JOURNAL OF MAGNETIC RESONANCE IMAGING 22:6772 (2005)

Original Research

A Comparison of Noninvasive MRI-Based Methods of Estimating Pulmonary Artery Pressure in Pulmonary Hypertension
Roald J. Roeleveld, MD,1 J. Tim Marcus, PhD,2 Anco Boonstra, MD, PhD,1 Pieter E. Postmus, MD, PhD,1 Koen M. Marques, MD,3 Jean G.F. Bronzwaer, MD, PhD,3 and Anton Vonk-Noordegraaf, MD, PhD1*
Purpose: To assess the accuracy of several noninvasive MRI-based estimators of pulmonary artery pressure by comparing them with invasive pressure measurement. Materials and Methods: We compared ve MRI methods with invasive pressure measurement by catheterization, in one group of pulmonary hypertension (PH) patients. Doppler echocardiography was included as a reference method. Main inclusion criterion was a mean pulmonary artery pressure above 25 mmHg at catheterization. MRI velocity quantication was used to obtain pulmonary ow acceleration and ejection times, and pulse wave velocity. The ventricular mass index was also assessed on MRI. Two commercially available 1.5-T systems were used for this study. Results: Data from 44 patients were analyzed. Correlation of acceleration time with mean pressure was: r 0.21, P 0.21, correlation of the acceleration/ejection time ratio with systolic pressure was: r 0.26, P 0.01. The ventricular mass index showed the best correlation with mean pressure, with r 0.56, P 0.001. Using the pulse wave velocity and the cross-sectional area of the pulmonary artery, the mean pressure could not be estimated accurately. Conclusion: Accurate estimation of pulmonary artery pressure in PH patients was not feasible by the MRI estimators studied. These noninvasive methods cannot replace right heart catheterization at this moment. Key Words: pulmonary arterial hypertension; magnetic resonance imaging; echocardiography; pressure estimation; right ventricular mass J. Magn. Reson. Imaging 2005;22:6772. 2005 Wiley-Liss, Inc.

1 Department of Pulmonology, VU University Medical Center/Institute for Cardiovascular Research (ICaR-VU), Amsterdam, The Netherlands. 2 Department of Physics and Medical Technology, VU University Medical Center/Institute for Cardiovascular Research (ICaR-VU), Amsterdam, The Netherlands. 3 Department of Cardiology, VU University Medical Center/Institute for Cardiovascular Research (ICaR-VU), Amsterdam, The Netherlands. Contract grant sponsor: GlaxoSmithKline Netherlands BV, Zeist, The Netherlands. *Address reprint requests to: A.V.-N., VUmc, Dept. of Pulmonology, De Boelelaan 1117, 1081 HV Amsterdam, The Netherlands. E-mail: a.vonk@vumc.nl Received October 12, 2004; Accepted March 15, 2005. DOI 10.1002/jmri.20338 Published online in Wiley InterScience (www.interscience.wiley.com).

PULMONARY HYPERTENSION (PH) is a phenomenon occurring in advanced stages of many diseases affecting the pulmonary circulation. Detection and follow-up of an elevated pressure in the pulmonary circulation is an important factor in determining therapeutic strategies and prognosis. However, the pressure in the pulmonary circulation can only be assessed by right heart catheterization (RHC), which is an invasive procedure and thus not well suited for either screening or frequently repeated follow-up measurements. Echo has become the most widely used technique to estimate systolic pulmonary artery pressure (sPAP), using the peak velocity of the regurgitant jet through the tricuspid valves. However, several studies including large groups of patients reported a poor correlation between echo and sPAP in PH patients, showing that echo cannot replace the gold standard (13). MRI has been proposed as an accurate alternative for echo in estimating pulmonary artery pressure (PAP); several estimators based on different MRI techniques have been described in recent years (4 11). In a small group of 13 PH patients Tardivon et al (4) found a correlation (r 0.9) between mean pulmonary artery pressure (mPAP) and acceleration time (AT): the time from the onset of forward ow to the moment of maximum ow velocity (Vmax) in the main pulmonary artery, as measured by velocity-encoded cine MRI. Using a slightly different technique, the study by Wacker et al (9) showed similar results (r 0.82, P 0.0011, N 12). Marcus et al (10) demonstrated a correlation between sPAP and the AT/ejection time (ET) ratio, AT divided by ET (r 0.68, P 0.02, N 11). A different technique also based on velocity-encoded MRI was used by Laffon et al (12), in a group of 15 PH patients. The blood ow velocity and the cross-sectional area (CSA) of the main pulmonary artery were measured by MRI. Using these measurements as well as correlations of pressure wave velocity and pulse pressure with mPAP, it was possible to estimate a frame in which the actual mPAP should lie, with a reliability of 87%. In a later publication by Laffon et al (17), the method was modied, resulting in a correlation with the measured pressure of r 0.92. This modied Laffon method also takes into account some other patient
67

2005 Wiley-Liss, Inc.

68

Roeleveld et al.

data: length, weight, and heart rate. The nal method in our study was published by Saba et al (6). In their study, right and left ventricular mass (RVM and LVM, respectively) were measured by MRI, and the ventricular mass index (VMI) calculated as VMI RVM/LVM. The index was found to correlate well with invasive measurements (r 0.81). When taking VMI 0.6 as an indicator for PH, a sensitivity of 84% and a specicity of 71% was achieved in a group of 26 subjects, 19 of whom were PH patients. Although correlations between PAP and MRI estimators were signicant in all the studies mentioned above, these studies were performed in relatively small groups of patients and have not been repeated until now. Furthermore, inclusion criteria varied between the different studies, making a direct comparison between the methods used in these studies very hard. For these reasons, the clinical value of these estimation methods is still unclear. The goal of our study was to assess the accuracy of the different MRI methods in estimating PAP, using RHC as the gold standard and routine echocardiography as a comparison method. MATERIAL AND METHODS Study Subjects and Design The institutional ethics committee approved this study. All patients gave informed consent prior to inclusion. The patients were referred to the VU University Medical Center for evaluation of suspected PH between June 2000 and January 2003. Only patients with proven PH, dened as mPAP 25 mmHg (measured invasively) and normal wedge pressure were included. RHC was performed with a uid-lled, single lumen, multipurpose catheter via the transfemoral approach (6 Fr. MPA 1; Cordis Co., Miami Lakes, USA). Patients were not sedated during the procedure. All patients underwent extensive investigations to establish a cause of their PH, similar to the diagnostic strategy as published by Gaine and Rubin (13). For each patient, all mentioned measurements were completed within a seven-day period. During this period, patients were in a stable clinical condition and no changes in therapy were made, with the exception of anticoagulants to allow a safe RHC. Measurements performed outside the seven-day period were scored as not available. All data was gathered according to a prospective study protocol and analyzed after completing the data acquisition. MRI Acquisition MRI scans ware made by a dedicated operator, using either a Siemens 1.5-T Vision or Sonata system (Siemens Medical Solutions, Erlangen, Germany). With the aid of several localizer images, the short-axis (SA) orientation of the heart was selected. Subsequently, a series of parallel SA image planes was dened starting at the base of the left and right ventricle and including both ventricles from base to apex. At every SA plane, a cine acquisition was then performed at inspiratory breathhold. For the SA cine-imaging on the Vision sys-

tem, a gradient-echo pulse sequence was used with segmented k-space, 7 ky lines per heartbeat, and a data-acquisition window of 80 msec. Echo-sharing was used, yielding a temporal frame at every 40 msec. The acquisition settings were: excitation angle 25; eld of view 219 250 mm; matrix size 126 256. Slice thickness was 6 mm and interslice gap 4 mm, yielding a slice distance of 10 mm. For the Sonata system, a steady state free precession (True-FISP) pulse sequence was used with 11 ky lines per heartbeat and a dataacquisition window of 34 msec. Excitation angle was 60; eld of view 228 280 mm; matrix size 208 256. Flow in the main pulmonary artery was measured using MR velocity quantication. An image plane was positioned perpendicular to the main pulmonary artery, just above the level of the valves. Velocity encoding was perpendicular to the image plane, thus parallel to the ow in the main pulmonary artery. A two-dimensional gradient-echo pulse sequence with excitation angle of 25 and echo time of 5 msec was used. The phase encoding steps of two different acquisitions (repetition time 12 msec) were interleaved, alternating one with velocity encoding of phase and one without. Subtracting of the resulting phase maps compensated for phase changes caused by inhomogeneity of the magnetic eld, leaving only velocity-related phase changes. The resulting temporal resolution was 24 msec. The velocity sensitivity was set at 150 cm/second, eld of view was 225 360 mm, and matrix size was 140 256. The number of phases was adapted for each patient, thus ensuring that a complete cardiac cycle was recorded, regardless of the heart rate. The acquisition started directly after the R-wave of the electrocardiogram, which served as trigger. The MRI acquisition protocol has also been described in an earlier publication by Marcus et al (14). MRI Postprocessing The SA images were postprocessed using the MASS software package, version 5.0 (Department of Radiology, Leiden University Medical Center, and Medis Medical Imaging Systems, Leiden, The Netherlands). In diastole and systole, epi- and endocardial contours were manually traced, as well as the contours of the papillary muscles. The interventricular septum was considered part of the left ventricle. The muscle masses for the left and right ventricle were calculated by multiplying the muscle cross-sectional areas with the slice thickness and the specic density of cardiac muscle, 1.05 g/cm3, and adding the values from all slices. The VMI was then calculated by dividing RVM by LVM. Where applicable, the ow curves were corrected for offset errors using a method similar to the one described by Walker et al (15). The CSA was manually outlined, and its change during the cardiac cycle (CSA) calculated. The maximum blood ow velocity as well as AT and ET were read directly from the ow curves. A previous study by Laffon (12) resulted in the formulas c

(CSA*PAPPI * CSA)

(1)

Estimators of Pulmonary Artery Pressure in PH

69

and R c/V max (1/V max)(CSA*PAPPI * CSA) 1/2 (2) to calculate pulse wave velocity c and ratio R, in which is the blood volemic mass of 1060 kg/m3 and PAPP is the pulmonary artery pulse pressure. The correlations of pulse wave velocity c or the ratio R with mPAP and of PAPP and mPAP were taken from the original publication by Laffon (12). To obtain an estimate of mPAP for each patient, rst the ratio R is calculated using the CSA measured on MRI and an arbitrary PAPP of 15 mmHg. Using the R value together with the R vs. mPAP correlation results in an estimation of mPAP. This, in turn, is used to estimate PAPP via the PAPP vs. mPAP-graph. This estimated PAPP is then used in the next cycle of calculations, and these iterations were repeated until two successive estimates of mPAP were separated by less than 1%. The modied Laffon method was applied directly and required only the ow velocity in the pulmonary artery, its surface area and the height, weight, and heart rate of the study subjects. These parameters were directly entered in the formulae as supplied in the article. Doppler Echocardiography This procedure was performed by experienced cardiologists, who were aware of the suspected diagnosis of the patients but unaware of the study setup. The sPAP was estimated based on the modied Bernoulli equation: sPAP RAP 4v2, in which RAP is the right atrial pressure, and v is the peak velocity of tricuspid regurgitation. The RAP was estimated to be 15, 10, or 5 mmHg depending on the variation in size of the inferior vena cava with inspiration: at no collapse, RAP was estimated at 15 mmHg, with partial collapse RAP 10 and with a complete collapse RAP 5.

Figure 2. Correlation between pulmonary artery ow acceleration time divided by ejection time (AT/ET ratio) as measured by MRI velocity quantication, and the invasively measured systolic pulmonary artery pressure (sPAP). The dashed line represents the correlation line: r 0.257, P NS.

Statistical Analysis Linear regression analysis and Pearsons correlation coefcient were used to describe the relations between RHC and the various noninvasive estimations. A Pvalue 0.05 was considered statistically signicant. The method described by Bland and Altman (16) was used to investigate the levels of agreement between RHC and the noninvasive estimators. RESULTS General Findings and Invasive Measurements A group of 47 patients was included: 34 females and 13 males, aged 21 to 75 years (mean: 44 years). In accordance with the recently adopted Venice classication, 42 patients had pulmonary arterial hypertension (PAH), four had PH due to chronic thromboembolic disease, and one due to left-sided heart disease. At RHC, the sPAP ranged from 50 to 141 mmHg (mean: 83 mmHg), mPAP was 2779 mmHg (mean: 52 mmHg). The mean heart rate was 85 16 beats per minute (bpm). Data from MRI and echocardiography were available from 44 of the 47 included patients. MRI Results During the MRI acquisitions, the average heart rate was 83 15 bpm., not signicantly different from that at the invasive tests. The results of correlating AT with mPAP (r 0.211, P 0.21) and the AT/ET ratio with sPAP (r 0.257, P 0.096) are shown in Figs. 1 and 2, respectively. Using the correlations and graphs supplied by Laffon (12), the framing values of mPAP were calculated from the CSA and Vmax measured on MRI in 39 patients; considering the importance of the CSA in this estimation method, ve scans were discarded, as the ow cines were judged to be acquired too close to the pulmonary artery valves. The results are shown in Fig. 3: in 12 of the 39 patients (31%), the actual mPAP was adequately framed by the minimal and maximal estima-

Figure 1. Correlation between pulmonary artery ow acceleration time (AT) as measured by MRI velocity quantication, and the invasively measured mean pulmonary artery pressure (mPAP). The dashed line represents the correlation line: r 0.211, P NS.

70

Roeleveld et al.

Figure 3. Results of the pressure estimation method as described by Laffon et al (12) (see Materials and Methods) and invasively measured mean pulmonary artery pressure (mPAP). Each couple of triangles and their connecting line form a framing window in which the actual pressure should lie. This was the case in 32% (12/39) of our patients.

Figure 5. Correlation between the ventricular mass index as assessed by short axis cine MRI, and invasively measured mean pulmonary artery pressure (mPAP). The horizontal line at y 0.6 represents a cutoff value, a mass index 0.6 was taken as indicative for pulmonary hypertension. The dashed line represents the correlation line: r 0.056, P 0.0001.

tion. The modied Laffon method, also applied in 39 patients, resulted in an estimation of mPAP that correlated with the actual mPAP with a correlation coefcient of r 0.21 (P not signicant [NS]) as shown in Figure 4. The LVM ranged from 78 to 202 g (mean: 133 g), RVM from 55 to 185 g (mean: 97 g). The VMIs calculated were between 0.41 and 1.65, with a mean of 0.74. The correlation coefcient between VMI and mPAP, shown in Fig. 5, was: r 0.56, P 0.001. When accepting the proposed value of VMI 0.6 as indicative of PH, nine patients would fail to meet that criterion, even though they had mPAPs above 25 mmHg. Our inclusion criteria did not allow for a calculation of a false-positive rate, but the false-negative rate was thus 20%. Echocardiography On nine of the 44 echos performed, estimation of sPAP failed because of either poor image quality or absence of a tricuspid regurgitation jet. The 35 sPAP estimations ranged from 41 to 223 mmHg. As shown in Fig. 6, the

correlation between sPAP on echo and by RHC was low, yet statistically signicant: r 0.38, P 0.026. Two obvious outliers are observed in the graph; without these, the correlation reads: r 0.55, P 0.001. Figure 7 shows the Bland-Altman plot of the comparison between RHC and echo with the 95% limits of agreement interval from 64% to 54%, and a bias of 5%. DISCUSSION Comparing diagnostic techniques is usually initially done in normal, healthy subjects. For a parameter with a certain range of normal values this may result in a good correlation between the actual and estimated physiological values. However, whether this will also be the case in the diseased state has to be demonstrated as well. This is especially crucial if the investigated parameter suddenly has values that are several times larger than the normal physiological range for which the estimation method was originally validated. This is the

Figure 4. Correlation between the mean pulmonary artery pressure (mPAP) estimated using the modied Laffon method, and invasively measured mPAP. The dashed line represents the correlation line: r 0.21, P NS.

Figure 6. Correlation between Doppler echocardiographic estimation and invasive measurement of systolic pulmonary artery pressure (sPAP) in mmHg. The dashed line represents the correlation line: r 0.375, P 0.026.

Estimators of Pulmonary Artery Pressure in PH

71

Figure 7. Bland-Altman plot of the agreement between Doppler echocardiographic estimation and invasive measurement of systolic pulmonary artery pressure (sPAP). The dashed lines represent the upper and lower 95% condence interval limits of agreement, 64.4% and 54.4%, respectively. The bias of 5.0% is not represented in the graph, as it would be very hard to distinguish from the horizontal axis.

situation in patients with pulmonary hypertension. Furthermore, when investigating a population with both healthy and pulmonary hypertensive subjects, it is conceivable that the overall correlation between methods is quite good, while within each subgroup (healthy or PH) the correlation is far less. Therefore we decided to evaluate the described techniques only in subjects with the disease. This is, besides the number of investigated patients, the most important difference compared to earlier studies. It may also explain the lower correlation coefcients found in this study in comparison with the original studies. Including only patients, and not a random group of subjects or a control group, does however imply that no conclusions on sensitivity, specicity or predictive values of the methods under investigation can be drawn. Since all parameters could be derived from the same large group of PH patients, this study allows a fair comparison between the different MRI techniques to estimate PAP, using RHC as a gold standard. For clinical and logistical reasons MRI, echo and invasive measurements could not be performed simultaneously. The seven-day time window we set for the measurements was arbitrary, and based upon the average time needed to complete all diagnostic investigations in a patient with suspected PH. The gap between pressure estimations by MRI and RHC was on average ve days. Furthermore, caution was taken that heart rate during the MRI measurements was similar to that during the invasive procedure, and no signicant difference in heart rate was observed. For routine clinical use, any noninvasive estimation method should ideally combine high accuracy and relative simplicity of performance. Using an estimator that is inaccurate is simply not justied, and using one that takes considerably more resources (time, special equipment, etc.) than the gold standard itself loses much of its practical value as well. Compared to MRI, echo has the advantages of being widely available, less expensive, and easy to perform. The added value of MRI should therefore be found in a higher accuracy.

The methods based on properties of the MRI ow curve are relatively easy to perform, as both MRI acquisition as well as postprocessing can be done in a matter of minutes. The correlation of AT and the AT/ET ratio with PAP were, however, so poor that these methods are unlikely to become of any use in clinical practice. Using the MRI-measured pulse wave velocity c and the cross-sectional surface area of the pulmonary artery in estimating mPAP in the way described by Laffon et al (12), turned out to be a very labor-intensive operation. Furthermore, it did not lead to an adequate estimation of mPAP. Although some of the actual mPAP values were effectively framed by the interval of estimations, most (27 out of 39) were not. This in spite of the very large intervals of up to 70 mmHg between minimal and maximal estimate. Apparently, the lower mPAP values, up to approximately 50 mmHg, are overestimated, whereas values above 65 mmHg seem to be underestimated. Recalculating the correlations on which the method by Laffon was based (c vs. mPAP and PPAP vs. mPAP) lead to substantially different results than the ones published and using these for the iterative calculations also failed to lead to a reliable estimation of mPAP (data not shown). This is mainly caused by a very poor correlation of PAPP with mPAP in our patient group, a correlation that is crucial to the method. Other limitations, such as the difculty in correctly measuring the minimal CSA and possible pressure wave phenomena have been acknowledged by the authors themselves as well. A modied method was published later (17) and included several physical parameters of the study subjects. In our present study, we demonstrate that this estimation is also too inaccurate to be used in practice. Of all MRI-based parameters tested, only the VMI, as described by Saba et al (6), was found to correlate signicantly with mPAP. Earlier work by Pattynama et al (18) and Alfakih et al (19) have already shown that these types of measurements are also sufciently reproducible. However, whether a correlation coefcient of 0.56 and a false-negative rate of 20% (if VMI 0.6 is taken as indicative of PH) is enough for this method to be of use in clinical practice is doubtful. As ventricular masses will not change instantaneously, this method will also not be useful in rapidly changing hemodynamic situations, e.g., the few PH patients who are responsive to vasodilators. Finally, both the acquisition as well as the postprocessing take a certain amount of time, which may be a problem depending on hospital logistics. Of course, the drawing of the endo- and epicardial contours on the short-axis images may be automated instead of done manually, saving a lot of time, but reliability of the method then comes to depend on the accuracy of the contour-detection algorithm. In this study, all patients underwent routine echocardiography, which was used for comparison with the MRI estimators. These investigations were performed by experienced cardiologists who were unaware of this study, reecting everyday medical practice. Echocardiographic estimation of sPAP in our study showed a signicant correlation with invasively measured sPAP (r 0.38, P 0.05), but also a considerable amount of

72

Roeleveld et al.
in patients awaiting lung transplantation. J Heart Lung Transplant 2001;20:833 839. Chan KL, Currie PJ, Seward JB, Hagler DJ, Mair DD, Tajik AJ. Comparison of three Doppler ultrasound methods in the prediction of pulmonary artery pressure. J Am Coll Cardiol 1987;9: 549 554. Tardivon AA, Mousseaux E, Brenot F, et al. Quantication of hemodynamics in primary pulmonary hypertension with magnetic resonance imaging. Am J Respir Crit Care Med 1994;150:1075 1080. Mousseaux E, Tasu JP, Jolivet O, Simonneau G, Bittoun J, Gaux JC. Pulmonary arterial resistance: noninvasive measurement with indexes of pulmonary ow estimated at velocity-encoded MR imagingpreliminary experience. Radiology 1999;212:896 902. Saba TS, Foster J, Cockburn M, Cowan M, Peacock AJ. Ventricular mass index using magnetic resonance imaging accurately estimates pulmonary artery pressure. Eur Respir J 2002;20:1519 1524. Kondo C, Caputo GR, Masui T, et al. Pulmonary hypertension: pulmonary ow quantication and ow prole analysis with velocity-encoded cine MR imaging. Radiology 1992;183:751758. Frank H, Globits S, Glogar D, Neuhold A, Kneussl M, Mlczoch J. Detection and quantication of pulmonary artery hypertension with MR imaging: results in 23 patients. AJR Am J Roentgenol 1993;161:2731. Wacker CM, Schad LR, Gehling U, et al. The pulmonary artery acceleration time determined with the MR-RACE-technique: comparison to pulmonary artery mean pressure in 12 patients. Magn Reson Imaging 1994;12:2531. Marcus JT, Vonk-Noordegraaf A, Roeleveld RJ, et al. Impaired left ventricular lling due to right ventricular pressure overload in primary pulmonary hypertension: noninvasive monitoring using MRI. Chest 2001;119:17611765. Murray TI, Boxt LM, Katz J, Reagan K, Barst RJ. Estimation of pulmonary artery pressure in patients with primary pulmonary hypertension by quantitative analysis of magnetic resonance images. J Thorac Imaging 1994;9:198 204. Laffon E, Laurent F, Bernard V, De Boucaud L, Ducassou D, Marthan R. Noninvasive assessment of pulmonary arterial hypertension by MR phase- mapping method. J Appl Physiol 2001;90:2197 2202. Gaine SP, Rubin LJ. Primary pulmonary hypertension. Lancet 1998;352:719 725. Marcus JT, Vonk-Noordegraaf A, De Vries PM, et al. MRI evaluation of right ventricular pressure overload in chronic obstructive pulmonary disease. J Magn Reson Imaging 1998;8:999 1005. Walker PG, Cranney GB, Scheidegger MB, Waseleski G, Pohost GM, Yoganathan AP. Semiautomated method for noise reduction and background phase error correction in MR phase velocity data. J Magn Reson Imaging 1993;3:521530. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307 310. Laffon E, Vallet C, Bernard V, et al. A computed method for non invasive MRI assessment of pulmonary arterial hypertension. J Appl Physiol 2004; 96: 463 468. Pattynama PM, Lamb HJ, Van der Velde EA, Van der Geest RJ, Van der Wall EE, De Roos A. Reproducibility of MRI-derived measurements of right ventricular volumes and myocardial mass. Magn Reson Imaging. 1995;13:53 63. Alfakih K, Plein S, Thiele H, Jones T, Ridgway JP, Sivananthan MU. Normal human left and right ventricular dimensions for MRI as assessed by turbo gradient echo and steady-state free precession imaging sequences. J Magn Reson Imaging. 2003 Mar;17:3239. Rich S, DAlonzo GE, Dantzker DR, Levy PS. Magnitude and implications of spontaneous hemodynamic variability in primary pulmonary hypertension. Am J Cardiol 1985;55:159 163. Galie N, Ussia G, Passarelli P, Parlangeli R, Branzi A, Magnani B. Role of pharmacologic tests in the treatment of primary pulmonary hypertension. Am J Cardiol 1995;75:55A 62A. Mukerjee D, St George D, Knight C, et al. Echocardiography and pulmonary function as screening tests for pulmonary arterial hypertension in systemic sclerosis. Rheumatology (Oxford). 2004 Apr; 43:461 466.

errors; less than 15% of all performed echos was accurate within 10% of the RHC. A physiological variation coefcient in pulmonary pressures of 8% has been reported. Estimation of the 95% condence interval for an individual patient indicated that for a signicant difference, the PAP must show a change of 22% (20,21). When taking a window of 22% as allowed difference between echo and RHC, 49% of the echos can be seen as adequate. These results emphasize that although echocardiography is a useful noninvasive screening method for the diagnosis of PH, this technique is of little value in the reliable estimation of PAP, which is in line with the ndings of two recent studies (1,22). Several other methods of estimating pulmonary hemodynamics with the aid of MRI are known, but these were left out of our study for various reasons; methods concerning pulmonary perfusion and ow were also not under investigation in this study. Mousseaux et al (5) found a close correlation (r 0.89, P 0.0001) between the ratio of the maximal slope for the ascending inow rate to the acceleration volume and pulmonary vascular resistance (PVR), but not with PAP, as we sought. Kondo et al (7) found the ratio of maximal to spatial mean velocity in the pulmonary artery to be higher in PH patients than in healthy subjects. PH patients were also found to have a lower stroke volume, and qualitatively different ow proles during the cardiac cycle. Although those studies are of clinical value, the described MRI method was not aimed to estimate PAP, and thus was not included in our study. The method published by Murray et al (11) was able to estimate PAP, using the ratio of the diameters of the main pulmonary artery and midthoracic descending aorta (r 0.7, P 0.01). This method was not tested in our study, as we expected a much lower correlation because of the inclusion of several patients with systemic hypertension, mainly related to an underlying disease of the scleroderma spectrum. The study of Frank et al (8) was not repeated as its main parameter (correlation of PAP with the thickness of the right ventricular wall) was quite closely related to the method of Saba et al. (6) using the VMI. We considered this latter method superior, as measuring only the thickness at one certain point of an irregularly shaped body can easily result in measurement error, whereas the VMI is calculated from the entire muscle volume. In conclusion, accurate estimation of the PAP in patients with known PH is not feasible by means of the MRI estimators studied. These noninvasive methods cannot replace RHC at this moment. However, for screening purposes, a highly reliable, noninvasive method would be quite useful in clinical practice, and further research in this direction should be stimulated. REFERENCES
1. Arcasoy SM, Christie JD, Ferrari VA, et al. Echocardiographic assessment of pulmonary hypertension in patients with advanced lung disease. Am J Respir Crit Care Med 2003;167:735740. 2. Homma A, Anzueto A, Peters JI, et al. Pulmonary artery systolic pressures estimated by echocardiogram vs cardiac catheterization

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13. 14.

15.

16.

17.

18.

19.

20.

21.

22.