Quantification of liver function using gadoxetic acid-enhanced MRI

Chronic liver diseases (CLD) are a major worldwide health problem. Although the prevalence of CLD from most etiologies has been stable or has even substantially declined, particularly in the case of hepatitis C virus (HCV) or hepatitis B virus (HBV) due to eradication or vaccination therapy, respectively [38], the prevalence of NAFLD has steadily increased, and is now the leading cause of CLD worldwide, affecting 80 to 100 million individuals in the USA alone [39, 40]. Therefore the early diagnosis of NAFLD and accurate assessment of liver disease severity are the keys to optimal patient management, where early treatment and lifestyle modification can arrest, and even reverse, deranged hepatic function [38] and histopathology [41].

Bastati et al. [42] showed that the RLE, after GA administration, can differentiate simple steatosis patients from nonalcoholic steatohepatitis (NASH) patients. This method using cutoff value of 1.24 achieved an area under the receiver operator characteristic curve (AUROC) of 0.85 (95% CI 0.75–0.91). The authors suggested that this method may be a useful screening tool in selecting potential NAFLD patients for liver biopsy. However, despite its high sensitivity (97%), the specificity of RLE was only 63%, likely due to the liver fibrosis that was present in 68% of the cohort. As in many CLDs, fibrosis is a well-known confounder that decreases the RLE [43] (Fig. 3a–e). Similar, previous, retrospective studies [44, 45] and unpublished prospective data from our group confirmed the above-mentioned results.

Liver fibrosis is a key determinant in the natural history of chronic liver diseases (CLDs) [46]. Per definition, liver fibrosis is a healing response to myriad injuries that result in CLD. Common insults include viral (A, B, C and D, E) and nonviral infections, alcohol, and NAFLD, while congenital hepatic fibrosis, Wilson’s disease, and a host of cholestatic liver diseases [47] are among the rare etiologies. Progressive liver fibrosis leads to end-stage liver cirrhosis as a result of fibrogenesis [48]. Moreover, because fibrosis may be reversible, early diagnosis is imperative so that early therapy can be instituted. Therefore, knowing the extent and degree of liver fibrosis in CLD patients is of clinical importance, as it can influence patient´s prognosis, surveillance, and treatment [48]. Liver biopsy currently remains the reference standard for the detection and staging of liver fibrosis, despite the risks associated with its invasiveness, inter-observer variability, potential for sampling errors, and poor patient acceptance, which makes it inappropriate for long-term monitoring or follow-up [49‐51]. Routine biochemical and hematologic tests fail to help quantify liver fibrosis in approximately 50% of patients [51, 52]. There is a large body of literature describing the role of gadoxetic acid-enhanced MRI in the staging of liver fibrosis, using easily obtainable quantitative measures [53]. Feier et al. [54] found that RLE is a good discriminator for the presence of hepatic fibrosis, and Watanabe et al. have found that the contrast enhancement index (CEI) is an efficient biomarker for staging hepatic fibrosis [55]. Further studies have shown that RLE is an independent predictor of fibrosis and is highly accurate for staging F2 fibrosis and cirrhosis (stage F4), with AUROCs of 0.82 and 0.83, respectively [53]. Other groups have found that the hepatobiliary phase (HBP) T1 relaxation time significantly correlated with the fibrosis stage and necroinflammatory activity grade. However, although the HBP T1 relaxation time had a high diagnostic accuracy for stage 3 fibrosis (AUROC of 0.82), its diagnostic accuracy for grade 3 necroinflammatory activity was relatively low (AUROC of 0.68) [56, 57]. The authors found that HBP T1 relaxation times were better at differentiating stage 2 from stage 3 fibrosis than at distinguishing grade 2 from grade 3 necroinflammatory activity (Fig. 4a–c).

Cirrhosis, the end-stage of hepatic fibrosis, is equivalent to stage 4 fibrosis in the Meta-analysis of Histological Data in Viral Hepatitis (METAVIR) system, or stages 5–6 in the Ishak scoring system [58], as well as in the newly introduced systems, including the Yale (Jain-Garcia), Laennec [59], and PIR systems [60].

Today, we know that cirrhosis is a dynamic process that may or may not progress. Indeed, early cirrhosis may be reversible. Thus, the former one-stage clinical and histologic description of “cirrhosis” is simplistic, and misrepresents current dogma [61]. Therefore, in the clinical setting, the term cirrhosis has been increasingly replaced by the term chronic liver disease (CLD), and thus, it is more important to distinguish between compensated CLD versus decompensated CLD, since, once patients cross this clinical boundary into decompensation, the odds are, by comparison, extremely low that they can avoid liver failure without transplantation or retransplantation [28, 62].

As a cure for chronic hepatitis C in responders where eradication of the viral load exceeds 95% post-treatment [63, 64], the focus of attention has now shifted to the regression of HCV-induced liver fibrosis, cirrhosis, and portal hypertension in these patients. Our group [65] hypothesized that the RLE might identify patients who have persistent hepatic inflammation and fibrosis, as well as portal hypertension despite HCV eradication. This is clinically relevant, as surveillance for portal hypertension and hepatocellular carcinoma are particularly important in patients with CLD. We evaluated changes in relative liver enhancement (RLE) obtained by GA-MRI in the hepatobiliary phase and changes in splenic volume (SV) after hepatitis C virus (HCV) eradication, as well as their predictive value for the development of (further) hepatic decompensation during follow-up. We observed an inverse correlation between the changes in RLE and SV (P < 0.001). In the non-responder patients, there was a decrease in RLE of − 11% (− 25% to − 3%; P = 0.019) and an increase in SV of 23% (7–43%; P = 0.004) (both P < 0.001 versus the responder patients). Interestingly, GA-MRI-nonresponse was associated with a substantially increased risk of (further) hepatic decompensation two years after the end of treatment: 80% versus 8%; P < 0.001. We concluded that GA-MRI might distinguish between individuals at low- and high-risk for (further) hepatic decompensation (GA-MRI-nonresponse) after HCV eradication. This could allow for individualized surveillance strategies.

The degree of parenchymal uptake and biliary excretion in GA-MRI may be useful for the evaluation of liver function as cirrhosis evolves [28, 66]. Tamada et al. found that hepatic parenchymal enhancement, i.e., RLE on GA-MRI, is affected by the severity of cirrhosis [67]. Tsuda and others, using RLE, also found that the liver’s signal intensity on the HBP of GA-MRI was lower in cirrhotic than in normal livers [68]. Conversely, Kanki et al. reported that the RLE derived from GA-MRI did not necessarily decrease with increased morphologic features of cirrhosis [69]. Other groups reported the value of T1 relaxometry in cirrhosis [70, 71], either unenhanced, or, after contrast media (CM) injection [72]. More recently, Sandrasegaran K. et al. [73] conducted a retrospective study to determine the value of quantitative parameters of GA-MRI in predicting prognosis in cirrhotics. Examining a cohort of 63 patients who had GA-MRI and a two-year clinical follow-up, they calculated the enhancement ratio (ER), contrast enhancement index (CEI), and contrast enhancement spleen index (CES). Comparing the usefulness of these parameters to clinical scores, such as the Child–Pugh score (CPS) and the model for end-stage liver disease (MELD) score, for predicting adverse outcomes, such as variceal bleeding (VB), hepatic encephalopathy (HE), and mortality at two years, they concluded that an ER of 15 or a CES of 20 were equivalent or better predictors of major morbidity and mortality than these commonly used clinical scores in cirrhotic patients.

Several studies have addressed the value of preoperative GA-MRI in predicting the risk of liver failure should major liver resection be performed [74‐76]. This estimate of hepatic function reserve is crucial to avoid post-hepatectomy liver insufficiency (PHLI) [36, 74].

It is well known that, in the setting of metastatic liver disease, chemotherapy can adversely affect liver function. In particular, metastatic colorectal cancer (CRC) patients who receive specific drugs, such as oxaliplatin and or irinotecan, are predisposed to develop chemotherapy-associated steatohepatitis (CASH) and sinusoidal obstruction syndrome (SOS) [77, 78].

In those patients with less than 25–30% remnant liver function, PHLI is the major cause of postoperative morbidity and mortality. Because morphologic volume may overestimate functional volume [30, 74], volumetry alone (based on CT or conventional MRI) may not deliver an accurate prediction of PHLI in the presence of CLDs, where steatosis and/or fibrosis likely exist.

GA-MRI has proven superior for the estimation of both global and regional function, and studies have shown it to be a more reliable prognosticator than indocyanine green clearance (ICG) [30, 36].

In a retrospective study, our group [74] looked at how well functional future liver remnant (functFLR), as calculated from the RLE on GA-MRI and volumetry on multidetector computed tomography (MDCT) done within 10 weeks of a planned major resection, predicted post-hepatectomy liver failure (PHLF) following major liver resection compared to well-established clinical tests. In a multivariate analysis, we found that a decreased functFLR was independently associated with the probability of PHLF (0.561; P = 0.002). MRI software development is needed to eliminate the use of CT for volumetry, which is the major drawback of this technique.

Furthermore, Cho et al. showed that the RLE is an accurate discriminator for predicting which patients will develop PHLI after major liver resection [79]. Wibmer et al. found that decreased RLE in liver surgery candidates was associated with a higher risk of PHLI and perioperative mortality [30].

Kim DK et al. concluded that GA-MRI-derived measurements predicted PHLF better than the ICG clearance test in hepatocellular carcinoma (HCC) patients who underwent hepatectomy [36].

Currently, GA-MRI and MR cholangiography have been increasingly used to assess liver graft dysfunction, since, unlike conventional gadolinium chelates, these techniques can provide both anatomic and functional information [80, 81]. Our group reported that, among the above-mentioned quantitative parameters derived from GA-MRI, the RLE is one of the most promising for graft evaluation [27]. Global and regional contrast media uptake and excretion of the graft can be calculated, providing the probability of graft survival. We could show that the RLE 20 min after contrast injection was directly related to the probability of one-year retransplantation-free survival in proportional hazard regression analysis (P = 0.005) [27].

Quantitative assessment of liver function using the calculation of the above-mentioned scores, such as RLE, is more tedious and time-consuming and normally cannot be performed as part of a routine MR examination. Furthermore, radiologists are used to assessing morphologic MR features and are less familiar with calculating lengthy and difficult equations. Therefore, Bastati et al. [27] introduced the functional liver imaging score (FLIS), derived from three hepatobiliary phase features on gadoxetic acid-enhanced MRI, each ranging from 0 to 2 on an ordinal scale. Following visual assessment of the liver parenchymal enhancement quality (EnQS), the rate of biliary contrast excretion (ExQS), and the persistence of contrast within the portal vein (PVsQS), the three scores are summed, and can range from 0 to 6, yielding the FLIS. Hence, the use of the term ‘semi-quantitative’ to describe the subjective and objective components that contribute to the FLIS (Fig. 5). Because the FLIS requires no signal intensity measurements, equations, or specific software, and is independent of MRI field strength and vendor, it can easily be incorporated into routine clinical practice.

Imaging techniques, including US, CT, and conventional MRI, have been used to diagnose cirrhosis and its sequelae based on morphological features. However, the prototypical features of cirrhosis bear no relationship to objective (i.e., functional) parameters of cirrhosis severity [6, 82]. Just as the FLIS, derived from GA-MRI, allows estimation of liver function in OLT grafts, it can be used in patients with chronic liver disease (CLD).

Previously described methods with which to assess the hepatobiliary phase uptake include the relative liver enhancement, the hepatic uptake index, the contrast enhancement index, and T1 relaxation values [18, 22, 24]. These methods all require complex computations and have vendor, field strength, and sequence dependencies that complicate their clinical application. FLIS is more practical for routine use than quantitative parameters, as it requires no SI measurements or calculations. The use of such an easy and reproducible score in clinical practice may potentially lead to better management of patients with CLD.

In a retrospective study, 265 patients with CLD, who had undergone GA-MRI, were assigned a FLIS score ranging from 0 to 6, based on the sum of three hepatobiliary phase features, as described above [28]. Patients were stratified into the following three groups according to fibrosis stage and a presence or history of hepatic decompensation: nonadvanced CLD; compensated advanced CLD (CACLD); and decompensated advanced CLD (DACLD). The predictive value of FLIS for first and/or further hepatic decompensation and for transplant-free survival, i.e., mortality, was investigated using Kaplan–Meier analysis, log-rank tests, and Cox regression analysis. Intra-observer and inter-observer intraclass correlation coefficients for FLIS were excellent, demonstrating the robustness and reproducibility of this scoring system [28].

The results demonstrated that the FLIS is an independent predictor of liver-related events (e.g., first hepatic decompensation) and transplant-free survival (mortality) in patients with different causes of chronic liver disease (CLD). Both decompensated and compensated patients with CLD with a reduced FLIS showed a three- to seven-fold higher risk of mortality, respectively, even after adjusting for established prognostic factors (Fig. 6a–d).

GA has an excellent track record in focal liver lesion detection and characterization, as well as in liver function estimation. Transient severe motion (TSM) artifacts, which may degrade arterial-phase images in 18% of patients, on average [83], is the main limitation of GA. However, various techniques, including free-breathing, obtaining multiple arterial phases, saline dilution of GA, and/or slower injection rates, have successfully alleviated TSM [84, 85]. Furthermore, the 20-min delay between dynamic and HB phase imaging need not be wasted. By injecting GA at the start of the exam, dynamic imaging, T2 w-images, DWI, MR elastography, and other techniques, such as T2 mapping, can be performed while waiting. GA is preferable to gadobenate, not only because of the shorter interval between contrast injection and late-phase imaging but also because gadobenate’s negligible hepatocyte uptake limits its role in estimating liver function.

In conclusion, the estimation of liver function is a crucial clinical issue for identifying the broad spectrum of hepatobiliary disorders, for monitoring the progression of chronic liver disease (CLD), for determining the optimal therapeutic strategy, as well as for the prevention of post-interventional hepatic failure.

GA-MRI can be used to evaluate liver morphology, as well as calculate several quantitative parameters, including RLE, HUI, CEI, and the semi-quantitative FLIS parameter, which is independent of field strength, exam parameters, scanner type, or vendor.