Background & Aims: The purpose of our study was to prospectively compare the success rate and diagnostic accuracy of magnetic resonance elastography, ultrasound elastography, and aspartate aminotransferase to platelets ratio index (APRI) measurements for the noninvasive staging of fibrosis in patients with chronic liver disease. Methods: We performed a prospective blind comparison of magnetic resonance elastography, ultrasound elastography, and APRI in a consecutive series of patients who underwent liver biopsy for chronic liver disease in a university-based hospital. Histopathologic staging of liver fibrosis according to the METAVIR scoring system served as the reference. Results: A total of 141 patients were assessed. The technical success rate of magnetic resonance elastography was higher than that of ultrasound elastography (133/141 [94%] vs 118/141 [84%]; P = .016). Magnetic and ultrasound elastography, APRI measurements, and histopathologic analysis of liver biopsy specimens were technically successful in 96 patients. The areas under the receiver operating characteristic curves of magnetic resonance elasticity (0.994 for F ≥ 2; 0.985 for F ≥ 3; 0.998 for F = 4) were larger (P < .05) than those of ultrasound elasticity, APRI, and the combination of ultrasound elasticity and APRI (0.837, 0.709, and 0.849 for F ≥ 2; 0.906, 0.816, and 0.936 for F ≥ 3; 0.930, 0.820, and 0.944 for F = 4, respectively). Conclusions: Magnetic resonance elastography has a higher technical success rate than ultrasound elastography and a better diagnostic accuracy than ultrasound elastography and APRI for staging liver fibrosis. Background & Aims: The purpose of our study was to prospectively compare the success rate and diagnostic accuracy of magnetic resonance elastography, ultrasound elastography, and aspartate aminotransferase to platelets ratio index (APRI) measurements for the noninvasive staging of fibrosis in patients with chronic liver disease. Methods: We performed a prospective blind comparison of magnetic resonance elastography, ultrasound elastography, and APRI in a consecutive series of patients who underwent liver biopsy for chronic liver disease in a university-based hospital. Histopathologic staging of liver fibrosis according to the METAVIR scoring system served as the reference. Results: A total of 141 patients were assessed. The technical success rate of magnetic resonance elastography was higher than that of ultrasound elastography (133/141 [94%] vs 118/141 [84%]; P = .016). Magnetic and ultrasound elastography, APRI measurements, and histopathologic analysis of liver biopsy specimens were technically successful in 96 patients. The areas under the receiver operating characteristic curves of magnetic resonance elasticity (0.994 for F ≥ 2; 0.985 for F ≥ 3; 0.998 for F = 4) were larger (P < .05) than those of ultrasound elasticity, APRI, and the combination of ultrasound elasticity and APRI (0.837, 0.709, and 0.849 for F ≥ 2; 0.906, 0.816, and 0.936 for F ≥ 3; 0.930, 0.820, and 0.944 for F = 4, respectively). Conclusions: Magnetic resonance elastography has a higher technical success rate than ultrasound elastography and a better diagnostic accuracy than ultrasound elastography and APRI for staging liver fibrosis. See Jacqueminet S et al on page 828 in CGH; See editorial on page 299. See Jacqueminet S et al on page 828 in CGH; See editorial on page 299. Liver biopsy is the current reference examination for the assessment of liver fibrosis. However, it is a costly procedure that carries a small risk of severe complications and is difficult to accept for patients. In addition, its accuracy remains debated because of sampling variability caused by the small size of the hepatic samples and the heterogeneity of liver fibrosis.1Bedossa P. Dargere D. Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C.Hepatology. 2003; 38: 1449-1457Crossref PubMed Scopus (1993) Google Scholar The noninvasive assessment of liver fibrosis has become a real challenge given that chronic liver diseases affect hundreds of millions of patients worldwide. Multiple data now emphasize that fibrosis is dynamic and, with effective intervention, reversible.2Rockey D.C. Bissell D.M. Noninvasive measures of liver fibrosis.Hepatology. 2006; 43: S113-S120Crossref PubMed Scopus (269) Google Scholar Successful treatment of viral hepatitis, autoimmune liver disease, alcohol-related disease, and other chronic liver diseases results not only in clinical improvement but also in decreased histologic fibrosis. Although experimental studies have revealed targets to prevent fibrosis progression in rodents, the efficacy of most treatments has not been proven in humans. The development of reliable noninvasive markers of liver fibrosis is essential to assess the prognosis of the disease and the response to treatment.2Rockey D.C. Bissell D.M. Noninvasive measures of liver fibrosis.Hepatology. 2006; 43: S113-S120Crossref PubMed Scopus (269) Google Scholar, 3Friedman S.L. Liver fibrosis—from bench to bedside.J Hepatol. 2003; 38: S38-S53Abstract Full Text Full Text PDF PubMed Google Scholar Several noninvasive methods have been proposed to stage liver fibrosis, including biochemical tests and imaging methods. The biochemical tests are composite scores (aspartate aminotransferase to platelets ratio index [APRI], FibroTest [BioPredictive, Paris, France], and so on) or serum markers of fibrosis such as hyaluronic acid. However, the value of these diagnostic methods remains debated.2Rockey D.C. Bissell D.M. Noninvasive measures of liver fibrosis.Hepatology. 2006; 43: S113-S120Crossref PubMed Scopus (269) Google Scholar, 3Friedman S.L. Liver fibrosis—from bench to bedside.J Hepatol. 2003; 38: S38-S53Abstract Full Text Full Text PDF PubMed Google Scholar Among the imaging methods, elastography has been shown to be a reliable method to stage liver fibrosis. It is based on the observation that fibrosis leads to increased tissue stiffness. Most clinical studies have been performed with ultrasound elastography.4Castera L. Vergniol J. Foucher J. et al.Prospective comparison of transient elastography, Fibrotest, APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C.Gastroenterology. 2005; 128: 343-350Abstract Full Text Full Text PDF PubMed Scopus (2080) Google Scholar, 5Sandrin L. Fourquet B. Hasquenoph J.M. et al.Transient elastography: a new noninvasive method for assessment of hepatic fibrosis.Ultrasound Med Biol. 2003; 29: 1705-1713Abstract Full Text Full Text PDF PubMed Scopus (2164) Google Scholar, 6Ziol M. Handra-Luca A. Kettaneh A. et al.Noninvasive assessment of liver fibrosis by measurement of stiffness in patients with chronic hepatitis C.Hepatology. 2005; 41: 48-54Crossref PubMed Scopus (1269) Google Scholar Recently, magnetic resonance (MR) elastography has emerged as an alternative method to assess liver elasticity.7Huwart L. Peeters F. Sinkus R. et al.Liver fibrosis: non-invasive assessment with MR elastography.NMR Biomed. 2006; 19: 173-179Crossref PubMed Scopus (356) Google Scholar, 8Huwart L. Sempoux C. Salameh N. et al.Liver fibrosis: noninvasive assessment with MR elastography versus aspartate aminotransferase-to-platelet ratio index.Radiology. 2007; 245: 458-466Crossref PubMed Scopus (319) Google Scholar, 9Yin M. Talwalkar J.A. Glaser K.J. et al.Assessment of hepatic fibrosis with magnetic resonance elastography.Clin Gastroenterol Hepatol. 2007; 5: 1207-1213Abstract Full Text Full Text PDF PubMed Scopus (744) Google Scholar To the best of our knowledge, no study has been reported about the comparison of MR and ultrasound elastography for the assessment of liver fibrosis. Therefore, the purpose of our study was to prospectively compare the technical success rate and diagnostic accuracy of MR elastography, ultrasound elastography, and APRI measurements for staging hepatic fibrosis in patients who underwent liver biopsy for chronic liver disease. This study was a single-center, prospective, blind comparison of MR elastography, ultrasound elastography, and APRI in a consecutive series of patients who underwent liver biopsy in the Department of Gastroenterology of St-Luc University Hospital, Université Catholique de Louvain, for suspicion of chronic diffuse liver disease between November 2005 and February 2007. The study protocol was in accordance with the Declaration of Helsinki and was approved by the ethics committee of our institution. Patients were enrolled after giving written informed consent. MR elastography, ultrasound elastography, and APRI measurements were performed within 2 days of liver biopsy. The order of the examinations was not randomized because it was considered that this order should not have influenced the results in patients with chronic liver diseases. For organizational reasons (access to the MR scanner), MR elastography was performed immediately before ultrasound elastography. A blood test, including measurements of aspartate aminotransferase levels and platelet counts, was systematically performed the day before the liver biopsy. To assess the reproducibility of MR and ultrasound elastography, these examinations were repeated within 1 month in all patients in whom MR elastography, ultrasound elastography, APRI measurements, and analysis of liver biopsy specimens were technically successful and who agreed to have repeated examinations. The 3-dimensional MR elastography method has been described in detail previously.7Huwart L. Peeters F. Sinkus R. et al.Liver fibrosis: non-invasive assessment with MR elastography.NMR Biomed. 2006; 19: 173-179Crossref PubMed Scopus (356) Google Scholar, 10Muthupillai R. Lomas D.J. Rossman P.J. et al.Magnetic resonance elastography by direct visualization of propagating acoustic strain waves.Science. 1995; 269: 1854-1857Crossref PubMed Scopus (1661) Google Scholar, 11Sinkus R. Tanter M. Catheline S. et al.Imaging anisotropic and viscous properties of breast tissue by magnetic resonance-elastography.Magn Reson Med. 2005; 53: 372-387Crossref PubMed Scopus (304) Google Scholar Briefly, low-frequency longitudinal mechanical waves of 65 Hz were transmitted into the right liver by a transducer placed against the last ribs at the back of the patient in supine position. The reconstruction of the local shear elasticity requires the presence of shear waves inside the organ. Shear waves were obtained by utilizing the fact that longitudinal waves generate shear waves due to mode conversion at interfaces everywhere inside the liver. The reason to excite in the first place with longitudinal waves is that these get less attenuated by tissue, leading to an efficient way for shear wave generation deep inside the body. Shear waves afterward become separated from the longitudinal contribution by applying the curl operator on the total displacement vector field. Images were obtained on a 1.5-T whole-body MR scanner (Gyroscan Intera; Philips Medical Systems, Best, The Netherlands) using a 4-elements torso coil. Five sagittal slices through the right liver were acquired with a slice thickness of 4 mm, field of view of 250 mm, matrix size of 642, echo time of 61 milliseconds, repetition time of 431 milliseconds, and 2 signal averages. The patients breathed freely, and respiratory gating was performed with a navigator on the right hemidiaphragm. Four dynamics were obtained by changing the phase offset between the mechanical excitation and the MR sequence to assess the amplitude and phase of the displacement after Fourier transformation. The motion-encoding gradients were applied successively in the 3 orthogonal directions to capture all the components of the 3-dimensional displacement vector. The total acquisition time was about 20 minutes, depending on the efficiency of the respiratory gating navigator. The phase images were analyzed with the Voigt model to obtain shear elasticity maps, that is, the real part of the complex shear modulus is attributed to the solid component of the material while the imaginary part accounts for losses. The shear elasticity (kPa) of the liver was measured as the mean value within the largest rectangular region of interest that fitted into the liver on the elasticity map of the central slice. The first series of measurements (ie, within 2 days of liver biopsy) was performed by a junior radiologist with 4 years of experience in MR imaging. The second series of measurements (within the month) was performed by a junior physicist with 2 years of experience in MR imaging. These 2 observers were blinded to the clinical, biochemical, and ultrasound data of the patient and to the results of histopathologic analysis. One-dimensional transient ultrasound elastography measurements were performed with a FibroScan (EchoSens, Paris, France). The technique and examination procedure have been described previously.5Sandrin L. Fourquet B. Hasquenoph J.M. et al.Transient elastography: a new noninvasive method for assessment of hepatic fibrosis.Ultrasound Med Biol. 2003; 29: 1705-1713Abstract Full Text Full Text PDF PubMed Scopus (2164) Google Scholar This method measures the velocity of the shear wave, which is directly related to Young's elastic modulus (kPa). It should be noted that, within tissues, the Young's modulus equals 3 times the shear elasticity modulus measured with the 3-dimensional MR elastographic method. Measurements were performed in the right lobe of the liver with the tip of the transducer probe placed in the intercostal space. The measurement depth was between 25 and 65 mm. Ten validated measurements were performed in each patient. The median value was considered representative of the Young's elastic modulus of the liver. The interquartile range of the 10 measurements was also recorded for each patient, and the ratio interquartile range/median value of liver stiffness was calculated. Ratios <0.2 indicate reliable measurements.12Lucidarme D. Foucher J. Le Bail B. et al.The ratio interquartile range/median value of liver stiffness measurement is a key factor of accuracy of transient elastography (FibroScan) for the diagnosis of liver fibrosis.Hepatology. 2007; 46: S181Google Scholar The whole examination lasted less than 5 minutes. The first and second series of measurements were performed by 2 different senior radiologists who had performed more than 100 examinations with the FibroScan. These 2 observers were blinded to the clinical, biochemical, and MR data of the patient and to the results of histopathologic analysis. Aspartate aminotransferase levels and platelet counts were measured with a Synchron Clinical System LX20 autoanalyzer (Beckman Coulter, Fullerton, CA) and an Advia 120 Hematology System autoanalyzer (Bayer, Leverkusen, Germany), respectively. The APRI index was calculated as follows: Aspartate Aminotransferase (/Upper Limit of Normal) × 100/Platelet Count (109/L). Percutaneous liver biopsy was performed by senior operators using the Menghini technique with a 1.4-mm-diameter needle (Hepafix; Braun, Melsungen, Germany). In 20 patients with ascites and/or trouble of blood crasis, liver biopsy was performed through a transjugular approach using a Ross-modified Colapinto catheter needle with a diameter of 1.5 mm (Cook, Bjaeverskov, Denmark). After biopsy, the liver samples were fixed in formalin, paraffin embedded, and stained with H&E and Masson's trichrome. All biopsy specimens were analyzed by 2 senior hepatopathologists blinded to the biological and clinical data and to the results of MR and ultrasound elastography. Liver biopsy specimens were considered suitable for fibrosis staging when they contained at least 10 portal tracts or obvious regenerating nodules. The stage of fibrosis was evaluated semiquantitatively on Masson's trichrome–stained slides according to the METAVIR scoring system.13Intraobserver and interobserver variations in liver biopsy interpretation in patients with chronic hepatitis C The French METAVIR Cooperative Study Group.Hepatology. 1994; 20: 15-20Crossref PubMed Scopus (1838) Google Scholar The METAVIR scoring system was initially described for chronic hepatitis C and was later applied to other chronic liver diseases.14Foucher J. Chanteloup E. Vergniol J. et al.Diagnosis of cirrhosis by transient elastography (FibroScan): a prospective study.Gut. 2006; 55: 403-408Crossref PubMed Scopus (1035) Google Scholar With this score, F0 represents no fibrosis; F1, portal fibrosis without septa; F2, portal fibrosis and few septa; F3, numerous septa without cirrhosis; and F4, cirrhosis. The fibrosis stage was assessed independently by each pathologist. In case of discrepancies, a consensus was obtained. However, in patients with nonalcoholic steatohepatitis and alcoholic liver disease, the use of the Brunt classification has been recommended because of the presence of a specific perisinusoidal fibrosis in these diseases.15Brunt E.M. Janney C.G. Di Bisceglie A.M. et al.Nonalcoholic steatohepatitis: a proposal for grading and staging the histological lesions.Am J Gastroenterol. 1999; 94: 2467-2474Crossref PubMed Scopus (3152) Google Scholar This scoring system differs from the METAVIR scoring system mainly for the staging of mild fibrosis. It is defined as follows: stage 0, no fibrosis; stage 1, perisinusoidal or periportal fibrosis; stage 2, perisinusoidal and periportal fibrosis; stage 3, bridging fibrosis; and stage 4, cirrhosis. The consequence is that in the case of nonalcoholic steatohepatitis and alcoholic liver disease, a METAVIR F0 score can already be a Brunt stage 1 and a F1 score can be a Brunt stage 2. To evaluate the possible influence of using a common staging system (METAVIR) for all patients in our study, a secondary and unblinded analysis for the presence of perisinusoidal fibrosis was performed in patients with nonalcoholic steatohepatitis or alcoholic liver disease who were staged F0 or F1 with the METAVIR scoring system. The grade of necroinflammatory activity was evaluated semiquantitatively on H&E-stained slides according to the Ludwig scoring system with its 2 main criteria: portal and lobular inflammation.16Ludwig J. The nomenclature of chronic active hepatitis: an obituary.Gastroenterology. 1993; 105: 274-278PubMed Google Scholar The portal inflammation was graded from 0 to 4, with P0 representing none or minimal inflammation; P1, portal inflammation; P2, mild limiting plate necrosis; P3, moderate limiting plate necrosis; and P4, severe limiting plate necrosis. The lobular inflammation was graded from 0 to 4, with L0 representing no inflammation; L1, inflammation but no necrosis; L2, focal necrosis or acidophilic bodies; L3, severe focal cell damage; and L4, damage including bridging necrosis. Steatosis was graded according to the Brunt scoring system as follows: 0, none; 1, steatosis in 1%–33% of hepatocytes; 2, steatosis in 33%–66% of hepatocytes; 3, steatosis in 66%–100% of hepatocytes.15Brunt E.M. Janney C.G. Di Bisceglie A.M. et al.Nonalcoholic steatohepatitis: a proposal for grading and staging the histological lesions.Am J Gastroenterol. 1999; 94: 2467-2474Crossref PubMed Scopus (3152) Google Scholar The length of each liver biopsy was established in millimeters, and the number of portal tracts was counted. The sample size was fixed to allow at least 80% power to detect by the McNemar's test at a 2-sided 5% significance level, a difference in sensitivity between MR elastography and ultrasound elastography of 0.10 considering a sensitivity of ultrasound elastography for F ≥ 2 close to 0.67 (similar to that found in the study by Castera et al4Castera L. Vergniol J. Foucher J. et al.Prospective comparison of transient elastography, Fibrotest, APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C.Gastroenterology. 2005; 128: 343-350Abstract Full Text Full Text PDF PubMed Scopus (2080) Google Scholar), and considering a prevalence of patients with F ≥ 2 around 60%. Under these conditions, a sample size of 95 patients evaluated by the 2 tests was required. The difference in technical success between MR elastography and ultrasound elastography was assessed with the χ2 test. Further analysis was performed in the patients in whom MR elastography, ultrasound elastography, APRI measurements, and histologic analysis of liver biopsy specimens were technically successful. The agreement between the pathologists for the scoring of liver fibrosis was assessed with weighted κ coefficients. The reproducibility of MR and ultrasound elastography was evaluated with intraclass correlation coefficients and coefficients of repeatability. The coefficients of repeatability were calculated as 1.96 times the standard deviations of the differences between the 2 measurements made by the 2 observers. Diagnostic performance of MR and ultrasound elastography was assessed on the first series of measurements obtained within 2 days of liver biopsy. The relationships between fibrosis stage and MR elasticity, ultrasound elasticity, and APRI measurements were tested with the nonparametric Kendal's coefficients of correlation. Magnetic resonance elastography, ultrasound elastography, and APRI were evaluated for their predictive performance using areas under the receiver operating characteristic (ROC) curves. Areas under ROC curves of MR elastography, ultrasound elastography, and APRI were compared using the method proposed by DeLong et al.17DeLong E.R. DeLong D.M. Clarke-Pearson D.L. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach.Biometrics. 1988; 44: 837-845Crossref PubMed Scopus (14326) Google Scholar To assess our results in a population of patients with moderate fibrosis, we performed a complementary sensitivity analysis comparing the areas under the ROC curves in the subpopulation of patients with F ≤ 2. In addition, MR elastography, ultrasound elastography, and APRI were introduced in a stepwise multivariate model to identify the best predictor(s) of fibrosis and the possible gain in prediction when several methods were used. A well-known problem of predictive models is the frequent overestimation of their performance because the models are evaluated on the samples used for their construction. Such a phenomenon called “optimism” should be assessed for appropriate choice and validation of the multivariate model. To assess the quality of the model (and thus the robustness of our conclusions), we used 10-fold cross-validation and bootstrapping procedure (N = 1000 bootstrap samples).18Steyerberg E.W. Harrell Jr, F.E. Borsboom G.J. et al.Internal validation of predictive models: efficiency of some procedures for logistic regression analysis.J Clin Epidemiol. 2001; 54: 774-781Abstract Full Text Full Text PDF PubMed Scopus (1802) Google Scholar Briefly, for K-fold cross-validation, the original sample was partitioned into K subsamples. Of the K subsamples, a single subsample was retained as the validation data for testing the model, and the remaining K − 1 subsamples were used as training data. The cross-validation process was then repeated several times, with each of the K subsamples used as the validation data. Bootstrapping (ie, based on repeatedly analyzing subsamples of the data) allows calculation of several parameters estimating the optimism of the model. As proposed by Steyerberg et al,18Steyerberg E.W. Harrell Jr, F.E. Borsboom G.J. et al.Internal validation of predictive models: efficiency of some procedures for logistic regression analysis.J Clin Epidemiol. 2001; 54: 774-781Abstract Full Text Full Text PDF PubMed Scopus (1802) Google Scholar bootstrapping was used to evaluate the quality of the model by calculation of calibration slopes (well-calibrated models have a slope equal or close to 1) and Brier scores (models scores range from 0 [perfect] to 0.25 [worthless]).18Steyerberg E.W. Harrell Jr, F.E. Borsboom G.J. et al.Internal validation of predictive models: efficiency of some procedures for logistic regression analysis.J Clin Epidemiol. 2001; 54: 774-781Abstract Full Text Full Text PDF PubMed Scopus (1802) Google Scholar Sensitivity, specificity, and positive and negative predictive values for the classification of F0 versus F1 − F4 (F ≥ 1), F0 − F1 versus F2 − F4 (F ≥ 2), F0 − F2 versus F3 − F4 (F ≥ 3), and F0 − F3 versus F4 (F = 4) were computed for MR elastography at the maximum total of sensitivity and specificity. In addition, the misclassification rates were given for each individual fibrosis stage. Moreover, the possible effects on the MR elasticity measurements of the type of liver biopsy (percutaneous vs transjugular), the length of the hepatic samples (with 3 groups: <15 mm, 15–25 mm, >25 mm), and the cause of liver disease (chronic viral hepatitis B and C vs alcoholic and nonalcoholic steatohepatitis) were studied using analysis of variance. Lastly, possible effects of steatosis or portal and/or lobular inflammation on the MR elasticity measurements were studied using 3-way analysis of variance on ranks.19Akritas M. Arnold S.F. Brunner E. Nonparametric hypothesis and rank statistics for unbalanced factorial designs.J Am Stat Assoc. 1997; 92: 258-265Crossref Scopus (170) Google Scholar Results are given as mean ± SD. Statistical analysis was performed with SAS 9.13 software (SAS Institute Inc, Cary, NC). A total of 146 patients underwent liver biopsy for chronic liver disease between November 2005 and February 2007. Five patients refused to participate, and 141 patients entered the study. Ultrasound elastography could be performed in 118 of 141 patients (84%). Ultrasound elastography was unsuccessful in 13 patients with ascites. Ascites was subclinical and detected with imaging in 5 of these patients. Ten other failures were caused by obesity. The body mass index of these patients was 32.8 ± 1.8 kg/m2 (29.8–35.1 kg/m2). Magnetic resonance elastography could be performed in 133 of 141 patients (94%) (P = .016 vs ultrasound elastography). The 8 failures were caused by claustrophobia in 3 patients, low hepatic signal related to hemochromatosis in 3 patients, and obesity in 2 patients. These 2 patients could not fit into the magnet bore, which was narrowed by the transducer. Measurements of APRI were obtained in 141 of 141 patients (100%). Liver biopsy specimens were suitable for fibrosis staging in 127 of 141 patients (90%). Further analysis was performed in the 96 patients in whom MR elastography, ultrasound elastography, APRI measurements, and histologic analysis were successful (Figure 1). There were 45 men and 51 women. Their age was 54 ± 13 years (range, 22–83 years), and their mean body mass index was 25.9 ± 4.0 kg/m2 (range, 18.7–33.1 kg/m2). Body mass index was >28 kg/m2 in 12 of 96 patients (12%). The cause of chronic liver disease was chronic viral hepatitis in 65 patients (chronic hepatitis C in 60 and chronic hepatitis B in 5), alcohol abuse in 14, nonalcoholic steatohepatitis in 8, α1-antitrypsin deficiency in 1, drug toxicity in 2, and unknown in 6. The length of the liver biopsy specimens was 30 ± 11 mm (range, 12–66 mm) and was ≥25 mm in 70 of 96 patients (73%). The number of portal tracts was 15 ± 6 (range, 6–33) and was ≥11 in 87 of 96 patients (91%). The distribution of fibrosis stage at the consensus reading was F0 in 22 of 96 patients (23%), F1 in 22 (23%), F2 in 19 (20%), F3 in 15 (15%), and F4 in 18 (19%), showing a homogeneous distribution of fibrosis stages. The 2 pathologists were initially in agreement for 81 of the 96 liver biopsy specimens analyzed (weighted κ coefficient, 0.90; 95% confidence interval [CI], 0.84–0.95) with no significant rating bias. Among the patients with alcoholic or nonalcoholic steatohepatitis, one who was staged F1 according to METAVIR had portal and perisinusoidal fibrosis and would have been classified as stage 2 according to Brunt. The elasticity was 2.49 ± 0.17 kPa in this patient. Portal inflammation was graded P0 in 17 of 96 patients (17%), P1 in 43 (45%), P2 in 35 (37%), P3 in 1 (1%), and P4 in 0 (0%). Lobular inflammation was graded L0 in 32 of 96 patients (33%), L1 in 34 (35%), L2 in 30 (32%), L3 in 0 (0%), and L4 in 0 (0%). Steatosis was graded 0 in 40 of 96 patients (42%), 1 in 32 (34%), 2 in 18 (19%), and 3 in 6 (5%). At ultrasound elastography, the ratio interquartile range/median value was 0.16 for F0, 0.17 for F1, 0.16 for F2, 0.19 for F3, and 0.19 for F4. The reproducibility of MR and ultrasound elastography could be assessed in 56 patients. The intraclass correlation coefficients of MR elasticity and ultrasound elasticity were 0.97 (95% CI, 0.92–0.99) and 0.94 (95% CI, 0.51–0.97), respectively. The 95% CIs of the interobserver differences of measurements were −0.016 to 0.088 for MR elasticity and 0.028 to 2.192 for ultrasound elasticity. The coefficients of repeatability were 0.385 for MR elastography and 8.149 for ultrasound elastography. In the 96 patients in whom elastography, APRI measurements, and histologic analysis were successful, MR elasticity, ultrasound elasticity, and APRI increased according to the stage of liver fibrosis (Figure 2). The MR elasticity was 2.15 ± 0.19 kPa (range, 1.89–2.43 kPa) for F0, 2.41 ± 0.11 kPa (range, 2.19–2.63 kPa) for F1, 2.85 ± 0.23 kPa (range, 2.49–3.18 kPa) for F2, 3.49 ± 0.53 kPa (range, 2.84–4.46 kPa) for F3, and 5.25 ± 0.65 kPa (range, 4.13–6.73 kPa) for F4. The measurements were correlated to the fibrosis stage: r = 0.84, P < .0001 for MR elasticity; r = 0.56, P < .0001 for ultrasound elasticity; and r = 0.36, P < .0001 for APRI. Table 1 shows the areas under the ROC curves of MR elasticity, ultrasound elasticity, and APRI and the combinations of the noninvasive methods for the different fibrosis thresholds. The areas under the ROC of MR elastography were significantly larger than those of ultrasound elastography, APRI, and the combination of ultrasound elastography and APRI (P = .003, P < .0001, and P = .005 for F