Discussion
Several studies have evaluated the use of readily available laboratory test results to predict significant fibrosis or cirrhosis in patients, such as Forn's score, the FibroIndex, the Lok index, and APRI. Because all these fibrosis markers need platelet counts and many need AST, patients whose platelet count has changed by other diseases or who have severe active hepatitis sometimes might have values largely different from actual fibrosis. Many studies have reported the evaluation of liver function or fibrosis with Gd-EOB-DTPA-enhanced MRI effects in liver damage. RE of liver parenchyma increased significantly with time until at least 20 min in normal or moderate liver dysfunction, but Child–Pugh class C did not show any increase after the portal phase. Tajima et al. reported that in both normal and chronic liver disease groups, liver enhancement increased gradually up to 30 min in the hepatobiliary phase, and the peak value at 30 min with normal liver function was higher than that in chronic liver dysfunction.
In our data, RE gradually increased until 25 min in the hepatobiliary phase in normal liver, but in advanced liver dysfunction, RE did not increase significantly, and the ratio of RE gradually increased over time, so we decided to use data that had been calculated using several methods both at 15 and 25 min in the hepatobiliary phase to compare its accuracy and usefulness. All MRI methods at 25 min had higher AUROC and higher correlation than at 15 min (data not shown). The best correlation of MRI and the liver biopsy was with LI at 25 min, which also had the highest AUROC for discriminating significant, extensive fibrosis and liver cirrhosis in the estimation set, followed by RE, with LS being the least accurate of these four methods. The LI score at 25 min was higher than the liver function tests that were reported previously, such as Forn's score, the FibroIndex, the Lok index, and APRI. Also in the validation set, LI at 25 min had the highest score of all these predictive methods.
SI in the hepatobiliary phase may be influenced by several factors, such as decreased hepatocytes, deficiency of hepatocyte function, and indocyanine green clearance. The uptake of gadoxetate disodium and its excretion are performed by the anion transporting polypeptide Oatp1 and Mrp2 mainly, and the balance of these effects may decide the SI of liver parenchyma followed by a decreased signal in hepatic damage and cirrhosis. Some previous reports have evaluated fibrosis using Gd-EOB-DTPA. Tsuda et al. reported the possibility of predicting fibrosis in non-alcoholic steatohepatitis (NASH) by using NASH model rats. Watanabe et al. reported that Gd-EOB-DTPA correlated with the fibrosis stage, but their report had a small number of patients, included several courses of hepatic injury, and used only the LMSI ratio. Many reports have used RE, and some have used the liver-to-spleen ratio; but, no reports have compared the accuracy of these methods. In chronic liver disease, especially in the advanced stage such as cirrhosis, many changes to the internal organ circulation occur, and in some cases, there may be inequality in the LMSI ratio and LSSI ratio. In this study, we tried to calculate the intensity ratio by using an internal control with intervertebral disc intensity because intervertebral discs have no enhancement after Gd-EOB-DTPA injection and compared all methods to decide the most useful method to evaluate hepatic fibrosis. LI method at 25 min had the highest AUROC in the prediction of significant or extensive fibrosis, so we recommend using LI and prefer to use the 25-min hepatobiliary data after injection. Detecting focal liver lesions 10 min after injection can be acceptable, but for the purpose of fibrotic evaluation, longer sampling, such as 25 min, is recommended.
Although LI has almost the same accuracy of fibrotic evaluation as using previous reported hematologic methods (no significant differences), we can obtain high accuracy with 0.79–0.86, and these values are higher than previous serum fibrosis markers; it may be an advantage to evaluate fibrosis and scanning HCC occurrence simultaneously and especially very useful in patients that are impossible to evaluate using the platelet count because of other reasons that have changed its value. The LI method can evaluate patients who are obese or have ascites, although these patients are not suitable to evaluate by Fibroscan, which is another non-invasive method reported previously. On the other hand, there are some problems with the LI25 method: (i) In general treatment, it is sometimes slightly difficult to find the best disk position because some patients have disk disease or spinal curvature; (ii) longer time is needed to obtain these pictures because 15 min is usually enough to detect a neoplasm in the liver; and (iii) MRI is more expensive than serum examination.
It is necessary to compare our methods to other devices such as MR elastography and transient elastography (Fibroscan), and prospective studies are needed to decide the most evaluable method. At the moment, we suggest that using a combination of hematologic biomarkers may lead to more accurate diagnosis without liver biopsy. Further, using EOB-MRI has a possibility of reflecting liver function accurately rather than its fibrosis, more detailed investigation is needed of liver function and EOB-MRI.