation of aortic waveform characteristics. J Hypertens 2003;21: 2195–9. 15. O’Rourke MF, Avolio A, Qasem A: Clinical assessment of wave reflection. Hypertension 2003;42:e15– 6. 16. Avolio AP, O’Rourke MF: Applicability and benefit of arterial transfer functions. J Hypertens 2003;21:2199 –2201. 17. Kelly R, Fitchett D: Noninvasive determination of aortic input impedance and external left ventricular power output: a validation and repeatability study of a new technique. J Am Coll Cardiol 1992;20:952–963. 18. Van Bortel LM, Balkestein EJ, van der Heijden-Spek JJ, Vanmolkot FH, Staessen JA, Kragten JA, Vredeveld JW, Safar ME, Struijker Boudier HA, Hoeks AP: Non-invasive assessment of local arterial pulse pressure: comparison of applanation tonometry and echotracking. J Hypertens 2001;19:1037–1044. 19. Segers P, Carlier S, Pasquet A, Rabben SI, Hellevik LR, Remme E, De Backer T, De Sutter J, Thomas JD, Verdonck P: Individualizing the aorto-radial pressure transfer function: feasibility of a model-based approach. Am J Physiol Heart Circ Physiol 2000;279:H542–H549. 20. Chen CH, Nevo E, Fetics B, Pak PH, Yin FC, Maughan WL, Kass DA: Estimation of central aortic pressure waveform by mathematical transformation of radial tonometry pressure. Validation of generalized transfer function. Circulation 1997;95:1827–1836. 21. Takazawa K, Tanaka N, Takeda K, Kurosu F, Ibukiyama C: Underestimation of vasodilator effects of nitroglycerin by upper limb blood pressure. Hypertension 1995;26:520 –523. 22. Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307– 310. 23. Chen C-H, Ting C-T, Nussbacher A, Nevo E, Kass DA, Pak P, Wang S-P, Chang M-S, Yin FCP: Validation of carotid artery tonometry as a means of estimating augmentation index of ascending aortic pressure. Hypertension 1996;27:168 –175. 24. Hope SA, Tay DB, Meredith IT, Cameron JD: Comparison of generalized and gender-specific transfer functions for the derivation of aortic waveforms. Am J Physiol Heart Circ Physiol 2002;283: H1150 –H1156. 25. Pauca AL, O’Rourke MF, Kon ND: Prospective evaluation of a method for estimating ascending aortic pressure from the radial artery pressure waveform. Hypertension 2001;38:932–937. 26. O’Rourke MF, Kim M, Adji A, Nichols WW, Avolio A: Use of arterial transfer function for the derivation of aortic waveform characteristics. J Hypertens 2004;22:431– 4. 27. Hope SA, Meredith IT, Cameron JD: Effect of non-invasive calibration of radial waveforms on error in transfer-function-derived central aortic waveform characteristics. Clin Sci (Lond) 2004;107:205–211. 28. Söderstöm S, Nyberg G, O’Rourke MF, Sellgren J, Ponten J: Can a clinically useful aortic pressure wave be derived from a radial pressure wave? Br J Anaesth 2002;88:481– 488. 29. Segers P, Qasem A, De Backer T, Carlier S, Verdonck P, Avolio A: Peripheral “oscillatory” compliance is associated with aortic augmentation index. Hypertension 2001;37:1434 –1439. 30. Fetics B, Nevo E, Chen CH, Kass DA: Parametric model derivation of transfer function for noninvasive estimation of aortic pressure by radial tonometry. IEEE Trans Biomed Eng 1999;46:698 –706.