Phase-Contrast Magnetic Resonance Flow Quantification in Renal Arteries
Comparison With 133Xenon Washout Measurements
To assess the accuracy of 2D phase-contrast magnetic resonance (2D PC MR) renal artery flow measurements, data obtained with this technique were compared with those acquired with the 133Xenon-washout procedure. In addition, the 2D PC MR flow data were related to functional renal information as derived from selective arterial and venous renin sampling. In 53 patients suspected of having renovascular hypertension, MR angiography of the renal arteries was performed, followed by a 3-step angiographic procedure: (1) selective venous and arterial renin sampling; (2) assessment of the renal blood flow by means of the 133Xenon washout technique, and (3) conventional renal angiography. After initial assessment, 71 kidneys were left for analysis. The overall prevalence of renovascular disease ≥50% stenosis was 18%. Mean renal blood flow as assessed with the 2D PC MR technique showed a significant correlation with the 133Xenon washout flow measurements, with a Pearson correlation coefficient of 0.69 (2-tailed; P<0.01). PC MR blood flow measurements correlated poorly with the presence and/or severity of renovascular disease on conventional angiography (r=0.1, P=0.36). Likewise, no statistically significant correlation with either renal venous renin levels or the renin ratio could be identified. Measurement of renal artery blood flow with the use of a 2D PC MR technique is technically feasible. However, the mean renal artery blood flow correlates poorly with either the presence of renovascular disease on angiography or with renin levels. Further improvement of this technique is necessary before it can be applied on a larger scale.
Magnetic resonance angiography (MRA) is a promising noninvasive imaging modality for the screening of renovascular disease.1–3 Recent publications indicate that high-resolution 3D gadolinium-enhanced MR angiography is the preferred MR angiographic technique for assessment of the renal arteries.4–7 However, similar to other imaging methods used in the diagnosis of renal artery stenosis, most MRA techniques solely rely on the morphological assessment of the vasculature. To assess the hemodynamic consequences of a particular arterial lesion, additional functional tests are still required.
Published reports concerning several noninvasive techniques (eg, duplex ultrasound and/or nuclear clearance techniques) show a great variety in diagnostic accuracy as well as various specific limitations of these tests in the assessment of renal artery blood flow.8–10 Therefore, the demand for a reliable, noninvasive screening method remains valid.11
Magnetic resonance phase-contrast (PC MR) protocols have been proposed to provide the desired functional information through direct, noninvasive 2D PC MR renal artery flow quantification. The accuracy of this technique was verified both in vivo and in vitro by comparison with other noninvasive (eg, duplex ultrasound and/or nuclear clearance techniques) and invasive techniques (implantation of ultrasound probes) or by internal validation by comparing aortic flow measurements above the renal arteries with the total flow through the renal arteries and aorta distal of these arteries.12–14
In the current study, the 133Xenon washout technique was used as an independent test for comparative evaluation of the acquired 2D PC MR renal artery flow data in hypertensive patients. The 133Xenon washout technique requires selective arterial catheterization to allow local disposition of the tracer and therefore has the potential to provide information about blood flow of each kidney separately.
The primary objective of this investigation was to assess the accuracy of the 2D PC MR flow measurements in patients suspected of having renovascular hypertension by comparing the results of these measurements with those obtained by 133Xenon washout.
In addition, the 2D PC MR flow data were related to functional renal information, as obtained by selective arterial and venous renin sampling to assess the presence of renal artery stenosis. In this respect, conventional renal angiography was used as the standard of reference for the diagnosis of renal artery disease.
Over an 11-month period, all patients who were referred by the Department of Internal Medicine for angiographic evaluation of the renal arteries were prospectively enrolled in the study. Selection criteria included malignant, accelerated, and/or treatment-resistant hypertension, hypertensive retinopathy grade III or IV, loss of renal function after ACE inhibition, unilateral loss of kidney volume, or unexplained deterioration of renal function (change in serum creatinine >20 μmol/L within 12 months).
Three weeks before angiography, all antihypertensive medication was discontinued. In addition, patients were asked to use a salt-restricted diet containing 55 mmol of sodium per day during the last week before admission. The rationale for the latter is that data on renal blood flow or renin secretion should always be interpreted in relation to sodium intake and that standardization, therefore, is mandatory. Adherence to this diet was checked by measuring 24-hour sodium output at the time of admission to the hospital.
The institutional review board had approved the study protocol, and patients were required to sign an informed consent form before they were included in the study.
The protocol of this study was fitted with a strict sequence of diagnostic modalities to avoid mutual influences; therefore the MRA examination was always performed 1 day before the conventional angiography was planned.
During the angiographic procedure, three separate, successive steps were taken: (1) selective renal venous and arterial renin sampling; (2) assessment of renal blood flow by means of the 133Xenon washout technique; and (3) conventional intra-arterial renal angiography. Only after completion of the first two steps was contrast material administered.
MR Imaging Technique
All images were acquired on a 1.5-T, whole-body MR system (Gyroscan NT, release 6; Philips Medical Systems), using the quadrature body coil for signal transmission and reception. No abdominal compression was used.
After localizing the renal arteries, a non–cardiac-gated 3D PC MR angiographic (3D PC-MRA) sequence was performed to visualize the renal arteries.12 T1-weighted gradient-echo images were obtained in the transverse plane with the following parameters: TR/TE 27.0/8.5, flip angle 20°, two signals averaged, 75% echo sampling, 200×140 mm FOV, 103×128 mm, acquisition matrix, using a 256 reconstruction matrix, and flow-encoded velocity range (Venc) of 45 cm/s. Fourier interpolation was used to reconstruct sections thinner than those acquired.
To reduce the venous signal from the inferior caval vein, a transverse presaturation slab (60.0 mm thick) was placed inferior to the lower pole of the kidneys. The maximum intensity projection (MIP) algorithm was used to create angiographic projections in the transverse, coronal, and sagittal planes. Both the MIPs and the original transverse images were used to select an acquisition plane to the artery studied for flow quantification.
A 2D PC MR technique was used for flow data acquisition during normal breathing with the following parameters: TR/TE: 14/6; flip angle 20°, FOV: 300×210 mm, slice thickness: 6 mm; acquisition matrix: 256×108, and Venc: 120 cm/s. The scan was retrospectively triggered by using a peripheral pulse unit, yielding 25 heart phases at a frequency of 65 beats/min within a scan time of ≈2 minutes. The flow data were obtained in a plane perpendicular to the right and left renal artery, 8 to 15 mm from the origin.
To ensure optimal vessel demarcation and measurement reproducibility, all quantitative 2D PC MR flow measurements were analyzed with the use of a contour detection program on an off-line workstation (Sun Microsystems). The automated region of interest definition algorithm selected the region on the 2D PC MR images within an operator-defined ellipsis drawn generously around the vessel on the modulus images. Instantaneous flow (mL/s) was calculated from the individual velocity images by integrating velocity (cm/s) across the area (cm2) of the vessel. Mean flow in each vessel was calculated as the average of instantaneous flows across the cardiac cycle.
The procedure was carried out in the angiosuite of the department of radiology, which is equipped with both a radiographic system and NaI (Tl) gamma spectroscopy system. After selective catheterization of the renal vessels and before any intra-arterial contrast material was given; blood samples for determination of active plasma renin concentration (APRC) were drawn simultaneously from aorta and both renal veins.
Subsequently, mean renal blood flow (MRBF) for each kidney was assessed by means of the 133Xenon washout technique.15 Only after completion of these measurements, diagnostic intra-arterial renal angiography was performed with a commercially available digital subtraction system (Integris 5000; Philips Medical Systems). Angiographic images of the abdominal aorta and renal arteries were obtained in the anteroposterior and left and right oblique views, with injection of 30 mL iohexol (Omnipaque 300; Nycomed) through a 4F Universal Flush catheter (Cordis) positioned at the level of the renal arteries. Selective renal angiography was performed in the majority of cases, with injection of 12 mL iohexol through a 5F end-hole Cobra-2 or Simmons-2 catheter (Cordis).
Three experienced interventional radiologists who were blinded to the MR angiographic results independently read and interpreted the conventional angiograms. For cases in which the three observers had conflicting interpretations of the angiograms (>10% difference in degree of stenosis), a fourth interventional radiologist was asked to establish a definite diagnosis, based on his review of all three separate interpretations as well as on the angiographic images.
Intra-Arterial 133Xenon Washout Technique
The principles of the inert gas washout technique are based on the work of Kety.16 For any substance carried to an organ by the blood, it is evident that the amount that enters the organ within a specified time interval must equal the amount of the substance, which, during the same time interval, leaves the organ, plus the amount that is accumulated and metabolized in the organ. In the case of an inert unmetabolized substance such as xenon, no conversion takes place. Thus, derived from several equations, monitoring the disappearance of 133Xenon from the organ will allow calculation of flow per unit volume of tissue.15
The 133Xenon washout curves were analyzed as described earlier.15–18 In brief, after subtraction of background radiation, the disappearance of 133Xenon from the kidney, measured by an extracorporeal gamma-probe, is analyzed mathematically by means of a 2-phase exponential decay. Occasionally, a monophasic decline in activity is observed and the curve analyzed accordingly. The MRBF is calculated as the weighted average of the fast and the slow components.18 All analyses were performed with the use of Prism 3.0 software.
To convert the relative blood flow per kidney as determined from 133Xenon washout technique to absolute blood flow per kidney, it is necessary to know the total tissue volume of the separate kidneys. For this purpose computer tomography (CT) was introduced as an independent imaging modality, since the MR scout images were thought to be insufficient for adequate assessment of the kidney volume. Spiral CT was performed with a commercially available double-helix scanner (Elscint Ltd). After localizing the level of the kidneys, a volume acquisition was performed with the following protocol: 120 kV at 350 mA, 30 seconds of continuous exposure, 2.5-mm collimation, and 1.3 mm/s table speed. The axial source images were postprocessed on an off-line workstation (Sun Microsystems). The volumes of the kidneys were calculated by multiplying the renal surface area on every individual slice (mm2) by the slice thickness (mm) and adding up all slice volumes.
When performed this way, the 133Xenon washout study provides an accurate estimate of renal blood flow and in our hands had a variability of 8% for repeated measurements.19
Conventional angiography was regarded as the standard of reference for detection of renal artery stenosis. A significant stenosis was defined as luminal narrowing >50% of the vessel diameter. Renal arteries with evidence of fibromuscular dysplasia were graded as significantly stenosed (≥50%).
Statistical differences were assessed by means of the 2-sample Student’s t test. Associations between the renal artery blood flow quantification techniques, or active plasma renin ratio (APRCvein−APRCaorta)/APRCaorta and conventional renal angiography, were calculated by means of the Pearson correlation coefficient and orthogonal regression analysis.
Fifty-three patients with 106 kidneys were prospectively enrolled in this study: The mean age was 55 years (range, 30 to 79), and 60% were men. Mean systolic blood pressure measured at three separate consecutive occasions was 188 mm Hg; mean diastolic blood pressure was 109 mm Hg. Serum creatinine averaged 109 μmol/L. Conventional renal angiography showed supernumerary arteries in 16 kidneys (15%), with one kidney that showed three separate renal arteries. Total occlusion of the renal artery was found in three kidneys.
Kidneys with supernumerary renal arteries were excluded from analysis because a mismatch between the ipsilateral individual renal arteries in these cases could occur during the blinded first two steps of the angiographic procedure: sampling and 133Xenon washout studies. Also, the kidneys with a total occlusion of the renal artery were discarded.
In 14 of the remaining 87 kidneys (16%), the 2D PC MR quantitative flow images showed insufficient quality for automatic contour detection caused by improper slice positioning and/or to flow voids during systole caused by proximal stenoses. In another two (unilateral) cases it was not possible to obtain 133Xenon washout information, probably because of poor positioning of the intra-arterial catheter, thus leaving 71 corresponding 2D PC MR and 133Xenon washout flow measurements for further analysis.
In this group of 71 kidneys, conventional angiography showed 4 arteries with mild renovascular disease (20% to 50% stenosis) and significant renal artery stenosis (≥50%) in 13 arteries (overall prevalence18%). Significant bilateral disease was seen in 3 patients.
2D PC MR Versus 133 Xenon Washout Flow Measurements
Assessment of MRBF with the 2D PC MR technique correlated significantly with the 133Xenon washout flow measurements (Figure 1), with a Pearson correlation coefficient of 0.69 (2-tailed; P<0.01). Excluding two outlying results with relatively high renal artery blood flow measurements obtained with both techniques would have resulted in a correlation coefficient of 0.51. The MRBF to the right kidney was 322 mL/s and to the left kidney was 347 mL/s (overall mean flow per kidney, 334 mL/s) with the use of the 2D PC MR technique for flow quantification. With the use of 133Xenon washout for flow quantification, these respective flow data were 323, 318, and 321 mL/s.
Flow Measurements in Relation to Presence of Renal Artery Stenosis
The 2D PC MR renal artery blood flow measurements correlated poorly with the presence and/or severity of renovascular disease (% of stenosis) on conventional renal angiography (r=0.1, P=0.36) (Figure 2). Furthermore, in patients with unilateral renovascular disease on angiography, no significant difference in 2D PC MR blood flow measurements between both kidneys could be detected.
Adjusting the 2D PC MR flow measurements to the renal volume by calculating the renal flow index: Flow (mL/min)/kidney volume (cm3) and relating these to the presence and/or severity of renovascular pathology on conventional angiography failed to show a statistically significant relation. Similar observations in relation to the presence of renovascular disease were noted with the use of the 133Xenon washout technique for renal artery flow quantification, further acknowledging the good correlation between the two independent techniques.
APRC in the aorta averaged 24 mU/L. In both the right and left renal veins, mean renin concentrations were 27 mU/L. No statistically significant correlation between the renal artery flow measurements with the use of 2D PC MR or 133Xenon washout and the active plasma renin ratio (APRCvein−APRCaorta)/APRCaorta could be identified (r=0.4, P=0.1, and r=0.3, P=0.26, respectively).
The present data show that the use of 2D PC MR blood flow measurements enables hemodynamic evaluation of the renal arteries in a cohort of hypertensive patients with clinical suspicion of renovascular disease. The absolute flow measurements in the renal arteries are comparable with results of other published series.14,20
As demonstrated in Figure 1, the 2D PC MR renal artery flow measurements correlated significantly with the corresponding renal artery flow data as obtained independently by the 133Xenon washout technique, with a Pearson correlation coefficient of 0.69 (2-tailed; P<0.01). This is a promising level of correlation, especially considering that these flow measurements are not completely identical; that is, 2D PC MR relates directly to the renal artery flow, whereas 133Xenon washout represents the flow through the renal microcirculation.
The calculated correlation is influenced to some extent by the two outliers shown in Figure 1, with relatively high renal artery blood flow measurements obtained with both techniques. Analysis without these two values would have resulted in a correlation coefficient of 0.51. The dispersion in the scatterplot in Figure 1 rises with the increase in measured MRBF, indicating a stronger correlation between the two methods at flow levels that are particularly relevant for the distinction between hemodynamically significant and nonsignificant renovascular disease.
Despite this correlation, we failed to identify a statistically convincing relation between the presence and/or severity of renovascular pathology on conventional, intra-arterial renal angiography, and either 2D PC MR or 133Xenon washout flow quantification data. Moreover, in patients with unilateral renovascular disease, we could not find a significant difference in blood flow between the two kidneys with either technique.
Likewise, no statistically significant correlation could be identified between active plasma renin ratio as a measure of functional hemodynamic abnormalities and the renal flow data with the use of either 2D PC MR or the 133Xenon washout technique.
Given the correlation between the two independent flow quantification techniques, the lack of discriminating power of 2D PC MR flow quantification for assessing the presence of renovascular disease cannot, a priori, be attributed to inaccuracies of the applied MR technique.
Our observations contrast with those from several publications in international literature on this topic.13,21 This seems, at least partially, related to methodological differences. For instance, as opposed to Binkert et al,21 we did not inject paramagnetic contrast before the phase-contrast flow acquisition, which might have allowed for a better delineation of the renal arteries and thereby a more accurate detection of the transverse vessel area.22 Moreover, we did not include a qualitative analysis of the flow velocity curve, since a reduction of MRBF is considered to be the most important parameter for differentiating between significant and nonsignificant renal artery stenosis.13 Finally, the differences in prevalence of renovascular disease between the respective studies make dissimilarities in inclusion criteria and patient selection probable, which will affect the result of these examinations and make mutual comparisons cumbersome.
Apart from the procedural differences, more fundamental explanations of physiological nature also must be considered. Data in literature indicate that the degree of stenosis does not correlate well with flow measurements.23 Therefore, relating a single deviating renal artery flow measurement to the presence and/or severity of renal artery stenosis may prove to be unjustified. Serial selective renal artery flow measurements under controlled circumstances may be necessary to detect changes of mean renal artery flow volume over time. Furthermore, a chronic and slowly progressive disorder such as renovascular disease triggers various compensatory mechanisms in one or both kidneys to maintain a normal renal artery blood flow, possibly even up to a very late stage of the disease. Thus, adaptation of endothelial function may take place at a much earlier stage of renovascular disease than what is now considered significant stenosis.19 Moreover, the negligible difference in flow measurements between normal and abnormal kidneys can also be explained by either adaptation of the abnormal kidney or by underlying disease, for example, nephrosclerosis, in the “normal” kidney.
Even with serial renal artery flow measurements, the adaptation of a kidney to a stenotic lesion might not be fully appreciated. Serial quantitative studies of renoparenchymal perfusion with pharmacological challenges such as captopril may be necessary to obtain the essential functional information that allows to define the therapeutic options for a particular patient.
A major shortcoming of this study relates to the lack of a sufficient number of patients with significant renovascular disease on conventional angiography. This limits statistical assessment of the measurements in relation to the degree of stenosis and makes additional research imperative. A second limitation is relatively high number of excluded original examinations, particularly reflecting the delicacy of the 2D PC MR flow quantification technique in patients. Furthermore, we did not repeat the 2D PC MR flow renal artery flow measurements to assess the reproducibility of these flow measurements. However, in a previously published study we have shown that quantitative measurement of renal artery blood flow with 2D PC MR is highly reproducible.12 Still, the spread of MR results in the current study calls for further improvements before widespread clinical application of this technique can be advocated.
In conclusion, the good correlation between two independent flow quantification techniques, that is, 2D PC MR and 133Xenon washout, indicates that mean renal artery blood flow assessment with a 2D PC MR technique is feasible. However, the poor correlation of both techniques with either the presence of renovascular disease on conventional angiography or the active plasma renin ratio demonstrates a lack of discriminating power of our study protocol for the functional evaluation of the renal arteries.
Despite the good correlation of 2D PC MR flow with another independent, method (133Xenon washout) to measure renal blood flow, many more studies are necessary to establish the place of MR flow quantification for the detection of the presence and degree of renal artery stenosis. Also, more work is needed to evaluate whether MR flow data obtained either under basal conditions or after, for instance, captopril challenge, are able to detect whether a stenosis leads to abnormal physiology and ultimately true renovascular hypertension.
- Received September 10, 2002.
- Revision received October 7, 2002.
- Accepted November 5, 2002.
Schoenberg SO, Bock M, Knopp MV, Essig M, Laub G, Hawighorst H, Zuna I, Kallinowski F, van Kaick G. Renal arteries: optimisation of three-dimensional gadolinium-enhanced MR angiography with bolus-timing-independent fast multiphase acquisition in a single breath-hold. Radiology. 1999; 211: 667–679.
Breitenseher von M, Kainberger F, Hübsch P, Trattnig S, Baldt M, Barton P, Karnel F. Screening of renal artery stenosis: early results of colour Doppler sonography. Fortschr Röntgenstr. 1992; 156:3: 228–231.
Wierema TKA, Houben AJHM, Kroon AA, Koster D, van der Zander K, van Engelshoven JMA, de Leeuw PW. Nitric oxide dependency of renal blood flow in hypertension with and without renal artery stenosis. J Am Soc Nephrol. 2001; 12: 1836–1843.