We have shown that CEUS is a highly reliable method of distinguishing benign from malignant liver lesions in most patients treated for pediatric malignancies. Our findings add to the growing body of evidence that CEUS can replace other imaging modalities for this indication [19, 20]. A critical aspect in achieving a high degree of diagnostic accuracy with CEUS in the oncology setting is the need to consider the clinical scenario in which the examination is performed. This is evidenced in our study by the improved sensitivity, specificity, PPV, NPV and accuracy of the “un-blinded” reviews compared to the “blinded” reviews. We also found that the less experienced radiologists had more indeterminate classifications than the more experienced radiologist, which may have partially accounted for the moderate agreement between reviewers. This is notable since the study radiologists had the same clinical information available during the “unblinded” review. We attribute this to several factors. First, we acknowledge that there is a learning curve in the performance and interpretation of CEUS findings. Although not addressed by this study, our findings suggest that fewer indeterminate interpretations are made as radiologists gain more experience and confidence in the modality. In our study, this was coupled with inherent subject matter expertise of the more experienced study radiologist resulting from years of practice in a predominantly oncology setting compared to the less experienced radiologist and radiology trainee. This highlights a challenge to the generalizability of CEUS in the evaluation of incidental liver lesions especially in patients with a history of cancer. Knowledge of the biologic behavior of the primary malignancy is essential since some pediatric cancers can metastasize long after diagnosis while others rarely metastasize to the liver. We found that another important feature to consider is the likelihood of liver cancer (primary or metastatic) in children with cancer predisposition syndromes. The likelihood of infection should also be considered and may be reflected by laboratory values or presence of fever [21].

Perhaps most importantly, in the oncology setting, CEUS can help avoid the anxiety often associated with a “wait-and-watch” approach. At our institution when a liver lesion is discovered on routine imaging, CEUS is performed either the same or the next day. Because CEUS has high sensitivity, specificity, NPV and accuracy we can usually confidently report to the treating physician that the patient has a lesion that is highly likely to be benign or requires additional evaluation. This has led to a paradigm shift at our institution. Rather than bringing the patient back for early follow-up (wait and watch) or obtaining additional imaging with CT or MRI, we proceed directly to CEUS for a more cost-effective, sedation and radiation free evaluation that allows prompt and highly accurate results, improving overall clinical management.

The primary objective of our study was to determine if CEUS could accurately distinguish benign from malignant liver lesions, and we did not endeavor to further characterize the benign lesions. In the study by Smith et al. of 46 pediatric solid tumor patients who developed FLLs, 14 (36%) were classified as FNH (possibly FNH-like) either based on clinical and imaging findings (n = 10) or pathologic inspection (n = 4). Other benign etiologies included cysts, perfusion abnormalities and focal fat. In that study, the time to development of FNH (FNH-like) lesions was longer than for other benign etiologies and malignant lesions. Others have reported FNH (FNH-like) lesions among children at approximately 5.6 years after completion of antineoplastic and cytotoxic chemotherapy for a malignant tumor [22, 23]. In our study, 51 subjects (51/68, 75%) were treated for solid malignancies while 17 (17/68, 25%) had hematologic disease. Among the solid malignancies neuroblastoma (n = 10) and CNS tumors (n = 9) were the most common diagnoses. However, among all diagnoses, leukemia was the most common (n = 11). Although we cannot be certain of the etiology of many of the liver lesions in our study, we found that most were benign and the mean time to discovery was 5.72 years after completion of therapy. However, this finding should be interpreted with caution since without systematic intermittent imaging screening across the cohort we cannot be certain of the time point at which the liver lesions developed. It remains unclear whether there is a common inciting factor among various pediatric cancer diagnoses and treatments or if there are underlying predisposing factors to developing benign liver lesions in this cohort. Importantly, however, it should not be assumed that patients who develop liver lesions during long term follow-up after completion of therapy have only benign disease. Recurrent cancer in long term survivors of pediatric malignancies remains the number one cause of death [24], and ongoing worry about recurrent disease occurs in more than 25% of parents of childhood cancer survivors [25]. Therefore, the minimal time and cost of performing CEUS in this patient population is far outweighed by the benefit.

Perhaps the most informative findings in our study are the etiologies of biopsied indeterminate and false positive results on the “unblinded” reviews. Among the 4 indeterminates classified by the more experienced radiologist, 3 were malignant, and occurred in very young children (5 months, 19 months, and 3 years old) with underlying cancer predisposition syndromes (trisomy 18, Beckwith Wiedemann, Li Fraumeni respectively). These lesions showed only subtle washout (n = 2) or isoenhancement (n = 1) in the delayed phase. The 4th indeterminate lesion was a pathologically proven FNH-like lesion that occurred in an adolescent boy who was 12.5 years off therapy for medulloblastoma. There was concern that the patient was at risk for a second malignancy and the CEUS features were consistent with fibrolamellar HCC which has a predilection to occur in adolescent boys. Kong and colleagues reported hypoenhancement in the delayed phase in 10.7% (3/28) of FNH lesions in adults [26] and this pattern was seen in 2 of the 4 FNH lesions that were biopsied in our study. Others have speculated that this might be explained by retention of the contrast material in the damaged liver parenchyma that surrounds the FNH-like lesion [10]. Additional studies evaluating the CEUS enhancement patterns of FNH vs. FNH-like lesions are needed to better understand the expected findings in these conditions. All 4 false positive lesions in the more experienced radiologist’s (MBM) “unblinded” review showed contrast washout/hypoenhancement in the portal venous and delayed phases. Three of the 4 lesions were biopsied; 2 represented inflammatory or infectious processes and 1 showed hepatocytes in a liver that was iron overloaded. Another biopsied lesion, which was interpreted as benign on CEUS (i.e., no washout), also showed hepatocytes with iron overload. We are unable to explain the discrepancy in the CEUS enhancement patterns of lesions occurring in iron overloaded livers in our study subjects although it is possible that they represent developing regenerative nodules or FNH-like lesions. Importantly, among 5 biopsied infectious and inflammatory lesions in our study, 3 showed hypoenhancement in the portal venous and delayed phases. One, a microabscess/necrotizing granuloma (Fig. 3), showed substantial hypoenhancement in all three phases. Our findings are consistent with others who have shown that some benign lesions, such as scars, granulomas and inflammatory pseudotumors, may exhibit arterial phase iso or hypoenhancement and late phase hypoenhancement similar to malignant lesions [13, 27, 28]. Such lesions may also show a central area of non- or hypoenhancement depending on the degree of central liquefaction and replacement by purulent material [28, 29]. Taken together these findings suggest that the lack of arterial phase hyperenhancement or presence of arterial iso or hypoenhancement, coupled with washout/hypoenhancement in the portal venous or delayed phases, may be an important feature in distinguishing infectious or inflammatory from malignant etiologies of liver lesions.

While a direct comparison of CEUS to hepatocyte specific MRI was beyond the scope of this study, hepatocyte specific MRI has the benefit of allowing a more global assessment of the entire liver, may identify subtle lesions and can provide additional lesion characterization that aids in distinguishing benign from malignant etiologies [30,31,32]. Therefore, we recommend that hepatocyte specific MRI be performed when the results of CEUS are equivocal, are not in keeping with the clinical index of suspicion and when a malignant etiology is suspected (see Fig. 4). However, hepatocyte specific MRI must be interpreted with caution in the setting of iron overload. Because iron deposition in hepatocytes significantly decreases background liver signal, the typical hepatocyte specific contrast enhancement features of benign and malignant liver lesions may be significantly altered. In such cases, CEUS may be the preferred method to characterize liver lesions.

There are several limitations to this retrospective study. Although we had a limited sample size our study offers insight into a unique patient population that has not been previously reported. Additionally, histologic proof was available in only 19 of the 68 lesions, however, given that liver lesions are often found incidentally and are typically of low malignant potential, it would be unethical to perform biopsy on all patients. While general patterns of ultrasound contrast-enhancement are established for benign and malignant lesions, the patterns of treated metastatic disease are not well understood. It is possible that some of our cases determined to be benign based on clinical follow-up reflect treated metastatic disease that was not previously detected. However, a benign lesion vs. treated metastatic disease may carry the same functional outcome. The ultrasound contrast agent used was not uniform throughout the study although most lesions (90%) were imaged with one agent. There were technical challenges during examination of very young patients, some of whom had difficulty holding still, and breathing motion sometimes made it difficult to maintain the lesion within the field of view. However, a radiologist was present for every examination and repeat injections were given, as needed, to improve diagnostic confidence. Another limitation is that the experienced study radiologists may have had some recall bias after being unblinded to the subject’s clinical history. We attempted to mitigate this by allowing sufficient time to pass between the clinical exam and the review for study purposes.