Genitourinary Imaging
Original Research
Screening for Transplant Renal Artery Stenosis: Ultrasound-Based Stenosis Probability Stratification
Audio Available | 
Share
OBJECTIVE. The objective of our study was to evaluate which spectral Doppler ultrasound parameters are useful in patients with clinical concern for transplant renal artery stenosis (TRAS) and create mathematically derived prediction models that are based on these parameters.
MATERIALS AND METHODS. The study subjects included 120 patients with clinical signs of renal dysfunction who had undergone ultrasound followed by angiography (either digital subtraction angiography or MR angiography) between January 2005 and December 2015. Five ultrasound variables were evaluated: ratio of highest renal artery velocity to iliac artery velocity, highest renal artery velocity, spectral broadening, resistive indexes, and acceleration time. Angiographic studies were categorized as either showing no stenosis or showing stenosis. Reviewers assessed the ultrasound examinations for TRAS using all five variables, which we refer to as the full model, and using a reduced number of variables, which we refer to as the reduced-variable model; sensitivities and specificities were generated.
RESULTS. Ninety-seven patients had stenosis and 23 had no stenosis. The full model had a sensitivity and specificity of 97% and 91%, respectively. The reduced-variable model excluded the ratio and resistive index variables without affecting sensitivity and specificity. We applied cutoff values to the variables in the reduced-variable model, which we refer to as the simple model. Using these cutoff values, the simple model showed a sensitivity and specificity of 96% and 83%. The simple model was able to categorize patients into four risk categories for TRAS: low, intermediate, high, and very high risk.
CONCLUSION. We propose a simple model that is based on highest renal artery velocity, distal spectral broadening, and acceleration time to classify patients into risk categories for TRAS.
Keywords: kidney, renal artery stenosis, renal graft, transplant
Transplant renal artery stenosis (TRAS) is the most common vascular complication of renal grafts with incidences ranging from 1% to 23%; the wide variation of incidences is likely due to institutional variance in clinical screening protocols, ultrasound parameters, and thresholds for intervention [1–10]. TRAS has a variable and nonspecific clinical presentation: refractory hypertension, fluid retention, increased serum creatinine value, audible bruit, and flash pulmonary edema are the most common symptoms [11].
Ultrasound is the most widely used initial imaging modality in patients with clinical suspicion for TRAS. In combination with clinical presentation, the ultrasound information will dictate the need for additional studies such as MR angiography, CT angiography, or digital subtraction angiography (DSA), depending on institutional resources and preferences [12–14]. To minimize the number of unnecessary additional studies (which can be costly, invasive, and associated with administration of potentially toxic contrast material), investigators have assessed several ultrasound parameters regarding their value in evaluating for TRAS: the peak systolic velocity in the main renal artery, ratio of renal artery velocity to iliac artery velocity, intraparenchymal artery resistive indexes, and intraparenchymal arterial systolic acceleration time [1, 15–22]. However, studies involving these ultrasound parameters have typically focused on a single parameter and have shown different optimal cutoff values [23]. Thus, no consensus has emerged about how to best screen and assess for TRAS using ultrasound. For instance, in a recent meta-analysis on angioplasty and stenting for TRAS, at least 10 separate ultrasound criteria were used [24]. Additionally, the role of spectral broadening, an indicator of vascular flow turbulence that can be seen distal to a stenosis, has not been investigated as a criterion for identifying TRAS.
Therefore, the purpose of this study was to evaluate which spectral Doppler ultrasound parameters are useful in assessing patients with clinical concern for TRAS and create mathematically derived risk assessment models that are based on these parameters to guide the use of additional imaging studies.
| Materials and Methods |   |
|---|
Our institutional review board approved this retrospective HIPAA-compliant study. At our institution, patients with renal allografts are followed postoperatively by dedicated transplant nephrologists. In the setting of clinical features suggestive of TRAS (i.e., increased creatinine value, hypertension, edema, and bruit over the allograft), patients undergo ultrasound of the transplanted kidney. Before 2014, all patients with overall clinical and sonographic concern for TRAS underwent DSA. Since 2014, these patients undergo MR angiography to decrease the number of unnecessary DSA examinations. Patients with TRAS detected at MR angiography typically proceed to DSA. For the purposes of our study, all patients who had a kidney graft and underwent an angiography study (either DSA or contrast-enhanced MR angiography) between January 2005 and December 2015 were included in the study. Patients were excluded if they did not undergo ultrasound of the transplanted kidney at our institution before angiography. If a patient had more than one DSA study, only the first pairing of ultrasound and angiography was included in this study. In cases in which MR angiography was followed by DSA, the DSA was used for angiographic evaluation. A total of 120 patients (mean age, 55 years; 91 men, 29 women) met these criteria, of which 109 underwent DSA and 11 had MR angiography alone. The clinical features that led to the initial ultrasound study were also identified.
Ultrasound was performed using an ultrasound machine (Logiq E9, GE Healthcare; or Acuson Sequoia, Siemens Healthcare) with C6–2, C6–1, and C5–1 curved transducers; a 9L linear transducer; and 5-1–MHz and 4-1–MHz vector transducers. All ultrasound examinations were performed by Registered Diagnostic Medical Sonographer–licensed technologists and were reviewed for image quality at the time of the examination by a board-certified radiologist.
Our protocol for arterial vasculature evaluation of a transplanted kidney includes the following: angle-corrected spectral Doppler imaging of the iliac artery proximal to the anastomosis as well as the transplanted main renal artery in its proximal, mid, and distal portions (with an angle of insonation < 60°) and non–angle-corrected spectral Doppler imaging of the intraparenchymal arteries of the transplanted kidney.
For the purposes of this study, the following parameters were extracted: the ratio of the highest peak systolic velocity of the renal artery (proximal, mid, or distal) to the peak systolic velocity of the iliac artery, highest peak systolic velocity of the renal artery (proximal, mid, or distal), presence or absence of spectral broadening in the segment of vessel just distal to the highest renal artery velocity, resistive index of the intraparenchymal arteries, and longest acceleration time of the intraparenchymal arteries. In cases of multiple transplanted renal arteries, the renal artery with the highest velocity was used. The recorded peak systolic velocities and intraparenchymal arterial resistive indexes were validated by a single radiologist with 3 years' postfellowship experience in abdominal radiology who was blinded to the angiographic findings. The presence of spectral broadening (defined as complete filling of the spectral window) and acceleration time (defined as the time in seconds to reach the peak systolic velocity in the intraparenchymal arteries regardless of the waveform) were assessed independently and then in consensus by two radiologists who were blinded to the angiographic findings and had 3 and 37 years' postfellowship experience in abdominal radiology, respectively, because these parameters may be susceptible to interobserver variability.
All DSA studies were reviewed in consensus by two interventional radiologists with 2 and 5 years' postfellowship experience, respectively, to evaluate for the presence and degree of stenosis and who were blinded to the ultrasound findings. All MR angiograms were reviewed in consensus by two fellowship-trained abdominal radiologists with 3 and 6 years' postfellowship experience on a dedicated 3D workstation (Aquarius iNtuition, version 4.4.8, TeraRecon) using maximum-intensity-projection images and who were also blinded to the ultrasound findings. For both DSA and MR angiography, stenosis was calculated as follows: (1 − [diameter of the stenotic region / diameter of the adjacent normal segment]) × 100%. Patients were grouped into three groups depending on their degree of TRAS: essentially no stenosis (0–19%), mild stenosis (20–49%), and moderate or severe stenosis (≥ 50% stenosis) [25, 26].
MR angiography was performed using ferumoxytol (Feraheme, AMAG Pharmaceuticals) as the contrast agent as previously described [27, 28]. Written informed consent was obtained for the off-label administration of ferumoxytol, and a weight-dependent and diluted dose of ferumoxytol was administered IV. MR angiographic images were acquired on a 1.5-T MRI system (Signa HDxt, GE Healthcare) with an 8-element phased-array torso coil. MR angiography was performed using a 3D T1-weighted spoiled gradient-echo pulse sequence with elliptical centric k-space filling. Axial, sagittal, and coronal high-resolution steady-state MR angiographic series were acquired. The high-resolution MR angiographic plane with the least artifact was chosen for image analysis.
Agreement between the readers was measured with the kappa statistic for the presence of spectral broadening and with the intraclass correlation coefficient (ICC) for acceleration time. Agreement values of 0.01–0.20 were rated as slight; 0.21–0.40, as fair; 0.41–0.60, as moderate; 0.61–0.80, as substantial; and 0.81–1.0, as excellent [29].
Model 1: the full model—A multivariable logistic regression model with generalized logits that included all investigated predictor variables (i.e., ratio of highest renal artery velocity to iliac artery velocity, highest renal artery velocity, presence of spectral broadening, acceleration time, and resistive index) was used to estimate the probability of categorization into the following three categories: no stenosis, mild stenosis, or moderate or severe stenosis. Sensitivity and specificity of the predicted classification were computed. An additional full model with all predictor variables was used to estimate the probability of categorization into two categories: no stenosis or stenosis (mild, moderate, or severe stenosis). The sensitivity and specificity of the predicted classification with respect to having no stenosis versus having any stenosis were then computed.
Model 2: the reduced-variable model—Next, we created a second model by identifying potential predictor variables that could be eliminated from model 1 in a stepwise fashion without affecting model 1's sensitivity and specificity for predicting no stenosis or any stenosis (mild, moderate, or severe stenosis).
Model 3: the simple model—Finally, to develop a more clinically applicable TRAS risk classification, we probed the predictor variables that had remained in model 2 for optimal cutoff levels by assessing sensitivity and specificity for each continuous variable at different cutoff levels using ROC curves. We thereby created a third model, entering the predictor variables categorized according to their previously identified cutoff levels to develop a simple classification rule to predict no stenosis versus any stenosis. The sensitivity and specificity with respect to having no stenosis versus having any stenosis (mild, moderate, or severe stenosis) were computed for this classification.
All statistical analysis was performed with statistics software (SAS, version 9.3, SAS Software). Statistical significance was assessed at the 0.05 level (two-sided).
| Results | |
|---|
We obtained data for 120 patients: 71 with moderate or severe stenosis, 26 with mild stenosis, and 23 with no stenosis (Figs. 1–4). Of the subgroup of 11 patients who underwent MR angiography alone, one had moderate or severe stenosis, one had mild stenosis, and nine had no stenosis (Fig. 2). Ultrasound factors for TRAS and the clinical findings that elicited the initial ultrasound examination are listed in Table 1.
 View larger version (131K) | Fig. 1A —Low risk category: 69-year-old female kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function and hypertension. A, Ultrasound image of origin of transplanted renal artery shows normal velocity (177 cm/s). PSV = peak systolic velocity, EDV = end diastolic velocity, RI = resistive index. |
 View larger version (119K) | Fig. 1B —Low risk category: 69-year-old female kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function and hypertension. B, Spectral Doppler image obtained more distal in renal artery than A shows that there is no spectral broadening. |
 View larger version (107K) | Fig. 1C —Low risk category: 69-year-old female kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function and hypertension. C, Ultrasound image of intraparenchymal artery shows high resistive index with normal acceleration time of 0.05 second. |
 View larger version (261K) | Fig. 1D —Low risk category: 69-year-old female kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function and hypertension. D, Digital subtraction angiogram obtained after ultrasound shows widely patent transplanted renal artery. |
 View larger version (183K) | Fig. 2A —Intermediate risk category: 57-year-old male kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function. A, Ultrasound image of proximal renal artery shows velocity of 235 cm/s, which did not meet criterion of being elevated. PS = peak systole, ED = end diastole, RI = resistive index. |
 View larger version (214K) | Fig. 2B —Intermediate risk category: 57-year-old male kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function. B, Spectral Doppler image obtained more distal in renal artery than A shows that there is some clarity in spectral window, which did not meet criterion for spectral broadening. |
 View larger version (159K) | Fig. 2C —Intermediate risk category: 57-year-old male kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function. C, Ultrasound image of intraparenchymal artery shows resistive index of 0.59 with prolonged acceleration time of 0.15 second. |
 View larger version (336K) | Fig. 2D —Intermediate risk category: 57-year-old male kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function. D, Curved planar reformatted MR angiogram obtained after ultrasound shows widely patent renal artery (arrow) emanating from external iliac artery. |
 View larger version (177K) | Fig. 3A —High risk category: 25-year-old female kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function. A, Ultrasound image of transplanted renal artery failed to show elevated velocity. PS = peak systole, ED = end diastole, RI = resistive index. |
 View larger version (271K) | Fig. 3B —High risk category: 25-year-old female kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function. B, Spectral Doppler image obtained more distal in transplanted renal artery than A shows that there is complete filling of spectral window, which indicates spectral broadening. |
 View larger version (182K) | Fig. 3C —High risk category: 25-year-old female kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function. C, Ultrasound image of intraparenchymal artery shows low resistive index (0.52) and prolonged acceleration time of 0.16 second. |
 View larger version (384K) | Fig. 3D —High risk category: 25-year-old female kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function. D, Angiogram obtained after ultrasound shows stenosis (> 50%) involving proximal to midportion of transplanted renal artery. |
 View larger version (180K) | Fig. 4A —Very high risk category: 52-year-old male kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function and edema. A, Ultrasound image of midportion of transplanted renal artery shows increased velocity (429 cm/s). PS = peak systole, ED = end diastole, RI = resistive index. |
 View larger version (326K) | Fig. 4B —Very high risk category: 52-year-old male kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function and edema. B, Spectral Doppler image obtained more distal in renal artery shows that there is complete filling of spectral window, which indicates spectral broadening. |
 View larger version (191K) | Fig. 4C —Very high risk category: 52-year-old male kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function and edema. C, Ultrasound image of intraparenchymal artery shows low resistive index (0.46) and prolonged acceleration time of 0.18 second. AT = acceleration time. |
 View larger version (400K) | Fig. 4D —Very high risk category: 52-year-old male kidney transplant recipient who underwent ultrasound of renal allograft secondary to worsening renal function and edema. D, Digital subtraction angiogram obtained after ultrasound shows stenosis (> 50%) just distal to origin of transplanted renal artery. |
Agreement between readers was excellent for the presence of spectral broadening (κ = 0.83) and substantial for acceleration time (ICC = 0.61).
Factors most associated with stenosis included the ratio of the highest renal artery velocity to iliac artery velocity, highest renal artery velocity, the presence of spectral broadening, and acceleration time. The model produced accurate classifications of patients with no stenosis (91% correct) and patients with moderate or severe stenosis (93%); however, most (77%) of the patients with mild stenosis were misclassified as having moderate or severe stenosis (Table 2). The full model had very high sensitivity (97%) and specificity (91%) with respect to the classification of no stenosis versus any stenosis (mild, moderate, or severe stenosis) (Table 3).
After elimination of the resistive index and the ratio of the highest renal artery velocity to iliac artery velocity, the reduced-variable model retained high sensitivity (96%) and specificity (91%) with respect to classification of no stenosis or any stenosis (mild, moderate, or severe stenosis) (Table 4).
The optimal cut point for the highest renal artery velocity was 300 cm/s and for acceleration time, 0.1 second. The sensitivity and specificity using these cut points were 84% and 83%, respectively, for highest renal artery velocity, 91% and 78% for acceleration time, and 90% and 83% for spectral broadening. For the simple model, the sensitivity of the predictive classification for no stenosis versus any stenosis remained very high (96%) with a high specificity (83%). The prediction model is depicted in Table 5. According to this model, four risk categories were created: If none of the three factors was positive using the optimal cut points, there was a 5% chance of a stenosis being present; if one of the factors was positive, a 32–46% chance of a stenosis; if two of the factors were positive, an 84–92% chance of a stenosis; and if three of the factors were positive, a 99% chance of a stenosis.
| Discussion | |
|---|
In our study, we created several sequential models that were based on ultrasound factors that predicted the presence or absence of stenosis in a graft's renal arteries. The initial model, the full model, was based on five characteristics: the ratio of the highest renal artery velocity to iliac artery velocity, highest renal artery velocity, presence of spectral broadening within the transplanted renal artery, resistive indexes of the intraparenchymal arteries, and acceleration time of the intraparenchymal arteries. Our reduced-variable model—which excluded the ratio of the highest renal artery velocity to iliac artery velocity and the resistive indexes—had similar sensitivity and specificity in detecting TRAS. Finally, we created a clinically more useful and applicable model by introducing binary cutoff values for maximal renal artery velocity and acceleration times (300 cm/s and 0.1 second, respectively) that very closely replicated the accuracy of the two preceding models. Therefore, we propose our final model for clinical TRAS screening in patients with clinically concerning features given its high sensitivity and specificity coupled with its ease of use.
The three ultrasound factors in the final model (highest renal artery velocity, presence of spectral broadening, and acceleration time) accurately stratified kidney recipients into four risk categories: low risk when none of the three factors is present, intermediate risk when one of three factors is present, high risk when two of three factors are present, and very high risk for TRAS when three of three factors are present. Therefore, even if one factor (elevated renal artery velocity, spectral broadening, or prolonged acceleration time) is absent, there remains a strong possibility (84–92%) of TRAS.
Although some studies have proposed the ratio of the highest renal artery velocity to iliac artery velocity as a means to assess for TRAS, we found no advantage in using this value over the highest renal artery velocity. Additionally, because the ratio requires calculation and requires the measurement of two velocities, thus rendering it even more prone to error, we used only the highest renal artery velocity. In assessing for TRAS, various cutoff velocities in the graft's renal artery, ranging from 180 to 400 cm/s, have been proposed [3, 16, 19–22]. In our study population, we found a cutoff of 300 cm/s to be most effective. Although a lower cutoff value might be more sensitive in detecting TRAS, it would also lead to a decrease in specificity—and thus the possibility of unwarranted additional examinations, including invasive angiography. A study by Robinson et al. [23] highlights the importance of not relying solely on an elevated renal artery velocity (which they define as > 250 cm/s) because 26% of recipients with a normal graft at 9 months and 18% of recipients with a normal graft at 1 year showed these elevated renal artery velocities. In their study, a large number of false-positive ultrasound findings would have been identified if a ratio cutoff of 1.8 had been used [23].
In our study, there was a trend of a decreased average resistive index with an increased degree of stenosis. However, we found that this ultrasound factor was not helpful in discriminating between the presence or absence of TRAS. There are conflicting views about the value of intraparenchymal resistive indexes in TRAS. Ardalan et al. [15] found that a resistive index of less than 0.55 can be used to screen for TRAS. However, de Morais et al. [21] reported no significant difference between groups that had versus did not have TRAS. In native kidneys, renal artery stenosis can be associated with variable intraparenchymal resistive indexes, ranging from low to high. In the setting of a native renal artery stenosis, the higher the resistive index, the more parenchymal scarring is likely present and the less likely it is to respond to endovascular intervention. Therefore, although the intraparenchymal resistive index may not be useful in assessing for the presence of TRAS, it may have implications with regard to treatment response [30, 31].
Spectral broadening has not to our knowledge been investigated as a factor in determining the presence or absence of TRAS. Spectral broadening, or filling in of the spectral window, is present when the likelihood of different velocities under the spectral window is equal, indicating turbulent blood flow in medium-sized and large vessels. This is seen distal to areas of stenosis and at bifurcations [32]. We found this factor to be useful in assessing for the presence of TRAS. It is not rare that the entire transplanted renal artery cannot be visualized on ultrasound, owing to technical constraints and obscuring bowel gas or to the tortuosity of the transplanted renal artery, such as seen in Figure 3. Therefore, finding the area of stenosis that may depict increased velocity can be challenging in some cases. However, ultrasound's ability to depict more distal spectral broadening can help increase its sensitivity in detecting TRAS. Although there are degrees of spectral broadening, we used the binary definition of complete versus incomplete filling of the spectral window. Spectral broadening can be affected by sample volume, Doppler angle, and gain settings; however, in spite of these limitations, we found spectral broadening to be a useful factor in detecting TRAS.
Other studies have advocated using an acceleration time cutoff of 0.1 second [19, 21]. In a study by Gottlieb et al. [19], the addition of an acceleration time of 0.1 second or greater in patients with transplanted renal artery velocity greater than 200 cm/s provided an accuracy of 95% in detecting TRAS, compared with 62% when using the elevated transplanted renal artery velocity (> 200 cm/s) alone. Our results are in concert with these findings, not only in the cutoff value of 0.1 second but also in showing its additive value when combined with other ultrasound factors. We should note that our definition of acceleration time does not take into account the shape of the waveform, as has been proposed by other studies [33, 34]. In our practice, we have found considerable variability in interpretation of acceleration time using this other definition, as has been noted by other authors [35]. Using this more straightforward definition of acceleration time, we obtained substantial agreement between the two readers.
Our final model has several important clinical implications and features. First, it provides a framework for establishing uniformity in ultrasound assessment for TRAS. Second, it is simple and clinically readily and easily useable, because it does not require computation of a value using a complex mathematic formula. Third, it effectively classifies patients into four discrete risk categories: low, intermediate, high, and very high risk. This classification is extremely useful for the referring clinician to whom it provides reliable guidance and decision support when deciding whether additional—more invasive—studies may be warranted. Finally, the corollary finding in our initial full model that ultrasound cannot reliably distinguish between mild stenosis and moderate or severe stenosis may point to a role for less invasive imaging, such as MR angiography, as a next step to assess whether invasive angiography is warranted. This distinction is important because patients undergoing DSA who do not require angioplasty and stenting are at increased risk of developing acute kidney injury [36]. Although it would be optimal if ultrasound could distinguish among the different grades of stenoses, its ability to effectively screen patients who have no stenosis from those with stenosis using our algorithm could lead to a decrease in subsequent imaging examinations, including invasive DSA.
Our study has several limitations. First, our models are based on retrospective data. It would be useful to validate our models retrospectively with data from other institutions and prospectively. Second, our reference standard included both DSA and MR angiography studies. However, we believe that this limitation is mitigated by studies that have shown that MR angiography very closely approximates DSA in assessment for stenosis [12, 37]. Additionally, the examinations were performed by multiple sonographers using various transducers. However, the adherence to standard imaging protocols should have minimized differences. Finally, all our initial ultrasound studies involved selected patients—that is, patients with clinical evidence for renal dysfunction as the reason for their examination and overall clinical and sonographic evidence for stenosis. Therefore, our results are not necessarily applicable to other scenarios, such as assessments during the immediate postoperative period or of clinically asymptomatic patients.
In summary, we have shown that the most important factors in assessing for TRAS are highest transplant renal artery velocity (≥ 300 cm/s), presence or absence of spectral broadening in the transplanted renal artery, and intraparenchymal arterial acceleration time (≥ 0.1 second). We propose an easy-to-apply, clinically relevant model based on elevated maximum transplant renal artery velocity, the presence of distal spectral broadening, and prolonged acceleration time to classify patients into four risk categories for TRAS: low, intermediate, high, and very high risk.
| References | |
|---|
