Diagnostic accuracy of contrast-enhanced ultrasound for characterization of kidney lesions in patients with and without chronic kidney disease

The incidence of kidney cancer is increasing with over 60,000 new kidney cancer diagnoses projected for 2017. Kidney cancer is also deadly. Over 14,000 associated deaths are expected in 2017 [1]. If detected early, kidney cancer can be treated effectively with surgery alone. Benign cystic lesions are common, accounting for up to 30% of all identified lesions [2]. Distinguishing malignant kidney neoplasms from benign lesions is crucial to determining appropriate treatment. Diagnostic options include biopsy [3, 4] or interval imaging over months or years [5‐7]. Cystic lesions can be characterized with contrast-enhanced computed tomography (CT) or magnetic resonance (MR) imaging [8] using the Bosniak criteria [9, 10].

However, Bosniak classification requires contrast enhancement which may pose risk in some patients with compromised kidney function. Risks associated with iodinated contrast agents used with CT imaging include further impairment in kidney function for those with advanced stages of chronic kidney disease (CKD). Use of gadolinium-based contrast agents in this population also poses a rare but serious risk of nephrogenic sclerosing fibrosis. Incidence of this disease has virtually disappeared with more stringent screening prior to gadolinium exposure and use of agents with better risk profiles. However, gadolinium deposition in various organs is now being postulated even in patients with normal kidney function, although the clinical significance is not yet known [11, 12]. Patients with impaired kidney function experience prolonged exposure to gadolinium since it is cleared primarily by the kidneys and thus have an increased risk of gadolinium deposition [13]. Over 40 million people in the United States have chronic kidney disease (CKD), and this population is rapidly growing (USRDS 2016). These patients have substantially increased risk for the development of kidney cancer [14‐16]. However, the typical screening tools of contrast-enhanced CT and MR are often contraindicated in patients with advanced stages of CKD. In patients with moderate stages of CKD, kidney-protective measures can be taken but add additional steps, can be cumbersome and do not entirely eliminate risk. Even in these populations, if a patient has a high-risk lesion, the benefits of a contrast CT or MR study may outweigh the risks from contrast exposure. For patients with lower or moderate risk lesions, the risk of contrast-enhanced CT or MR may not be worth the benefit. Radiologic options for this population are limited [17]. Non-contrasted surveillance imaging is often performed, but this is inferior to contrast enhanced studies for lesion characterization.

A potential alternative imaging modality for evaluating kidney lesions is contrast-enhanced ultrasound (CEUS). CEUS has been used to characterize indeterminate lesions in multiple organs [18, 19], including the kidneys [20‐27]. CEUS has the ability to detect vascularization as the microbubble contrast agent remains purely intravascular [28, 29]. A unique aspect of CEUS is its ability to image contrast dynamics in real time [30, 31]. US contrast agents have a low serious adverse event rate of 0.006–0.009%, consisting primarily of anaphylactoid reactions that resolve when the contrast is cleared and typically do not necessitate hospitalization [32, 33]. Less serious side effects that are also rare and typically transient, resolving when the contrast agent is cleared, include headache, dizziness, flushing, nausea, flank pain and chest pain. An additional benefit setting them apart from other contrast agents is that they are excreted by exhalation through the lungs and are therefore not nephrotoxic and safe for patients with kidney disease. Despite the potential for CEUS as a diagnostic option among individuals with CKD, few studies have evaluated CEUS in this population, although interest is increasing [34, 35]. In this exploratory pilot study, we sought to investigate the diagnostic accuracy of CEUS for detecting enhancement of kidney lesions, and thus risk of malignancy, in populations with and without CKD and subsequently determine if this imaging modality has potential to be an alternative screening tool in patients with contra-indications to contrast CT or MR. We hypothesized that CEUS has sensitivity comparable to contrast-enhanced CT or MR among individuals with and without CKD.

Evaluation of kidney lesions by ultrasound has historically been limited by lack of enhancement. Contrast-enhanced CT/MR has been the imaging standard. However, iodinated contrast agents may worsen kidney function, and gadolinium agents may cause nephrogenic systemic fibrosis in patients with advanced CKD. The primary goal of this pilot study was to explore the detection of microbubble contrast in kidney lesions among patients with and without CKD and to test the accuracy of CEUS for characterizing these lesions as malignant or benign (Fig. 3). We found that the sensitivity of CEUS, whether compared to tissue diagnosis or follow-up imaging, is high and comparable to the reported sensitivities of CT (83–100%) and MR (81–100%) [21‐24, 40] among patients with and without CKD. High sensitivity provides high certainty that a negative test in an individual means the individual does not have the disease, essentially ruling out malignancy.

Although we found high sensitivity with CEUS, specificity was low. Whether compared to tissue diagnosis or tissue diagnosis/follow-up imaging, specificity of CEUS is lower than reported CT (51–96%) and MR (71–100%) specificities [21‐24, 40]. Low specificity means a large number of false positives, thus a positive test does not mean the individual has the disease and cannot rule in malignancy very well.

A test with high sensitivity and lower specificity (for example, mammography or CT for lung cancer) is suitable for screening purposes. Benign cystic masses are common in renal insufficiency and distinguishing malignant from benign cysts is a diagnostic challenge. CT or MR is the current standard, but performing CT/MR on all patients with cystic masses would be prohibitively costly and expose a large number of patients to radiation or potentially nephrotoxic contrast agents. CEUS’s ability to exclude malignancy with minimal risk of adverse effects means that the use of CT/MR could be narrowed down to only those patients with a positive CEUS.

In the subgroup of patients who also had CT/MR, the number of patients who were CEUS negative but CT/MR positive is smaller than the CEUS positive but CT/MR negative group, suggesting that CEUS tends to assign a higher Bosniak category (or solid designation) than CT/MR (Fig. 4 and Fig. 5). This analysis does not take diagnosis into account and can therefore only be used to suggest that CEUS is more likely to call a lesion malignant than CT/MR and does not lead to conclusions about false positive or negative rates.

There are three possible explanations: First, ultrasound is more sensitive than CT/MR for detecting the presence of intravascular contrast. Unlike CT and MR agents, microbubble contrast agents remain intravascular: they are not filtered by glomeruli and do not extravasate into the interstitial space [41]. This may lead to more prominent internal enhancement with CEUS than CT/MR agents [22, 23]. Second, ultrasound has greater spatial resolution than CT/MR: in some cases (Fig. 4), cysts that appear simple on CT/MR show internal features on ultrasound, including septations not seen on CT, resulting in a higher Bosniak grade using definitions based on CT or MR imaging. Better spatial and contrast resolution is an important advantage to examining lesions with CEUS; however, to reduce the false positive rate, the grading system will need to be modified to take this into account. Future, larger studies should be performed to develop a revised classification system [42]. Using objective measures of enhancement, similar to Hounsfield units for CT, may also improve CEUS accuracy. Furthermore, new CEUS imaging techniques such as super-resolution imaging [43], molecular imaging [44], and acoustic angiography [45] may further improve specificity to malignant lesions as they are translated into the clinic. The third possibility is that relative inexperience with CEUS compared to CT or MR leads to upstaging of kidney lesions.

Few studies have looked specifically at the use of CEUS for characterizing lesions in patients with CKD [46]. In this study, we examined patients with and without CKD. With more advanced CKD, overall accuracy declined, due largely to a decrease in specificity. While sensitivity only dropped from 100% for early CKD to 83% for advanced CKD, specificity dropped from 72% to 41%. The reason for this requires further investigation, but we observed that while we were able to detect contrast enhancement in the uninvolved parenchyma for all patients, parenchymal enhancement was more heterogeneous and reduced in patients with CKD compared to non-CKD patients. We therefore hypothesize that a decrease in background parenchymal enhancement accentuates the difference in lesion enhancement compared to surrounding parenchyma. This abnormal parenchymal enhancement is even further pronounced in advanced CKD and may cause the lesion to stand out more, creating the perception that it is enhancing. This would be expected to primarily affect specificity, as the false-positive rate would increase with worsening CKD. With this in mind, we hypothesized that subjects with vascular CKD pathology (diabetes or hypertension) would have more false positives than non-vascular etiologies (cystic, glomerular, drug-related) but found no significant differences, likely due to small numbers. This should be explored further in larger populations. Nonetheless, the high sensitivity for ruling out malignancy in the CKD population suggests that this modality could be used for screening following non-contrast ultrasound or CT for patients in whom contrast-enhanced CT/MR is contraindicated (Fig. 6). Larger studies are needed to confirm and extend our findings.

Based on our findings, we suggest that CEUS be considered to evaluate low to moderate-risk lesions in patients who have absolute or relative contra-indications to contrast CT or MR. If CEUS does not detect lesion enhancement, the patient may not need to undergo contrast CT or MR. If CEUS does detect lesion enhancement, further imaging should be considered. In addition, CEUS may have an application as a secondary test for lesions not definitively diagnosed by either contrast CT or MR. Due to the greater sensitivity to enhancement and special resolution, CEUS may be able to detect lesion enhancement where other modalities may not. This potential application should be evaluated in future studies.

Although CEUS is not yet FDA-approved for kidney imaging in the United States, it is increasingly used off label. Interpretation tools, such as the customized GUI, that address each component lesion characteristic of the Bosniak classification system, can assess reader consistency and inter-reader agreement. Observational training visits to institutions that have experience and familiarity with CEUS will improve implementation and interpretation accuracy.

This study has limitations. First, this is a pilot study with a relative small size population from a single center. Due to the small size, meaningful subgroup analyses, such as those at each stage of CKD and those with cystic vs. solid lesions, was not possible. Because of this limitation, the conclusions drawn are preliminary in nature and should be interpreted with caution. Future, larger, multicenter studies are needed to determine the true accuracy of CEUS among individuals with diverse stages of CKD. Second, the blinded readers in this study were academic genitourinary radiologists with fellowship training in ultrasound. They underwent a preliminary training session and performed CEUS literature review, but neither had prior experience with CEUS. However, their background is not atypical as physicians are now becoming more familiar with CEUS in the United States, particularly in academic centers. Reassuringly, the sensitivities observed were similar to those reported by more experienced centers [34]. As there is little available formal training, development of a training program and experience with interpretation of CEUS is needed before widespread implementation. Additionally, because CEUS is a perfusion study, applying objective perfusion parameters could increase accuracy. Third, the reference standard of tissue diagnosis was not available for all patients because many indeterminate lesions underwent imaging surveillance rather than biopsy or surgery. A minimum of one-year follow-up was accepted, but the ideal follow-up period is 3–5 years, depending on lesion complexity, as some renal cancers are slow growing. Until longer follow-up data on more patients is obtained, malignancy cannot be ruled out with a Bosniak IIF lesion, and these patients would still need to be followed, as is the case with contrast CT and MR. Fourth, both cystic and solid lesions were included. In order to more accurately determine accuracy and potential modifications to Bosniak criteria for CEUS in a CKD population, future studies will need to include primarily cystic lesions. Fifth, the referral source for the two groups, non-CKD and CKD, was different. This introduces potential bias since the non-CKD group came primarily from urology clinic where patients were referred for evaluation of highly suspicious lesions (i.e. likely higher risk lesions), and the CKD group came primarily from non-urologists who were longitudinally following indeterminate lesions (i.e. likely lower risk lesions). Lastly, as with any ultrasound study, obesity, lesion location and motion artifact from breathing are technical limitations.

The traditional Bosniak classification system was developed to classify kidney cystic lesions based on contrast-enhanced CT characteristics We adapted this classification system for B-mode US/CEUS and found that it has potential as a test to exclude malignancy based on a sensitivity comparable to CT/MRI, but it does not yet have good specificity. Low specificity might be attributed to the greater spatial resolution and contrast detection with CEUS compared to contrast-enhanced CT, leading to upstaging of lesions. It may also be attributed to lack of experience interpreting CEUS studies. Our results suggest that with further refinement of a classification system and more training, CEUS may be an alternative to contrast CT or MR for characterizing indeterminate kidney lesions, particularly in patients with CKD or other contraindications to contrast CT or MR. However, if the traditional Bosniak criteria are used for CEUS, malignancy can still be excluded but a positive CEUS result will require further diagnostic evaluation due to its low specificity. Improved specificity may be accomplished with training programs for radiologists and revisions to the Bosniak criteria, potentially adopting a classification scheme proposed by Barr et al. [24] which incorporates both solid and cystic lesions. Larger studies focused on cystic lesions, particularly in patients with later stages of CKD are needed to confirm and extend our findings and inform classification criteria revisions.