Abstract
OBJECTIVE:
We performed a prospective study to evaluate the value of contrast-enhanced ultrasound (CEUS) in quantitative evaluation of renal cortex perfusion in patients suspected of early diabetic nephropathies (DN), with the estimated GFR (MDRD equation) as the gold standard.
METHODS:
The study protocol was approved by the hospital review board; each patient gave written informed consent. Our study included 46 cases (21 males and 25 females, mean age 55.6 ± 4.14 years) of clinical confirmed early DN patients. After intravenous bolus injection of 1 ml sulfur hexafluoride microbubbles of ultrasound contrast agent, real time CEUS of renal cortex was performed successively using a 2–5 MHz convex probe. Time-intensity curves (TICs) and quantitative indexes were created with Qlab software. Receiver operating characteristic (ROC) curves were used to predict the diagnostic criteria of CEUS quantitative indexes, and their diagnostic efficiencies were compared with resistance index (RI) and peak systolic velocity (PSV) of renal segmental arteries by chi square test. Our control group included forty-five healthy volunteers. Difference was considered statistically significant with P < 0.05.
RESULTS:
Changes of area under curve (AUC), derived peak intensity (DPI) were statistically significant (P < 0.05). DPI less than 12 and AUC greater than 1400 had high utility in DN, with 71.7% and 67.3% sensitivity, 77.8% and 80.0% specificity. These results were significantly better than those obtained with RI and PSV which had no significant difference in early stage of DN (P > 0.05).
CONCLUSIONS:
CEUS might be helpful to improve early diagnosis of DN by quantitative analyses. AUC and DPI might be valuable quantitative indexes.
Keywords
Introduction
Diabetic nephropathy (DN) is by far the most common cause of end-stage renal disease (ESRD) in many countries and a main cause of diabetes-related morbidity and mortality worldwide [18]. Various hemodynamic factors contributed to the development of DN, including increased systemic and intra-glomerular pressure, as well as glomerular hyperfiltration [28]. Following the early stage of hyperfiltration and increased glomerular filtration rate (GFR), there was usually a time of clinical latency which may last up to 20 years, with subsequent decline in renal function [11]. Previously, researches had revealed that the early stage of DN was reversible. Patient who did not develop overt albuminuria (over 200 mg/min) could revert to normal kidney function if treated appropriately [17]. Therefore, early diagnosis and clinical follow-up evaluation of DN are very important.
At present, renal biopsy is the gold standard for diagnosing and staging of DN. However, it is an invasive procedure associated with many potential complications and is not suitable for routine clinical follow-up [2]. Clinically, increase in serum creatinine (SCr) was not sufficient to evaluate DN since SCr may be in the normal range despite decreased GFR [5]. As the assessment of tissue perfusion yields information about renal function, it is important to identify the early perfusion changes of DN noninvasively by noninvasive imaging methods. Over the past decade, application of diffusion-weighted (DW) functional magnetic resonance (MR) imaging quantified by the apparent diffusion coefficient (ADC) provided information on diffusion and perfusion simultaneously [32]. However, their clinical applications were restricted by several factors such as time-consuming, expensive or required complex measurements [9].
Ultrasound is an imaging modality with advantages including low cost, portability, availability, and absence of exposure to radiation or nuclear tracers [1]. Color Doppler flow imaging (CDFI) was considered as a possible imaging technique in detecting renal blood perfusion abnormalities. Unfortunately, it was limited in the detection of low-velocity flow, and flow in smaller (<2 mm in diameter) or deep vessels [1]. Contrast-enhanced ultrasound (CEUS) has recently been proposed as a new contrast-specific imaging modality to quantify renal perfusion [4, 19]. Because ultrasound microbubbles are blood-pool agents with no nephrotoxicity, they were distributed to the entire macro- and microvascular system without extravasation. CEUS represented a perfect tool for assessing renal parenchymal perfusion in real time [3, 33]. For more than 90% of total renal blood flow entering the renal cortex, we hypothesized that measurements of their renal tissue kinetics could be used to quantify renal perfusion changes in the early stage of DN.
In this study, we prospectively explored the value of CEUS in diagnosis of early perfusion changes in DN patients. We also evaluated whether CEUS quantitative methods can improve the predictive capability for DN.
Materials and methods
Patients
This study was performed with the approval of Zhongshan Hospital (approval number: zs-20130518). All patients gave their full informed consents to participate in our study. They were examined with the standardized procedures of CEUS.
The study enrolled 46 DN patients (21 females and 25 males, CKD stage II∼III) of overall mean age of 55.6 ± 4.14 years. The inclusion criteria for a suspicion of early DN were: (1) Patient had a type 2 diabetic history for 5∼10 years. (2) Patient had evidence of kidney damage for more than 3 months (eg, hypertension, proteinuria, elevated blood serum urea nitrogen levels (BUN >7.0 mmol/L,) or serum creatinine levels (SCr >130 μ mol/L). (3) Body mass index (BMI) of patients were between 18.5∼26.9. (4) Clinically, the included patients had kidney damage with decrease in GFR (30∼60 ml/min/1.73 m2) estimated by abbreviated MDRD equation. According to the Kidney Disease Outcome Quality Initiatives (K/DOQI) clinical practice guidelines for chronic kidney disease, the included patients were classified as early DN (CKD, stage II∼III) [21].
The exclusion criteria were: (1) Known history of renal artery stenosis defined as more than 50% reduction in diameter of renal artery by previous CT angiography (CTA) results. (2) Evidence of hydronephrosis on gray scale ultrasound. (3) Contrast agent allergy. (4) Severe heart or pulmonary disease. (5) Pregnancy. (6) Patients who could not hold their breath during CEUS procedure.
Forty-five sex and age matched healthy volunteers (21 males, 24 females) were enrolled as control groups, mean age, 55.6 years ± 3.14. Those healthy adults had no history of diabetes or nephropathy, with normal renal and cardiac function and not undergoing any pharmacologic treatment. They had normal renal sizes and shapes on gray scale ultrasound, normal BUN (BUN <7.0 mmol/L) and Scr levels (Scr <130 μ mol/L). In accordance with institutional review board-approved guidelines, all healthy volunteers received CEUS examinations and signed the consent forms prior to enrollment.
Protocol
All ultrasound examinations were performed with Philips iU22 unit (Philips Bothell, WA, USA). A C5-2 broadband curved array transducer with 2–5 MHz convex probe (Philips Bothell, WA, USA) was used. All ultrasound examinations and quantification analysis were performed by one experienced radiologist with 15 years’ experience of diagnostic ultrasound of kidney. SonoVue® (Bracco Imaging Spa, Milan, Italy), a second-generation contrast agent consisting of a stabilized suspension of sulfur hexafluoride microbubbles with a phospholipidic shell, approved for radiological clinical use in China, was used in all patients.
First, all patients underwent gray scale ultrasound to measure the size (longitudinally and cross-sectionally) of both two kidneys. We tried to choose a maximum longitudinal scanning plane for CEUS that included the entire kidney if possible. Second, spectral Doppler was used to assess the renal blood flow, RI (resistance index) and PSV (peak systolic velocity, cm/s) of renal interlobular arteries were calculated with automatic spectral wave form tracing method. Third, after ultrasound contrast agent was administered by bolus injection into the antecubital vein, we performed CEUS. First for the right kidney, after a time interval for about 20 min, the same CEUS procedure was performed for left kidney [7]. Renal perfusion images obtained with CEUS were stored digitally and analyzed offline, using a commercially available software tool (Qlab, Release 4.1, Philips, Bothell, WA).
CEUS Examination
Image acquisition
CEUS was performed using contrast harmonic imaging at a low MI <0.1. To each patient, SonoVue® was administered as a 1 ml bolus through an antecubital vein, immediately followed by 10 ml saline solution (0.9% NaCl) over 4-5 sec. Continuous scanning started immediately after the contrast injection, and examination of renal cortex perfusion was evaluated in real time and lasting 2 to 6 minutes. Imaging of renal perfusion was obtained in a maximum longitudinal scanning plane including the entire kidney if possible. During the CEUS examination, acoustic power was set at low level to minimize breaking microbubbles, with the focal zone set at the renal cortex in the deep part of the kidney in ultrasound image. Throughout the examination, the transducer was kept in the same scanning plane. Digital cine-clips were stored on the hard disk as DICOM (Digital Image Communications in Medicine) images. Patients were asked to hold their breath as long as possible and breathe gently and regularly afterward.
Quantification procedure
Digital cine-loops were transferred to a PC system for off-line quantification using Qlab software. A region of interest (ROI) was drawn over the mid superficial peripheral renal cortex and excluded the medulla.
As the signal intensity-related parameters were depth dependent [7], the investigator ensured selection of the same square and same sized ROI (5×5 mm square), ROIs were kept in a similar depth on the kidney cortex for each subject, while avoiding inclusion of the interlobar and arcuate arteries. For quantitative analysis of renal tissue perfusion, signal intensity in the ROI was measured and TICs was automatically generated. If the ROI deviated from the original position because of respiratory movements, we used the “breath compensation” function in Qlab software to identify and correct placements of the ROIs. For each kidney, analyses were repeated three times. We took the average of three ROI observations in order to minimize the transitional distance caused by respiration and for accuracy of analyses.
After SonoVue® was administered as a bolus, a smooth TIC fitted with a Gaussian curve was created. A Gamma-variate function: I(t) = A*t*exp(-αt) + C was the suitable curve fit approximation according to the kinetic model used [7, 31]. Series of quantitative parameters were automatically calculated. I(t) represented the pixel intensity as a function of time, the slope rate of ascending curve (A) was the scaling factor related to the wash-in of TICs, the slope rate of descending curve (α) was a rate constant reflected to the width of TICs. Area under curve (AUC) was derived from this equation as A/α2. Derived peak intensity (DPI) was calculated as (A/ α* exp1). Time to peak (TTP) was calculated as 1/α. The asymptotic value of C represented the baseline intensity.
Conventional CDFI parameters
The same C5-2 broadband curved array transducer with 2 to 5 MHz extended operating frequency range (Philips Bothell, WA, USA) was used to assess the RI and PSV by spectral Doppler measurements. The Doppler sample volume was positioned with Doppler angle kept between 30–60 degree. The PRF was set to avoid aliasing, the wall filter was optimized at low level to detect slow diastolic flow. Intra-renal Doppler signals were obtained from three representative interlobular arteries in the upper, middle and lower pole of renal cortex. The median value of PSV and RI of each intra-renal artery was assessed by using three different measurements performed by a single investigator.
Analyses and statistics
CEUS data analyses
Data are expressed as mean ± standard deviation. Before performing any statistical analyses between groups, data were checked for normal distribution using the Kolmogorov-Smirnov Z-test.
The two-samples Student’s t-test was applied for the comparison between DN group and control groups, also between two kidneys. All statistical analyses were performed with SPSS 15.0 software package (SPSS, version 15.0 Inc. Chicago, IL, USA). A difference was considered statistically significant with P < 0.05.
Receiver operating characteristic (ROC) analysis
For further analyses of the predictive efficiency of ultrasound quantitative indexes, we performed ROC analysis as a predictor of DN (n = 46) compared with the normal groups (n = 45).
The diagnostic performance including the sensitivities, specificities and overall accuracies of CEUS and CDFI were compared by Chi-square tests, using the estimated GFR (MDRD equation) as the ‘gold standard’. A difference was considered statistically significant with P < 0.05.
Results
Conventional CDFI parameters
Using unenhanced ultrasound scanning, the gray scale appearance of both medulla and cortex were similar in DN patients and the control groups.
With spectral Doppler measurements, statistical analyses revealed a non-significant difference in RI and PSV between normal control groups (n = 45) and DN patients (n = 46) (P > 0.05) (Table 1).
Real-time perfusion of renal cortex in DN patients
CEUS showed distinct real-time perfusion of renal cortex in both DN patients and control groups. From the longitudinal scan plane of ultrasound, we visualized rapid contrast enhancement of renal cortex due to high renal blood flow. From segmental renal arteries, interlobular arteries, arcuate arteries to small interlobular arteries, and immediately followed by enhancement of the renal cortex. Then the renal pyramids were gradually filled in with contrast agents and became isoechoic with the cortex. Then, the renal enhancing effect decreased as the contrast concentration decreased. From visual observation, no difference had been observed in the perfusion process of renal cortex between DN patients and the control groups (Fig. 1).
No adverse effect of sonographic contrast material was noted, and there was no hematuria or local pain.
Quantitative perfusion data of renal cortex
TICs in the renal cortex were acquired in all 46 DN patients and 45 healthy volunteers. After analysis with Qlab software, quantitative perfusion indexes of CEUS were effectively obtained in all patients and were selected for final data analysis.
In both groups, TICs of renal cortex perfusion were asymmetrical curves. TICs included three parts: steep ascending slope, peak and flat descending slope. Compared with normal control groups, renal cortex perfusion TICs of DN patients were characterized by a delayed and decreased enhancement in the renal cortex (Fig. 2).
Comparing with normal control groups, quantitative indexes of DN patients were characterized by increased AUC and TTP, decreased DPI. The slope rates of both ascending and descending curve had changed (A gradually increased, and α decreased). It took more time to reach the peak intensity (TTP increased). Among all the quantitative indexes in both kidneys of DN patients, AUC and DPI were significantly changed (P < 0.05) (Table 2). The interobserver agreement with kappa value was 0.647 (P < 0.05).
ROC analysis
We used ROC curve analysis to evaluate the diagnostic efficiency of CEUS quantitative indexes in DN patients. Our results indicated that diagnostic accuracies of DPI and AUC were better than 0.7. As DPI had larger area under curve in ROC analysis, we considered this quantitative index had the better diagnostic accuracy (Fig. 3).
Applied 12 dB as the cut-off value (>12 dB no disease, <12 dB disease), DPI measurements in the 46 patients yielded 33 true positives, 35 true negatives, 10 false positives and 13 false negatives (sensitivity 71.7%, specificity 77.8%, positive predictive value 76.7%, negative predictive value 72.9%).
Applied 1400 dB·sec as the cut-off value (<1400 dB·sec no disease, >1400 dB·sec disease), AUC measurements in the 46 patients yielded 31 true positives, 36 true negatives, 9 false positives and 15 false negatives (sensitivity 67.3%, specificity 80.0%, positive predictive value 77.5%, negative predictive value 70.5%) (Table 3).
Discussion
DN is likely to be more susceptible to intervention at an early stage. Traditionally, CDFI with spectral Doppler measurements of renal mean artery or segmental arteries is a noninvasive method for the investigation of renal hemodynamics. Increased RI might reflect a renal scarring process as results of intrarenal vessel area reduction and intrarenal vascular resistance increase [25]. However RI was nonspecific and might be influenced by various factors, such as increased intra-abdominal pressure, pulse rate, pharmacotherapy, and the site at which it was measured. It depended on the examiner and was limited in obese individuals [22, 27]. Also, RI values only correlated with macroangiopathy, it could not interrogate the microvasculature because bulk tissue movement was faster than capillary flow [22]. In our research, no differences were observed in RI or PSV of early DN patients (P > 0.05). Further Chi-square tests indicated that the diagnostic efficiencies of CEUS quantitative indexes were much better than RI and PSV. Our results provided the evidence that quantitative CEUS evaluation of renal cortex perfusion in DN patients might be superior to previous CDFI measurements, including RI and PSV values.
With the introduction of second generation ultrasound contrast agents, CEUS enabled dynamic assessment and quantification of microvascularisation up to capillary perfusion [7, 31]. Previously researchs concluded that CEUS with perfusion-analysis may be a useful predictor of graft function in the early post-transplantation period [4]. Since the nonlinear signals from microbubbles occur regardless of their motion and equally when they are stationary, CEUS detected the capillary bed, which reflected the largest part of the microcirculation [10]. Our results showed that CEUS provided real-time perfusion imaging of the renal cortex vascularization, even in the deep pole areas. CEUS provided rapid and continuous enhancement both in DN patients and healthy volunteers.
With quantitative analysis software, CEUS opened up new possibilities in the characterization of the microvasculature, it allowed precise quantification of the parameters in a certain region of interests objectively and accurately [4, 10]. More importantly, the contrast agents used in CEUS are not nephrotoxic and are excreted mainly via the respiratory tract [3]. The use of wash-in and wash-out curves of contrast agents are proposed as a more functional and distinct analysis of this issue [26]. Previously, some researchers concluded that CEUS was a valuable tool for the assessment of the tissue perfusion in native and transplanted kidneys by providing information on perfusion deficits of the parenchyma [30]. Patients with acute rejection have delayed signal increase in the transplant cortex [14]. In diabetic patients, the most important morphological features included tubular atrophy, glomerulosclerosis, arteriosclerosis, arteriolar hyalinosis and interstitial fibrosis [24]. A restricted blood fluidity can contribute to a limited perfusion especially in diabetic microangiopathy [15, 20]. Microvascular functional and structure damage were of vital importance in the progression of DN [8]. In the early stages, glomerular filtration and renal volume are increased, renal cortex increased the resistance to its perfusion [6, 20]. As a result of renal perfusion reduction, fewer contrast microbubbles entered the renal cortex. Then, renal auto-regulation mechanism was activated to keep the balance of perfusion. In our current study, shapes of the TICs observed in the renal cortex reflected a Gaussian curve [12, 23]. Our results showed that comparing with normal control groups, quantitative indexes of DN patients were characterized by increased AUC, A and TTP, decreased α and DPI. The slope rates of both ascending and descending curve had changed; it took more time to reach the peak intensity. Those results indicated that compared with normal groups, less contrast microbubbles entered the renal cortex microvascular bed with slow perfusion in unit time.
In TICs, the total amount of tracer passed through the ROI, the corresponding AUC was derived from this equation as A/α2, which reflected the relative blood volume in the tissue. DPI was calculated as (A/ α* exp1), which represented the peak intensity of perfusion [12]. As noticeably decreased microbubbles entering renal cortex, back scattering signals comparable with the progressive blood flow decline of renal cortex also decreased. Among all CEUS quantitative indexes, AUC was significantly increased and DPI was significantly decreased in statistical analyses. Further ROC curve analysis indicated that the diagnostic accuracy of DPI and AUC were better than 0.7. Applied 12 dB as the cut-off value of DPI and 1400 dB·sec as the cut-off value of AUC, relatively satisfactory diagnostic predictive efficiency were achieved of those CEUS quantitative indexes. These results suggested that changes in those quantitative indexes provided better insight into the vascular perfusion damage of renal cortex, they may be valuable to quantify the impaired microcirculation perfusion in DN patients.
Some basic limitations may exist for CEUS quantitative perfusion: first, in the analysis of TICs and perfusion indexes, the ROIs should be defined with the same size and shape, in the comparable depth of renal cortex. Second, the different causes of DN patients, whether the patients taking drugs may alter hemodynamics of renal cortex and affect the final quantitative results.
Our preliminary findings showed CEUS quantitative analysis might be helpful for evaluating early perfusion changes in DN patients. AUC and DPI might be valuable quantitative indexes.