Detection of Renal Artery Stenosis:

AJR:177, November 2001

1123

Detection of Renal Artery Stenosis:

Prospective Comparison ofCaptopril-Enhanced Doppler Sonography, Captopril-Enhanced Scintigraphy, andMR Angiography

OBJECTIVE

.

The objective of our study was to compare the value of captopril-enhancedDoppler sonography, captopril-enhanced renal scintigraphy, and gadolinium-enhanced MRangiography for detecting renal artery stenosis.

SUBJECTS AND METHODS

.

Forty-one patients with suspected renovascular hyper-tension were prospectively examined with captopril-enhanced Doppler sonography, captopril-enhanced renal scintigraphy, gadolinium-enhanced MR angiography, and catheter angiography.The sensitivity and specificity of each technique for detecting renal artery stenosis measuring50% or greater and 70% or greater were compared using the McNemar test. Positive and neg-ative predictive values were estimated for populations with 5% and 30% prevalence of renalartery stenosis. Kappa values for interobserver agreement were assessed for both gadolinium-enhanced MR angiography and catheter angiography.

RESULTS

.

For detecting renal artery stenosis measuring 50% or greater, the sensitivity ofgadolinium-enhanced MR angiography (96.6%) was greater than that of captopril-enhancedDoppler sonography (69%,

p

= 0.005) and captopril-enhanced renal scintigraphy (41.4%,

p

=0.001). No significant difference in specificity was observed among modalities. For renal arterystenosis measuring 50% or greater, positive and negative predictive values were respectively 62%and 86% for captopril-enhanced Doppler sonography, 49% and 76% for captopril-enhancedrenal scintigraphy, and 53% and 98% for gadolinium-enhanced MR angiography. Interob-server agreement was high for both gadolinium-enhanced MR angiography (

κ

= 0.829) andcatheter angiography (

κ

= 0.729).

CONCLUSION

.

Gadolinium-enhanced MR angiography is the most accurate noninvasivemodality for detecting renal artery stenosis greater than or equal to 50%. The use of captopril-en-hanced Doppler sonography in combination with gadolinium-enhanced MR angiography foridentifying renal artery stenosis needs to be evaluated with a cost-effectiveness analysis.

enal artery stenosis is the leadingcause of curable hypertension. Es-timates suggest that the prevalence

of renovascular disease as a cause of hyperten-sion ranges from 0.5% to 5% in the generalpopulation [1, 2] to as high as 45% in selectedpatients with suggestive clinical features [3].Catheter angiography, which is accepted as thegold standard for the detection of renal arterystenosis, is not an ideal screening method be-cause it is invasive and expensive. Catheter an-giography requires administration of iodinatedcontrast material and exposure to ionizing ra-diation. A reliable noninvasive diagnostic testis needed to select patients for invasive diag-nostic and therapeutic approaches.

During the last decades, several noninvasiveimaging modalities have been evaluated for their

ability to detect renal artery stenosis. Renal scin-tigraphy [4–10] and Doppler sonography [11–15] that show captopril-induced changes provideindirect evidence of the presence of renal arterystenosis and have proven helpful in screeningpatients with this condition. However, data con-cerning the reliability of these techniques are in-consistent and vary among studies. Manyauthors have reported disappointing results forboth techniques [7, 16–22]. More recently, sub-stantial advances have been achieved withgadolinium-enhanced three-dimensional MRangiography for the identification of renal arterystenosis [23–27]. Yet the value of noninvasivemodalities for the detection of renal artery steno-sis has not been sufficiently defined, and themost useful diagnostic strategy remains undeter-mined. Furthermore, the performance of a given

Salah D. Qanadli

1

Gilles Soulez

1

Eric Therasse

2

Viviane Nicolet

1

Sophie Turpin

3

Daniel Froment

4

Maryse Courteau

4

Marie-Claude Guertin

5

Vincent L. Oliva

1

Received August 9, 2000; accepted after revision April 26, 2001.

Supported by operating grant MA15225 of the Medical Research Council of Canada. S. Qanadli was supported by a grant of the Société Française de Radiologie.

1

Department of Radiology, CHUM, Hôpital Notre-Dame, 1560 Sherbrooke St. E., Montréal, Quebec H2L 4M1, Canada. Address correspondence to G. Soulez.

2

Department of Radiology, CHUM, Hôpital Hotel-Dieu de Montréal, University of Montréal, 3840 St. Urbain, H2W 1T8 Montréal, Quebec, Canada.

3

Department of Nuclear Medicine, CHUM, Hôpital Hotel-Dieu de Montréal, University of Montréal, H2W 1T8 Montréal, Quebec, Canada.

4

Department of Medicine, CHUM, Hôpital Notre-Dame, Montréal, Quebec H2L 4M1, Canada.

5

Department of Biostatistics, Hotel-Dieu de Montréal, University of Montréal, Montréal, Quebec H2L 4M1, Canada.

AJR

2001;177:1123–1129

0361–803X/01/1775–1123

© American Roentgen Ray Society

R

1124

AJR:177, November 2001

Qanadli et al.

screening modality can be influenced by theprevalence of true renovascular disease in thepopulation studied. Thus, prospective compari-sons are needed to evaluate the value of eachmodality and to identify a clear diagnostic strat-egy. Our study was designed to compare the ac-curacy of captopril-enhanced renal scintigraphy,captopril-enhanced Doppler sonography and ga-dolinium-enhanced MR angiography for the de-tection of renal artery stenosis in patients withclinically suspected renovascular hypertension,and to determine the predictive value of thesemethods for identifying renal artery stenosisboth in nonselected populations and in popula-tions selected on the basis of clinical features.

Subjects and Methods

Between January 1998 and May 1999, 41 pa-tients (15 men, 26 women; age range, 41–78years; mean, 64 years) were prospectively enrolledin this study. Patient selection was based on thepresence of one or several of the following clinicalfeatures: onset of hypertension before the age of25 years or after 45 years; severe hypertension(malignant hypertension, grade III or IV retinopa-thy, hypertensive encephalopathy, or diastolicblood pressure > 115 mm Hg); refractory hyper-tension (systolic blood pressure > 160 mm Hg ordiastolic blood pressure > 95 mm Hg despite opti-mal doses of three antihypertensive drugs); accel-eration of hypertension by more than 15% withinthe preceding 6 months; or abdominal or flankbruit. During this time, 30 patients were not in-cluded in the study for one or more of the follow-ing reasons: creatinine clearance of less than 40mL/min; hyperkalemia (potassium >

5.5 mmol/L),because of the risk of nephrotoxicity induced byiodinated contrast material; history of stroke ortransient ischemic attack with a carotid bruit, be-cause of the risk of hypotension induced by capto-pril; history of allergy to angiotensin-convertingenzyme inhibitors or iodinated contrast material;or contraindications to MR imaging (e.g., pace-maker, ocular metallic foreign bodies). Thirty-sixadditional patients refused to undergo all examina-tions, sometimes because they had undergone oneor more noninvasive studies with a normal result.The mean arterial systolic over diastolic bloodpressure of the study population was 162 ± 23over 85 ± 12 mm Hg and the mean creatinineclearance level was 107 ± 38 mL/min. Forty of thestudy patients had two kidneys and one had a soli-tary kidney, for a total of 81 kidneys studied.

All patients underwent intrarenal Dopplersonography before and after captopril administra-tion, captopril-enhanced scintigraphy, gadolinium-enhanced MR angiography, and catheter angiographywithin a 3-month period. The sequence of exami-nations depended on the accessibility of the imag-ing modality at the time of imaging. Angiotensin-converting enzyme inhibitors and calcium block-ers were discontinued 2–5 days (depending on the

half-life of the medication) before radionuclideand Doppler sonographic examinations [10]. Nosurgery or endovascular procedure was done be-tween any imaging modalities.

The study was approved by our institutionalethics and research committees, and written in-formed consent was obtained from all patients.

Doppler Sonography

Doppler sonographic examinations were per-formed with a Spectra unit (Diasonics, Milpitas,CA) equipped with a 3.5-MHz phased array trans-ducer. After we identified intrarenal arteries withcolor-flow imaging using a posterior oblique ap-proach, spectral velocity waveforms were obtainedat an angle of insonation of less than 60° from seg-mental arteries at the superior, mid (anterior andposterior), and inferior portions of the kidney. Weused the smallest velocity scale, the lowest wallfilter, and a sweep time of 2 sec. Each patient un-derwent two Doppler sonographic examinationsusing the same technique: the first, baseline exam-ination was followed by a second examination per-formed 1 hr after the oral administration of 25 mgof captopril, in keeping with the recommendationsof the consensus report on angiotensin-convertingenzyme inhibitors for detecting renovascular hy-pertension [10]. We used a pattern recognition ap-proach according to previously published criteria[15]. The most abnormal Doppler spectrum (pro-vided that it was reproducible) was selected by theinvestigator for each kidney before and after theadministration of captopril and was morphologi-cally classified into one of the three types de-scribed by Oliva et al. [15]. Type A represents anormal spectrum with an early systolic peak and asteep linear early systolic rise. Type B includes anormal spectrum without an early systolic peakbut with a steep linear early systolic rise. Type Crepresents abnormal spectrum with a decrease ofthe early systolic rise. A type C Doppler spectrumwas considered indicative of renal artery stenosis.Additional measurements of the resistive index,acceleration, and acceleration time of early sys-tolic rise were obtained for the selected Dopplerspectrum. Acceleration and acceleration timethresholds for positive results were set at 390 cm/sec

2

and 0.06 sec, respectively, for the baseline ex-amination and at 440 cm/sec

2

and 0.09 sec for thecaptopril-enhanced examination [15]. For eachkidney, Doppler findings were considered positiveif either the quantitative (acceleration, time of ac-celeration) or morphologic (pattern recognition)evaluation was abnormal. In cases of disagreementbetween quantitative and qualitative (pattern rec-ognition) evaluations, assessment of renal arterypatency was based on pattern recognition, giventhe high interobserver correlation previously es-tablished with this method (

κ

= 0.95) [15]. DirectDoppler imaging of the proximal renal arteriescould not be performed consistently because oftechnical limitations. Therefore, only intrarenalDoppler sonography was used for this comparativestudy in order to minimize the number of exclu-

sions. The investigator’s level of confidence in thecaptopril-enhanced Doppler sonography interpre-tation was rated on a five-point scale as follows:very high, high, fair, low, and very low.

All examinations were performed and analyzedby one of two investigators who were unaware ofthe findings of the other techniques.

Scintigraphy

Baseline and captopril-enhanced

99m

Tc-mercap-toacetyltriglycine (

99m

Tc-MAG3) scintigraphy wasperformed in all patients using a 1-day, 25-mg capto-pril protocol as recommended by the Working PartyGroup on Determining the Radionuclide of Choice[28].

99m

Tc-MAG3 is a protein-bound radiopharma-ceutical tracer, and its clearance is almost exclusivelythrough tubular secretion.

99m

Tc-MAG3 was pre-ferred to other tracers because of the high extractionefficiencies, its image quality, and its favorable dosim-etry. Patients were instructed to be well hydrated be-fore the examination. Because chronic administrationof angiotensin-converting enzyme inhibitors may re-duce the sensitivity of scintigraphy, this medicationwas withheld from all patients for 2–5 days before theexamination and was replaced by other drugs whenindicated. A large-field-of-view gamma camera inter-faced with a computer was positioned beneath the pa-tient to obtain standard posterior views of the kidneys.Images were stored in a 64

×

64 word-mode pixel ma-trix. Time–activity curves were generated. Data wereobtained for a minimum of 30 min. After the baselinestudy, an oral dose of 25 mg of captopril was adminis-tered, and the patient was instructed to drink 300–500mL of water. The patient was then placed in the su-pine position and blood pressure was monitored atfrequent intervals. The captopril-enhanced study wasinitiated 60 min after captopril administration. Resultswere interpreted according to the guidelines of the So-ciety of Nuclear Medicine [29]. The most importantcriterion for detecting renal artery stenosis was unilat-eral parenchymal retention of the radiopharmaceuticalafter captopril administration. A change in the 20-minto maximum uptake ratio of 0.15 or greater, an in-creased delay of 2 min before maximum uptake, orchanges superior or equal to 2 in the renogram grade(from a 5-level scale) were considered indicative ofrenal artery stenosis. Patients with abnormal baselinefindings indicative of reduced renal function that werenot modified after captopril administration were con-sidered to have an intermediate probability of renal ar-tery stenosis.

All examinations were performed and inter-preted by one investigator who was unaware of thefindings of the other imaging studies.

MR Angiography

MR angiography was performed with 1.5-Tunit (Magnetom Vision; Siemens, Erlangen, Ger-many) using a phased array body coil. Examina-tion of renal arteries consisted of two sequences inthe coronal plane: before and during a dynamic IVadministration of 0.2 mmol/kg of body weight ofgadopentetate dimeglumine (Magnevist; BerlexCanada, Montreal, Canada) to provide background

Sonography, Scintigraphy, and MR Angiography of Renal Artery Stenosis

AJR:177, November 2001

1125

subtraction and to increase vessel-to-backgroundcontrast. Images were acquired with the followingparameters in a single breath-hold: three-dimen-sional gradient-echo technique; TR/TE, 3.4/1.4msec; receiver bandwidth, 890 Hz per pixel; fieldof view, 300

×

300 mm

2

; matrix, 128

×

256; vol-ume coverage, 100 mm; slice thickness, 1.8 mm(after interpolation, the effective thickness was 0.9mm); scanning time, 25 sec. The contrast materialwas administered at a rate of 1.5–2 mL/sec. Thetransit time of the contrast material was deter-mined using a test-bolus sequence of 2 mL and dy-namic region-of-interest analysis of the signalintensity at the level of the renal arteries. Imageacquisition was started 2 sec before the signal in-tensity peak. In all patients, T1- and T2-weightedfast spin-echo sequences were obtained beforegadolinium-enhanced MR angiography to evaluatethe morphologic status of the kidneys, such as theparenchymal volume and signal. Maximum-inten-sity-projection reconstructions and multiplanarreformations were processed after subtraction.

Gadolinium-enhanced MR angiography exami-nations were reviewed by two independent investi-gators without knowledge of the results of any otherexamination. Renal angiograms were graded for im-age quality using a three-point scale: optimal, whena high degree of contrast enhancement was obtainedwithout motion artifacts; suboptimal, when thequality was sufficient for analysis of the main renalarteries but without a high degree of contrast en-hancement; and inconclusive, when poor opacifica-tion or major motion artifacts were observed.Combined analysis of source images, maximum-intensity-projection reconstructions, and multipla-nar reformations was used to analyze renal arteriesand to quantify stenosis. The percentage of stenosiswas calculated using a precision caliper, a magnify-ing lens, and the following formula: (

D

–

d

) /

D

×

100, where

D

is the diameter of the uninvolved seg-ment of renal artery and

d

is the diameter of thestenotic segment. In cases of multiple renal arteries,the most stenotic artery was considered. When morethan one stenosis were identified in a single renal ar-tery, the most severe stenosis was used for analysis.In cases of intravascular signal void, renal arterystenosis was considered greater than 70%.

Catheter Angiography

Catheter angiography was performed on a digi-tal subtraction system (DFP 2000; Toshiba Medi-cal System, Otawara-Shi, Japan) through thefemoral approach in all patients using a 5-Frenchpigtail catheter introduced with the Seldinger tech-nique. A standard posteroanterior projection of theabdominal aorta was obtained in all patients using40 mL of 32% iodinated contrast material (Visi-paque 320; Nycomed Imaging, Ontario, Canada)injected at a rate of 20 mL/sec. Additional projec-tions and selective angiograms were obtained ifnecessary at the discretion of the investigator.

All angiograms were reviewed independentlyby the same two investigators who reviewed gado-linium-enhanced MR angiographic examinations.

A minimum delay of 1 month was observed be-tween gadolinium-enhanced MR angiography andcatheter angiography interpretation sessions. Theidentification of patients was concealed to avoidbias resulting from patient recognition. The cathe-ter angiography interpretation session was doneafter the gadolinium-enhanced MR angiographyinterpretation session. Investigators analyzed im-age quality and measured renal artery stenosis inthe same manner as for gadolinium-enhanced MRangiography. To categorize kidneys and patientswith catheter angiography, two thresholds—50%and 70%—were used to define renal artery steno-sis. Discrepancies among investigators that led tothe classification of renal artery stenosis into dif-ferent categories at catheter angiography were re-solved by consensual interpretation to establish thestandard of reference.

Data Analysis

Kidneys in which at least one examination wasinconclusive were excluded from the comparativeanalysis. According to the findings in each modal-ity, kidneys and patients were classified using atwo-point scale as follows: absence of renal arterystenosis, or presence of renal artery stenosis. Cath-eter angiography was considered the standard ofreference. Indeterminate results with captopril-en-hanced Doppler sonography or captopril-enhancedscintigraphy were considered positive to facilitatestatistical analysis. This attitude is also in keepingwith our usual clinical practice.

The sensitivity and specificity for renal arterystenosis detection were calculated for each tech-nique on the basis of the findings at catheter an-giography using 50% and 70% thresholds for renalartery stenosis. The McNemar test was used tocompare the obtained values. For gadolinium-en-hanced MR angiography, the results of the first in-vestigator were used for the comparative analysis.

The predictive value of each technique for de-tecting renal artery stenosis greater than or equalto 50% was estimated using Bayesian analysis fora nonselected population (with a 5% renal arterystenosis prevalence [30]) and for a population se-lected on the basis of clinical criteria (the preva-lence of renal artery stenosis was set at 30%according to previously published data [1]).

The interobserver variability for interpreting gad-olinium-enhanced MR angiography and catheter an-giography was assessed using the kappa value andintraclass correlation coefficient. On the basis of thekappa value, agreement was defined as follows:poor, less than 0.20; fair

,

0.21–0.40; moderate, 0.41–0.60; good, 0.61–0.800; and excellent, 0.80–1.00. A95% confidence interval (CI) was assigned to thecalculated kappa value. The degrees of stenosis mea-sured with gadolinium-enhanced MR angiographyand with catheter angiography were compared inkidneys with renal artery stenosis greater than orequal to 50% using the Student’s

t

test.Statistical analysis was performed with a statis-

tical software system (SAS for Windows, version6.12; SAS Institute, Cary, NC). Differences were

considered statistically significant when

p

valueswere less than 0.05.

Results

Catheter angiography was considered opti-mal in 98% (investigator 1, 99%; investigator2, 97%) of kidneys and suboptimal in 2% (in-vestigator 1, 1%; investigator 2, 3%). Ninetyrenal arteries, including nine accessory or mul-tiple arteries, were identified in the 81 kidneys.Intrarenal artery Doppler waveforms were ob-tained in all kidneys studied. Doppler exami-nations were scored with a high or very highlevel of confidence in 66 kidneys (81%). Onlytwo examinations (2%) had a low or very lowlevel of confidence. Captopril-enhanced renalscintigraphic examinations were available andwere considered diagnostic in all kidneys.Gadolinium-enhanced MR angiography ex-aminations were considered optimal in 72% ofkidneys (investigator 1, 75%; investigator 2,69%) and suboptimal in 25% (investigator 1,22%; investigator 2, 28%). In one patient, boththe left and the right gadolinium-enhancedMR angiography renal angiograms (2%) wereconsidered inconclusive and were excludedfrom statistical analysis. Therefore, 79 kidneyswere available for the comparative study. Allbut three accessory renal arteries identified atcatheter angiography were visualized on gado-linium-enhanced MR angiography.

Catheter angiography revealed the pres-ence of renal artery stenosis greater than orequal to 50% in 31 (76%) of 41 patients andin 41 (52%) of 79 kidneys. The populationwith renal artery stenosis (mean degree ofstenosis, 68 ± 11%) consisted of 12 men and19 women having a mean age of 65 ± 9 yearsand a mean arterial systolic over diastolicblood pressure of 163 ± 22 over 84 ± 12 mmHg. Twenty-seven patients had atheroscle-rotic lesions, and four patients had fibromus-cular dysplasia. Three kidneys had a totallyoccluded renal artery. The nonstenotic popu-lation (mean degree of stenosis, 20 ± 19)consisted of three men and seven womenhaving a mean age of 60 ± 11 years and amean arterial systolic over diastolic bloodpressure of 160 ± 27 over 87 ± 17 mm Hg.

The mean standard deviation between thetwo investigators for the degree of stenosisdetermined with catheter angiography was9% (range, 0–41%). Concordance betweeninvestigators in quantifying the degree ofstenosis was excellent, with an intraclasscorrelation coefficient of 0.90. When kidneyswere categorized using the two-point ordinalscale with a 50% threshold, disagreement oc-

1126

AJR:177, November 2001

Qanadli et al.

curred in 11 kidneys, with a kappa value of0.729 (range, 0.581–0.877). When thethreshold was set at 70%, agreement re-mained good despite a slight decrease of thekappa value to 0.691 (range, 0.495–0.887).

Agreement between investigators for renalartery stenosis quantification with gadolin-ium-enhanced MR angiography (intraclasscorrelation coefficient = 0.88) was similar tothat observed with catheter angiography (in-traclass correlation coefficient = 0.90). Themean standard deviation between the two in-terpreters of gadolinium-enhanced MR an-giography for the degree of stenosis was

10% (range, 0–74%). Agreement betweeninvestigators measured with kappa coeffi-cients calculated with 95% CIs for identify-ing renal artery stenosis measuring 50% orgreater was excellent for gadolinium-en-hanced MR angiography (

κ

= 0.829 [95%CI, 0.699–0.959]) and good for catheter an-giography (

κ

= 0.0.729 [95% CI, 0.581–0.877]). However, kappa values were slightlylower for the 70% threshold than for the 50%threshold. With the 70% threshold, kappavalues were considered good (

κ

= 0.691[95% CI, 0.495–0.887]) for catheter angiog-raphy and moderate (

κ

= 0.592 [95% CI,

0.385–0.799]) for gadolinium-enhanced MRangiography. Compared with catheter an-giography, gadolinium-enhanced MR an-giography overestimated the degree ofstenosis (Fig. 1). Among 41 kidneys with re-nal artery stenosis greater than or equal to50%, the degree of stenosis observed was78% ± 22% for gadolinium-enhanced MRangiography as compared with 69% ± 14%for catheter angiography (

p

= 0.003).Sensitivity and specificity for detecting re-

nal artery stenosis with thresholds of 50% and70% are reported in Table 1. The sensitivity ofgadolinium-enhanced MR angiography was

C

BA

Fig. 1.—76-year-old man with severe hypertension.Radiologic investigation revealed discrepancy be-tween MR angiography and catheter angiography. A, Catheter angiogram reveals bilateral renal arterystenosis. Right renal artery stenosis (arrow ) was consid-ered meaningful using 50% threshold and insignificantusing 70% threshold (measurement of investigator 1:54%; investigator 2: 64%). Similarly, left renal arterystenosis (arrowhead ) was meaningful for 50% thresholdand insignificant for 70% threshold (investigator 1: 58%;investigator 2: 54%).B, Maximum-intensity-projection reconstruction obtainedfrom gadolinium-enhanced MR angiography shows bilat-eral renal artery stenoses. However, degree of stenosiswas overestimated as compared with catheter angiogra-phy. Right renal artery stenosis (long arrow) was consid-ered meaningful using 70% threshold by secondinvestigator (investigator 1, 63%; investigator 2: 71%) andleft renal artery stenosis (short arrow) was consideredmeaningful using 70% threshold by first investigator (investi-gator 1: 75%; investigator 2: 61%), thus resulting in false-positive results for gadolinium-enhanced MR angiography.C, Intrarenal Doppler waveform obtained from right kid-ney 1 hr after captopril administration. Systolic rise isperfectly straight (arrow ), indicating a normal finding.Left renal Doppler study had equally normal findings.

Sonography, Scintigraphy, and MR Angiography of Renal Artery Stenosis

AJR:177, November 2001

1127

significantly higher than that of captopril-en-hanced renal scintigraphy for both the 50%and 70% thresholds both in the kidney and inthe patient populations (Tables 2 and 3). Thesensitivity of gadolinium-enhanced MR an-giography was superior to that of captopril-en-hanced Doppler sonography, but the differencereached statistical significance only for the50% renal artery stenosis threshold. The sensi-tivity of captopril-enhanced Doppler sonogra-phy was significantly greater than that ofcaptopril-enhanced scintigraphy for both 50%and 70% thresholds for renal artery stenosis. Nosignificant difference of the specificity was ob-served among the modalities, except for gad-olinium-enhanced MR angiography (79.5%)

and captopril-enhanced Doppler sonography(94.9%), for a 50% threshold for renal arterystenosis in the kidney population (

p

= 0.03). Because of a high prevalence of renal artery

stenosis in our study population, positive andnegative predictive values were estimated usingBayesian analysis for a 30% renal artery steno-sis prevalence, as reported in highly selectedpopulations [1]. For detecting renal artery steno-sis greater than 50%, positive and negative pre-dictive values were respectively 62% and 86%for captopril-enhanced Doppler sonography,49% and 76% for captopril-enhanced scintigra-phy, and 53% and 98% for gadolinium-en-hanced MR angiography. In a nonselectedpopulation with a 5% prevalence of renal artery

stenosis, positive predictive values were 16%,10%, and 14% for captopril-enhanced Dopplersonography, captopril-enhanced scintigraphy,and gadolinium-enhanced MR angiography, re-spectively; and negative predictive values were98%, 96%, and 100%, respectively.

Three patients had stenosis greater than50% caused by fibromuscular dysplasia. Inthis group of patients, renal artery stenosiswas accurately diagnosed using scintigraphyin one patient, using Doppler sonography inone patient, and using MR angiography intwo patients.

Discussion

Despite the importance of identifying ren-ovascular disease, a reliable noninvasive tech-nique to detect renal artery stenosis in patientswith hypertension has not been clearly estab-lished. A reliable noninvasive technique is ofparamount importance to select patients withsuspected renal artery stenosis for subsequentrevascularization procedures and also to avoidunnecessary catheter angiography in patientswithout substantial renal artery stenosis. Incon-sistencies are apparent in the literature regardingnoninvasive techniques, especially sonography[16, 17, 19, 31] and radionuclide studies [7, 20,22]. Consequently, we designed our study toprospectively compare the three leading nonin-vasive techniques. Most studies define renal ar-tery stenosis as a reduction in diameter greaterthan 50% based on morphologic evaluation ofthe renal artery [32]. However, some authors ar-gue that only stenoses greater than 70% shouldbe considered hemodynamically significant[32–34]. For this reason, we evaluated the accu-racy of each technique using two renal arterystenosis thresholds, 50% and 70%.

Our results show that gadolinium-enhancedMR angiography has a high sensitivity for de-tection of renal artery stenosis and is probably

Note.—CDS = captopril-enhanced Doppler sonography, CRS = captopril-enhanced renal scintigraphy, MRA = gadolinium-enhanced MR angiography.

TABLE 1 Sensitivity and Specificity of Noninvasive Techniques Calculated for 50% and 70% Thresholds for Renal Artery Stenosis

Variable

Analysis by Kidney Analysis by Patient

Stenosis ≥ 50% Stenosis ≥ 70% Stenosis ≥ 50% Stenosis ≥ 70%

CDS CRS MRA CDS CRS MRA CDS CRS MRA CDS CRS MRA

SensitivityPercent 62.5 32.5 90.0 79.0 47.4 94.7 69.0 41.4 96.6 87.5 56.3 93.8

Number 25/40 13/40 36/40 15/19 9/19 18/19 20/29 12/29 28/29 14/16 9/16 15/16

SpecificityPercent 94.9 92.3 79.5 80.0 88.3 81.7 81.8 81.8 63.6 66.7 79.2 62.5Number 37/39 36/39 31/39 48/60 53/60 49/60 9/11 9/11 7/11 16/24 19/24 15/24

Note.—Data are p values, which were considered significant at less than 0.05 (McNemar test). CDS = captopril-enhancedDoppler sonography, CRS = captopril-enhanced renal scintigraphy, MRA = gadolinium-enhanced MR angiography. Dash (—)indicates not applicable.

TABLE 2 Comparison of Sensitivity and Specificity of Noninvasive Techniques for Detection of Renal Artery Stenosis in 79 Kidneys

Imaging Type

Stenosis ≥ 50% Stenosis ≥ 70%

Sensitivity Specificity Sensitivity Specificity

CDS CRS CDS CRS CDS CRS CDS CRS

CRS 0.001 — 0.564 — 0.034 — 0.096 —MRA 0.001 0.001 0.034 0.132 0.180 0.003 0.763 0.206

Note.—Data are p values, which were considered significant at less than 0.05 (McNemar test). CDS = captopril-enhancedDoppler sonography, CRS = captopril-enhanced renal scintigraphy, MRA = gadolinium-enhanced MR angiography. Dash (—)indicates not applicable.

TABLE 3 Comparison of Sensitivity and Specificity of Noninvasive Techniques for Detection of Renal Artery Stenosis in 40 Patients

Imaging Type

Stenosis ≥ 50% Stenosis ≥ 70%

Sensitivity Specificity Sensitivity Specificity

CDS CRS CDS CRS CDS CRS CDS CRS

CRS 0.005 — 1.000 — 0.025 — 0.180 —MRA 0.005 0.001 0.317 0.414 0.564 0.014 0.739 0.157

1128

AJR:177, November 2001

Qanadli et al.

the most useful noninvasive tool in a high-prev-alence population. We observed that normal re-sults on gadolinium-enhanced MR angiographycould convincingly exclude renal artery stenosisin 98% of cases. However, gadolinium-en-hanced MR angiography has a low positive pre-dictive value (53%) even in a selectedpopulation. This low value could be explainedby overestimation of the degree of stenosis withgadolinium-enhanced MR angiography as com-pared with catheter angiography. In fact, thepositive predictive value of captopril-enhancedDoppler sonography was greater (62%) thanthat of gadolinium-enhanced MR angiography.However, we could not reproduce the diagnos-tic performance of intrarenal captopril-en-hanced Doppler sonography reported in earlierstudies [13, 15]; this discrepancy may be relatedto their design, which consisted of a retrospec-tive review of patients who underwent capto-pril-enhanced Doppler sonography and catheterangiography, without systematic validation withcatheter angiography of negative Dopplersonography results. The sensitivity of captopril-enhanced Doppler sonography in our study(analysis by patients) was fair (69%) for detect-ing renal artery stenosis measuring 50% orgreater and good (87.5%) for detecting 70% re-nal artery stenosis. One limitation of our studyis the absence of evaluation of the main renal ar-tery with color Doppler sonography. The accu-racy of captopril-enhanced Doppler sonographycould be increased by systematically analyzingproximal renal arteries using a direct approach,but this technique is limited by technical failurein 25– 42% of cases [16, 17, 35]. Furthermore,this technique is often inadequate for identify-ing accessory renal arteries, which are presentin approximately 20% of patients [16, 17]. On apositive note, there is hope that the use of sono-graphic contrast agents will improve the evalua-tion of proximal renal arteries [36, 37].

We observed a poor accuracy for captopril-enhanced scintigraphy for detecting moderateand severe renal artery stenosis, with significantlylower sensitivity than that of gadolinium-en-hanced MR angiography and captopril-enhancedDoppler sonography. Our results suggest thatcaptopril-enhanced scintigraphy may not be auseful screening test for renal artery stenosis inpopulations comparable to ours. In the litera-ture, the diagnostic performance of scintigraphyis much higher than that we report, with resultsof 51–96% [8]. We have no clear explanationfor the poor performance of scintigraphy in ourstudy. All patients had unenhanced and capto-pril-enhanced examinations performed after thediscontinuance of angiotensin-converting en-

zyme inhibitors, and only 10 patients (25%) hadbilateral stenoses. The investigators who per-formed these examinations are specialists in renalscintigraphy. A possible explanation for our dis-crepant findings is that most of the results re-ported in the literature are based on retrospectivestudies. Those results can be influenced by a veri-fication bias often found in retrospective studies(only patients with positive findings undergo aconfirmatory examination). Consequently, theproportion of false-negative examinations maynot be evaluated properly, which could lead to anoverestimated sensitivity. In our study, the perfor-mance of scintigraphy was evaluated prospec-tively and appears disappointing.

A criticism of our study is the 75% prevalenceof renal artery stenosis in our patient population.The admitted prevalence of renal artery stenosisafter clinical selection of hypertensive patients is30% [1, 38]. Therefore, our study populationmay not be representative of the target popula-tion, who should meet the selection criteria thatwe used. The prevalence of renal artery stenosisamong all patients referred to our institution withclinically suspected renal artery stenosis duringthe study period was 41%. However, some clini-cians and patients were reluctant to pursue furtherexaminations in the study protocol after normalfindings on Doppler sonography or scintigraphy.In fact, 36 patients in that situation refused to un-dergo angiography, which explains in large partwhy the prevalence of renal artery stenosis wasfurther increased in our study population. An-other point of debate is that patients examined forrenal artery stenosis in our study were selected onsuggestive clinical features, which differs fromthe nonselected population of hypertensive pa-tients, of whom only 2–5% have renovasculardisease [30, 38]. Considering this prevalence dif-ference we estimated the predictive values ofeach screening test in reference to a prevalence asgreat as 30% (clinically selected patients) and aslow as 5% (nonselected patients).

Given the lower availability and higher costof gadolinium-enhanced MR angiography ascompared with captopril-enhanced Dopplersonography, the relative place of these twoscreening tests remains to be determined. Con-sidering the low positive predictive value ofgadolinium-enhanced MR angiography foridentifying renal artery stenosis in nonselectedpatients, the use of this method should be re-served for selected patients. Whether captopril-enhanced Doppler sonography should play arole in the selection remains to be clarified.

In summary, we believe that none of the eval-uated modalities could accurately detect renalartery stenosis in nonselected populations. Our

study shows that gadolinium-enhanced MR an-giography is more accurate than captopril-en-hanced Doppler sonography, which is moreaccurate than captopril-enhanced scintigraphyfor the detection of renal artery stenosis. Gado-linium-enhanced MR angiography is the mostreliable noninvasive screening test for detectingrenal artery stenosis in a population selected onthe basis of clinical criteria. The combination ofcaptopril-enhanced Doppler sonography andgadolinium-enhanced MR angiography forscreening patients with suspected renovasculardisease needs to be evaluated with a cost-effec-tiveness study.

References

1. Working Group on Renovascular Hypertension.Detection, evaluation, and treatment of renovas-cular hypertension: final report.

Arch Intern Med

1987

;147:820–8292. Derkx FH, Schalekamp MA. Renal artery steno-

sis and hypertension.

Lancet

1994

;344:237–2393. Davis BA, Crook JE, Vestal RE, Oates JA. Preva-

lence of renovascular hypertension in patientswith grade III or IV hypertensive retinopathy.

NEngl J Med

1979

;301:1273–12764. Chen CC, Hoffer PB, Vahjen G, et al. Patients at

high risk for renal artery stenosis: a simple methodof renal scintigraphic analysis with Tc-99m DTPAand captopril.

Radiology

1990

;176:365–3705. Dondi M. Captopril renal scintigraphy with 99mTc-

mercaptoacetyltriglycine (99mTc-MAG3) for de-tecting renal artery stenosis.

Am J Hypertens

1991

;4[suppl]:737S–740S 6. Mann SJ, Pickering TG, Sos TA, et al. Captopril

renography in the diagnosis of renal artery stenosis:accuracy and limitations.

Am J Med

1991

;90:30–407. Fommei E, Ghione S, Hilson AJ, et al. Captopril ra-

dionuclide test in renovascular hypertension: a Eu-ropean multicentre study—European MulticentreStudy Group.

Eur J Nucl Med

1993

;20:617–6238. Prigent A. The diagnosis of renovascular hyperten-

sion: the role of captopril renal scintigraphy and re-lated issues.

Eur J Nucl Med

1993

;20:625–6449. Taylor A Jr, Nally JV. Clinical applications of re-

nal scintigraphy.

AJR

1995

;164:31–4110. Taylor A, Nally J, Aurell M, et al. Consensus report

on ACE inhibitor renography for detecting renovas-cular hypertension: Radionuclides in Nephrourol-ogy Group—Consensus Group on ACEIRenography.

J Nucl Med

1996

;37:1876–188211. Stavros AT, Parker SH, Yakes WF, et al. Segmen-

tal stenosis of the renal artery: pattern recognitionof tardus and parvus abnormalities with duplexsonography.

Radiology

1992

;184:487–49212. Schwerk WB, Restrepo IK, Stellwaag M, Klose KJ,

Schade-Brittinger C. Renal artery stenosis: gradingwith image-directed Doppler US evaluation of re-nal resistive index.

Radiology

1994

;190:785–79013. Rene PC, Oliva VL, Bui BT, et al. Renal artery

stenosis: evaluation of Doppler US after inhibi-tion of angiotensin-converting enzyme with cap-topril.

Radiology

1995

;196:675–679

Sonography, Scintigraphy, and MR Angiography of Renal Artery Stenosis

AJR:177, November 2001

1129

14. Miralles M, Cairols M, Cotillas J, Gimenez A, SantisoA. Value of Doppler parameters in the diagnosis of re-nal artery stenosis.

J Vasc Surg

1996

;23:428–43515. Oliva VL, Soulez G, Lesage D, et al. Detection of

renal artery stenosis with Doppler sonography be-fore and after administration of captopril: value ofearly systolic rise.

AJR

1998

;170:169–17516. Berland LL, Koslin DB, Routh WD, Keller FS.

Renal artery stenosis: prospective evaluation ofdiagnosis with color duplex US compared withangiography.

Radiology

1990

;174:421–42317. Desberg AL, Paushter DM, Lammert GK, et al.

Renal artery stenosis: evaluation with color Dop-pler flow imaging.

Radiology

1990

;177:749–75318. Svetkey LP, Wilkinson R Jr, Dunnick NR, et al. Cap-

topril renography in the diagnosis of renovasculardisease.

Am J Hypertens

1991

;4[suppl]:711S–715S19. Postma CT, van Aalen J, de Boo T, Rosenbusch

G, Thien T. Doppler ultrasound scanning in thedetection of renal artery stenosis in hypertensivepatients.

Br J Radiol

1992

;65:857–86020. Roccatello D, Picciotto G, Rabbia C, Pozzato M,

De Filippi PG, Piccoli G. Prospective study oncaptopril renography in hypertensive patients.

AmJ Nephrol

1992

;12:406–41121. Kliewer MA, Tupler RH, Carroll BA, et al. Renal

artery stenosis: analysis of Doppler waveform pa-rameters and tardus-parvus pattern.

Radiology

1993

;189:779–78722. van Jaarsveld BC, Krijnen P, Derkx FH, Oei HY,

Postma CT, Schalekamp MA. The place of renalscintigraphy in the diagnosis of renal artery

stenosis: fifteen years of clinical experience.

ArchIntern Med

1997

;157:1226–123423. Steffens JC, Link J, Grassner J, et al. Contrast-en-

hanced, K-space-centered, breath-hold MR an-giography of the renal arteries and the abdominalaorta.

J Magn Reson Imaging

1997

;7:617–62224. Wilman AH, Riederer SJ, King BF, Debbins JP,

Rossman PJ, Ehman RL. Fluoroscopically triggeredcontrast-enhanced three-dimensional MR angiogra-phy with elliptical centric view order: application tothe renal arteries.

Radiology

1997;205:137–14625. Bakker J, Beek FJ, Beutler JJ, et al. Renal artery

stenosis and accessory renal arteries: accuracy ofdetection and visualization with gadolinium-en-hanced breath-hold MR angiography. Radiology1998;207:497–504

26. Gilfeather M, Yoon HC, Siegelman ES, et al. Re-nal artery stenosis: evaluation with conventionalangiography versus gadolinium-enhanced MRangiography. Radiology 1999;210:367–372

27. Schoenberg SO, Bock M, Knopp MV, et al. Renalarteries: optimization of three-dimensional gadolin-ium-enhanced MR angiography with bolus-timing-independent fast multiphase acquisition in a singlebreath hold. Radiology 1999;211:667–679

28. Blaufox MD, Dubovsky EV, Hilson AJ, Taylor A Jr,de Zeeuw R. Report of the Working Party Group onDetermining the Radionuclide of Choice. Am J Hy-pertens 1991;4[suppl]:747S–748S

29. Taylor AT Jr, Fletcher JW, Nally JV Jr, et al. Pro-cedure guideline for diagnosis of renovascularhypertension: Society of Nuclear Medicine. J

Nucl Med 1998;39:1297–130230. Ram CV, Clagett GP, Radford LR. Renovascular

hypertension. Semin Nephrol 1995;15:152–17431. Kliewer MA, Tupler RH, Hertzberg BS, et al.

Doppler evaluation of renal artery stenosis: inter-observer agreement in the interpretation of wave-form morphology. AJR 1994;162:1371–1376

32. Canzanello VJ, Textor SC. Noninvasive diagnosisof renovascular disease. Mayo Clin Proc 1994;69:1172–1181

33. Levin DC, Beckmann CF, Serur JR. Vascular resis-tance changes distal to progressive arterial stenosis:a critical re-evaluation of the concept of vasodilatorreserve. Invest Radiol 1980;15:120–128

34. Textor SC, Smith-Powell L. Pathophysiology ofrenal failure in ischemic renal disease. In: NovickAC, Hamilton G, eds. Renal vascular disease.London: Saunders, 1996:289–304

35. Helenon O, el Rody F, Correas JM, et al. ColorDoppler US of renovascular disease in native kid-neys. RadioGraphics 1995;15:833–854

36. Melany ML, Grant EG, Duerinckx AJ, Watts TM,Levine BS. Ability of a phase shift US contrastagent to improve imaging of the main renal arter-ies. Radiology 1997;205:147–152

37. Claudon M, Rohban T. Levovist (SHU 508 A) inthe diagnosis of renal artery stenosis: results ofcontrolled multicenter study. (abstr) Radiology1997;205:242

38. Eardley KS, Lipkin GW. Atherosclerotic renal ar-tery stenosis: is it worth diagnosing? J Hum Hy-pertens 1999;13:217–220