Magnetic resonance‐based biomarkers in nonalcoholic fatty liver disease and nonalcoholic steatohepatitis

Abstract Nonalcoholic fatty liver disease is a growing epidemic affecting 30% of the adult population in the Western world. Its progressive form, nonalcoholic steatohepatitis (NASH), is associated with an increased risk of advanced fibrosis, cirrhosis and liver‐related mortality. Therefore, the detection of NAFLD and risk stratification according to the severity of the disease is crucial for the management of patients with NAFLD. Liver biopsy for such risk stratification strategies is limited by its cost and risks; therefore, noninvasive alternatives have been developed. Among noninvasive biomarkers developed in NAFLD, magnetic resonance (MR)‐based biomarkers have emerged as key noninvasive biomarkers in NAFLD with the ability to accurately detect hepatic steatosis and liver fibrosis. The potential utility of MRI for the detection of NASH and functional liver assessment has also recently emerged. In the current review, we will discuss the data supporting the utility of MR‐based biomarker for the detection of features of NAFLD and its potential use in clinical practice and clinical research in NAFLD.


| INTRODUC TI ON
Nonalcoholic fatty liver disease (NAFLD) is now considered as one of the most prevalent aetiologies of chronic liver disease worldwide affecting approximately 30% of the adult population. 1,2 NAFLD encompasses a spectrum of severity from a simple accumulation of fat in the liver to nonalcoholic steatohepatitis (NASH) considered as the progressive form with higher risk to progress to advanced fibrosis or cirrhosis 3 which is associated with a higher risk of liver-related mortality. 4 In addition, NASH-related cirrhosis is currently the second leading indication for liver transplants in the United States. [5][6][7] Therefore, the detection of NAFLD and risk stratification according to the severity of the disease is crucial for the management of patients with NAFLD and pharmacological therapy for NASH has become an intensive field of research with promising new therapies under development. 8,9 As NASH is a disease defined by biopsy, any noninvasive imaging biomarker will per definition be compared to the biopsy-defined characteristics of NASH, that is steatosis, inflammation and ballooning but also stage of fibrosis. 10 However, this expensive and invasive procedure is not applicable for the screening of NASH and liver fibrosis at the level of high-risk population or to assess longitudinal change in NASH or liver fibrosis. In addition, liver biopsy is limited by sampling error and significant inter-and intra-observer variability. [11][12][13][14][15] Therefore, noninvasive, precise, reproducible, accurate surrogates are needed for the detection of the different stage of NAFLD including NAFLD, NASH or liver fibrosis and to monitor changes in stage of the disease over time.
Several noninvasive biomarkers of NAFLD assessment including serum biomarkers, clinical predictor rules or imaging-based measurements have been developed. 16,17 Magnetic resonance (MR)based biomarkers have emerged as key noninvasive biomarkers in NAFLD encompassing several modalities that enables accurate assessment for hepatic steatosis quantification or liver fibrosis assessment. Since NASH is defined by biopsies this limits the clinical utility of functional liver tests since they will per definition not compare directly with the static biopsy derived information. However, emerging data suggest that functional liver imaging may also play a role, especially in the development of new treatments for NASH, especially when the mode of action of these drugs can better be reflected by change in liver function. In the current review, we will discuss the data supporting the utility of MR-based biomarker for the detection of features of NAFLD and its potential use in clinic or clinical research in NAFLD. The methods are summarized in Table 1.

| Magnetic resonance spectroscopy
The presence of NAFLD is defined by the presence of hepatic steatosis ≥5% either by imaging or histology. 18 MRS noninvasively measures proton signals as a function of their resonance frequency. The signal intensity at frequencies corresponding to water or fat can be quantified, and the fat-signal fraction can be calculated. MRS is highly sensitive for the detection and quantification of even small amounts of liver fat, and MRS is considered as the most accurate noninvasive method to quantify liver fat. [19][20][21] However, the use of MRS in clinical practice is limited as it is not available on all clinical scanners and requires dedicated spectroscopic sequences and timeconsuming postprocessing analysis. In addition, MRS is restricted in spatial coverage owing to the volume selection required limiting measurements to a small portion of the liver. All these limitations impede the use of MRS for longitudinal monitoring.

| Magnetic resonance imaging
As MRS, magnetic resonance imaging exploits the difference of the resonance frequencies between water and fat proton signals. MRIproton density fat fraction (PDFF) take into account several confounding factors that may affect the MRI estimation of tissue fat concentration for an accurate quantification of hepatic steatosis. 22 PDFF is defined as the ratio of the density of mobile protons from triglycerides and the total density of protons from mobile triglycerides and mobile water. MRI-PDFF is a quantitative imaging biomarker that enables accurate, repeatable and reproducible quantitative assessment of liver fat over the entire liver. [23][24][25][26][27] Fundamental difference between histologic and MRI-PDFF assessment of hepatic steatosis relies on the feature measured by each modality. Histological assessment estimates the number of steatotic cells in the liver, while MRI-PDFF estimates the overall percentage of MRI-visible protons on fat molecules in the liver. 28 Therefore, as the fat content of a cell does not generally exceed 50%, MRI-PDFF percentages are almost always less than half the value derived by histology.
MRI-PDFF is well validated using magnetic resonance spectroscopy (MRS) as reference 27,29-32 and against histology-proven steatosis grade. 21,26,33,34 In a recent meta-analysis of 23 studies with 1,679 participants, MRI-PDFF was shown to have excellent linearity, bias and precision across different reconstruction methods, and MR scanners of different field strength and manufacturer. 35 Advantages of MRI-PDFF are to rapidly assess PDFF over the entire liver in a short breath hold (~20 seconds). PDFF maps are automatically reconstructed without user input or postprocessing. In addition, MRI-PDFF methods are FDA approved and are commercially available on the three major MRI vendors, GE Healthcare, Siemens and Philips, ensuring potential widespread availability. in over 50 intervention studies in NAFLD or NASH. An example of a PDFF map from the same patient before and after treatment is shown in Figure 1. In this case, the liver volume was also assessed since this can be of importance in understanding unexpected effects from interventional studies. An example of that is the increase in liver volumes induced by, for example, fenofibrates. 42 MRI-PDFF has thus emerged as noninvasive imaging biomarker suitable as an end-point in clinical trial in NASH for internal decision-making. 43 For regulatory purposes in phases 2B and 3, biopsies are however still required.

TA B L E 1 Magnetic resonance-based modalities available for the assessment of NAFLD
Finally, MRI-PDFF has been associated with longitudinal change in histologic feature of NAFLD including NASH and liver fibrosis. Patel et al 44 have shown that the relative reduction of liver fat quantified by MRI-PDFF is associated with a histologic response in NASH.
Finally, preliminary data have also suggested a prognostic value of

MRI-PDFF in NAFLD progression. Ajmera et al have shown that
baseline MRI-PDFF fat content is associated with longitudinal progression of fibrosis in patients with biopsy-proven NAFLD. 45 These preliminary data need to be validated in larger independent cohorts.
Overall, MRI-PDFF is emerging as one of the leading noninvasive quantitative biomarkers for the quantification of hepatic steatosis in term of accuracy, precision and reproducibility. Although its cost is a limitation for a use in routine clinical practice, its utility is valuable in the context of clinical trials and may also be useful as a prognostic factor of progression of regression of NAFLD in future.

| Magnetic resonance elastography
The MRE is an MRI-based technique that images the propagation of acoustic shear waves in the liver and applies a mathematical algorithm to compute cross-sectional images displaying the magnitude F I G U R E 1 An example of a MRI-PDFF map from the same patient before and after intervention. In this case, the entire liver is measured excluding major bile ducts and veins to improve precision. Here, also the liver volume was measured and reductions both in PDFF and liver volume are seen. The whole liver analysis also enables the histogram display and analysis of pixels throughout the liver as shown in the figure of the complex shear modulus of liver tissue. 49  reported in an independent cohort by Loomba et al. 59 The sensitivity of MRE to discriminate between lower stage of fibrosis F0 versus F1-4 was lower, and further studies with larger multicentre cohort are needed to confirm the optimal thresholds and diagnostic performance for the detection of individual stage of fibrosis. Further data are needed to determine the place of each modality and to develop optimal and cost-effective algorithm using step wise approaches for the assessment of liver fibrosis.

| O ther MR-based biomarker for the detec tion of NAFLD an d NA S H: T1 , correc ted T1 an d multipar ametric MRI
There are currently no direct methods clinically available to assess disease activity including ballooning and inflammation in NASH.
There are however several MRI methods that have been suggested to be related to disease activity and fibrosis such as magnetization transfer contrast and diffusion-weighted imaging and T1 mapping, where T1 mapping is most widely used method today. T1 mapping of liver disease was described already in 1981 67 and has also more recently been described for assessing liver fibrosis in cirrhotic patients. 68 It was also proposed that T1 is a representation of extracellular fluid in the liver by Banerjee et al 69  shown that cT1 could discriminate between groups with different activity scores. It should however be noted that an overlap between individual groups is present, like that of fibrosis assessment with MRE. In addition, data on the prognostic value of cT1 71 are available but these data are originating from small sample sizes and larger studies are therefore warranted. cT1 is currently being deployed in the UK-Biobank study, 72 so outcome data from much larger populations can be expected in the future.
There are also some questions to be addressed using T1 and cT1 to monitor longitudinal changes in NAFLD and NASH. A potential issue is the fluctuating stores of liver glycogen, glycogen binds large amounts of water. 73 This means that any intervention that alters liver glycogen stores also can induce changes in T1 independent of inflammation and fibrosis. Another potential issue is the use of R2* for correction of T1 since there is a strong dependency of R2* on liver fat as shown recently by Bashir et al. 74 In fact, in this study it was shown that liver fat is the most influential covariate of hepatic R2* both at 1.5 and 3T. This means that any intervention inducing a change in liver PDFF also will induce a change in cT1 owing to the change in R2*. It can of course still be so that reductions in inflammation and fibrosis can occur and induce changes in cT1, but interventional data need to be considered in the light of changes in hepatic PDFF. In addition, the Bashir paper showed there were only very limited number of subjects that suffered from abnormally high R2* and hence very few subjects that require correction. Furthermore, the relationship between hepatic PDFF and fibrosis is not linear but rather biphasic, 57

| A ssessment of liver func tion by MRI
In most other disease areas, circulating biomarkers or imaging biomarkers that define the disease are available such as HbA1c in type The use of gadoxetate disodium is indicated for intravenous use in T1-weighted MRI of the liver to detect and characterize lesions in adults with known or suspected focal liver disease. Its mechanism is that normal hepatocytes take up the contrast agent and make normal parenchyma brighter on T1-weighted images while tumour cells do not take up the contrast agent, hence increasing the contrast between focal lesions and normal liver tissue. However, it has also been shown that the relative signal enhancement following injection of gadoxetate disodium is reduced in patients with fibrosis. [76][77][78] This could be explained by two possible mechanisms: a. dilution of functioning hepatocytes by presence of fibrosis b. reduced uptake into the hepatocytes by reduced OATP1 action It has also been shown that gadoxetate disodium uptake is reduced also in subjects with hepatic inflammation and ballooning. 76 It could therefore be hypothesized that anything that should not be present in the liver, for example fibrosis and inflammatory cells, will dilute the concentration of functional hepatocytes and that the conditions associated with fibrosis and inflammation, for example oxidative stress and mitochondrial dysfunction, also will affect the transporters associated with gadoxetate disodium uptake and excretion.
Several studies have been performed utilizing dynamic imaging with gadoxetate disodium, 79,80 and this allows quantitative information to be extracted by compartmental modelling yielding data on uptake and excretion rate of gadoxetate disodium in the hepatocytes but also quantitative assessment of extracellular volume in the liver, the same parameter that is associated with T1 but here directly However, the following drawbacks need to be considered. The use of Gadoxetate Disodium is currently not indicated for patients with NAFLD, furthermore the use of a Gadolinium-based contrast agent may carry an increased risk with potential retention of gadolinium in the body 82 and are contraindicated in subjects with GFR < 30, and finally there are currently no interventional data in humans available using gadoxetate disodium in NAFLD and NASH.
Other functional methods assessing the liver include quantification of portal flow using phase-contrast MRI and portal pulsatility as a measure of vascular resistance in the liver using MRI. As described previously, the concept of stress testing is readily available in the diagnostic workup in other disease areas but not for NAFLD. There are however possibilities to combine MRI-based methods such as gadoxetate disodium imaging or 31P-MRS of ATP with functional challenge, for example fructose, to determine the functional reserve capacity of the liver to better understand the prognosis of a patient or effects of a pharmacological treatment. These methods are however not readily available, and further studies are needed to understand and validate the utility of these methods.

| CON CLUS ION
MRI is today an important tool in research of NAFLD and NASH patients. The most commonly used methods include PDFF, MRE and T1 measurements. Out of these, PDFF is well validated and frequently used and gives a more reproducible assessment of liver steatosis than liver biopsy. MRE, T1 measurements and functional MRI are also used in interventional studies but more data on relation between treatments effects seen with MRI/MRE and biopsy is required to fully understand the utility in clinical research. All these methods can also be used in the clinical workup of patients, and again here, PDFF is the best validated method. Larger prospective studies using some of the methods described in this paper are underway and will guide us about the clinical utility.

F I G U R E 3
Displayed is an example of dynamic gadoxetate disodium imaging. The time-intensity curves represent the signal in the aorta and the liver parenchyma. To the right is shown the compartmental modelling used

CO N FLI C T O F I NTE R E S T S
C.C reports no conflict of interests, and L.J is an employee and shareholder of Antaros Medical.

E TH I C A L A PPROVA L
This is a review paper; no separate informed consent has been obtained to write this.

DATA AVA I L A B I L I T Y S TAT E M E N T
No new data generated in this review article.