Author + information
- Sarah A Hope, BSc (Hons)MBChB* ()
- Ian T Meredith, MBBS, PhD, FACC and
- James D Cameron, MDMBBSBEMEngSc
We read with interest the study by Wilkinson et al. (1) in the March 20, 2002, issue of JACC. The investigators employed pulse wave analysis using sphygmorCor software (AtCor Medical, Sydney, Australia) for the derivation of central aortic waveforms from radial waveforms acquired by applanation tonometry, and they presented data on the derived aortic pulse pressure and augmentation index. The findings of the study are as expected; however, we are concerned that the technique described is being increasingly utilized, but remains poorly validated in the literature. The researchers state that the technique utilizes a “validated transfer function,” yet this assertion is not supported by the references they quote (2–5).
Karamanoglu et al. (2) derived a transfer function by frequency domain analysis of central aortic and applanation tonometry-acquired radial waveforms in 14 patients. They report systolic pressure alone without prospective validation. Takazawa et al. (4) applied the transfer function of Karamanoglu et al. (2) to an additional small cohort and found an 11 mm Hg discrepancy between measured and derived central aortic pressures, when tonometry-acquired pressures were calibrated to noninvasive brachial pressures, suggesting that this transfer function may not be generalizable to other subjects. These investigators also found the transfer function to underestimate central augmentation index on average, with considerable individual variability that was not quantified.
Segers et al. (5) applied the transfer function of Chen et al. (6), derived by a time domain, rather than a frequency domain, analysis. There may be important differences between transfer functions derived by frequency and time domain analyses in their ability to reconstruct central waveforms accurately, with potential advantages to the latter (7). Segers et al. (5) found wide individual variability in derived aortic augmentation index, with Bland-Altman 95% limits of agreement of approximately 30%. These findings support those of Fetics et al. (7), using a different transfer function derived by time domain analysis, who described a percentage error in augmentation index estimation of 54 ± 232% in 19 patients, but very good estimation of central systolic pressure, when tonometry-acquired pressures were calibrated to central aortic mean and diastolic pressures. These findings are consistent with the findings of Chen et al. (6) that augmentation index is highly dependent upon high-frequency components of the waveform, and the frequency response characteristics of applanation tonometry are such that high-frequency data is inevitably lost by this technique (8). The proprietary transfer function of the sphygomoCor software is unpublished, and it is unclear as to whether it has been derived by either frequency or time domain analysis. Pauca et al. (3) utilized the sphygmoCor software and prospectively evaluated it for the derivation of central blood pressures only, from invasively measured radial waveforms, when calibration of peripheral waveforms was not required. These data cannot validate the technique for use with noninvasively obtained radial waveforms.
We suggest that, although some transfer functions may adequately reproduce central aortic pressures from radial waveforms acquired by applanation tonometry, it is premature to suggest that the technique is well validated for noninvasive use, and published data suggest that the technique may be unreliable for the derivation of central aortic augmentation index. This technique should be properly validated before being applied to large-scale prospective intervention studies.
- American College of Cardiology Foundation
- Wilkinson I.B.,
- Prasad K.,
- Hall I.R.,
- et al.
- Karamanoglu M.,
- O’Rourke M.F.,
- Avolio A.P.,
- Kelly R.P.
- Chen C.H.,
- Nevo E.,
- Fetics B.,
- et al.