Estimation of Central Aortic Pressure Waveform by Mathematical Transformation of Radial Tonometry Pressure Data
To the Editor:
Chen et al1 report on the estimation of central aortic pressure waveform data using a generalized transfer function (TF). Their conclusion that “central aortic pressures can be accurately estimated from radial tonometry with the use of a generalized TF” is not supported by their data.
Intra-aortic and radial artery pressure data, collected from a cohort of 20 patients undergoing cardiac catheterization, were used to generate the TF. To use these same data to test out the TF is only appropriate for checking the internal validity of the mathematical transformations being applied. To be able to draw any useful conclusions about the overall applicability of such a generalized TF and its external validity requires further data to be prospectively collected from a separate group of patients (ie, different from those used to generate the original TF).
The application of generalized TFs using both applanation tonometric data from the radial artery2 and Finapres data from the digital artery3 suggests that when such prospective analyses are carried out, quite marked differences between estimated and measured central aortic pressures can be observed (B.T. Plunkett et al, unpublished data, 1996). Furthermore, with increasing blood pressure these differences become even more significant, with implications for the validity of such approaches in hypertensive patients.
Although Chen et al1 highlight the utility of this approach because “peripheral pressures can be measured noninvasively,” their noninvasively recorded peripheral pressure measurements had to be recalibrated to invasively measured data. Calibration experience with both radial tonometry and Finapres data collection suggests that these calibration errors can be substantial. Recalibration of such recordings to invasively collected data precludes application of the technique in outpatient or ward settings—the very places where most clinicians will be seeking to apply such techniques.
For these reasons, individualized TFs relying solely on noninvasive data collection (eg, simultaneous real-time carotid artery tonometry and Finapres), as some researchers have described,4 may actually permit more accurate resynthesis of central aortic pressure wave contours than generalized TFs.
There is a need for simple prospective studies to be reported from different groups of patients from those who provided data for the original creation of such generalized TFs. In this respect, it would be interesting to know how accurately the generalized TF of Chen and colleagues was able to predict central aortic pressures in the next 20 patients who were studied in their cardiac catheter laboratory. If they can calibrate these data solely using noninvasive pressure measurements, such as can be done on the ward or in the clinic, this would begin to offer useful information about the reliability of such a generalized TF. Without such prospective data and external validity checks, it is premature to make claims about “clinically acceptable accuracy.”
- Copyright © 1998 by American Heart Association
Chen C-H, Nevo E, Fetics B, Pak PH, Yin FCP, 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.
Karamanoglu M, O’Rourke MF, Avolio AP, Kelly RP. An analysis of the relationship between central aortic and peripheral upper limb pressure waves in man. Eur Heart J. 1993;14:160–167.
Plunkett BT, Lehmann ED, Kelly RP, Hayward CS, Avolio AP, O’Rourke MF. Relationship between the ascending aortic pressure waveform and the pressure wave determined in the finger by Finapres. Z Kardiol. 1996;85(suppl 3):136. Abstract.
Karamanoglu M, Feneley MP. A comparison of three methods for synthesis of the central aortic pressure waveform from the peripheral upper limb pulse in man. Aust NZ J Med. 1997;27:124. Abstract.
Dr Lehmann raises 2 concerns: (1) that broad applicability of the general transfer function (GTF) was inadequately tested because it was derived and applied to the same study group and (2) that we calibrated tonometry signals to central pressures, whereas peripheral calibration is clearly required for noninvasive implementation.
Testing the GTF in the same patient group from which it was derived did not compromise the 2 primary goals of our study: to determine if a GTF averaged from many patients performed adequately compared with individualized TFs from each patient (model robustness) and to test if a GTF derived under 1 condition (steady state) predicted data obtained under very different conditions (preload reduction) (predictive performance). Our study design did prevent predictive bias but did not eliminate interpatient variance. If differences between individual TFs (which were present and demonstrated by standard deviation about the GTF) were as critical to net TF performance as Dr Lehmann suggests, the variance of measured versus estimated pressure differences using the GTF would have been larger. In reality they were quite small. Furthermore, if changes in flow and mean pressure critically altered the TF, the GTF model would fail to predict responses during vena caval occlusion. We fully agree that predictive accuracy should also be tested in an independent group and have since performed such analysis (B. Fetics, BE, et al, unpublished data, 1998), applying the GTF from our prior studyR1 to 19 different patients. We again find close agree- ment between estimated and measured aortic systolic pres- sures (Psys): Psysmeasured=1.02×Psysest−5.1, r=0.995, P<10−6 (SEE=0.63 mm Hg); arterial compliance (C): Cmeasured=1.1×Cest−0.007, r=0.9, P<10−6 (SEE=0.053 mL/mm Hg); and good agreement between temporal waveforms (MSE=2.6±0.22 mm Hg).
There is no argument that individual patient TF differences exist, and for some estimation parameters, such as augmentation index or dicrotic notch definition, these differences likely preclude reliable use of a GTF. However, for many other useful parameters, such as systolic and pulse pressures and diastolic decay waveshape, TF differences do not appear critical to obtaining good estimates with clinically acceptable levels of accuracy. We again caution that our data are directly relevant to supine patients at rest and during preload reduction. It remains possible that greater discrepancies in the TF exist during exercise or with marked changes in systemic vascular resistance, and this will need further testing.
Regarding the second point, our goal was not to validate methods for noninvasive peripheral estimation of arterial systolic and diastolic pressures. Errors in noninvasive calibration, easily detected by mean pressure differences between cuff and central pressures, would have artificially enhanced interpatient TF variability. Our aim was to determine variability of the transform itself assuming correct calibration. Subsequent sensitivity analysis revealed only small errors in estimated systolic pressures for a ±10 mm Hg calibration error. Several methods for external pressure calibration exist and can work well,R2 and ongoing studies are using such methods. We too look forward to further studies of the GTF and testing of its applicability in the outpatient and ward settings, much the same as Dr Lehmann does.
Chen C-H, Nevo E, Fetics B, Pak PH, Yin FCP, 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.
Sharir T, Marmor A, Ting CT, Chen JW, Liu CP, Chang MS, Yin FCP, Kass DA. Validation of a method for noninvasive measurement of central arterial pressure. Hypertension.. 1993;21:74–82.