Liver Fibrosis Noninvasive Assessment with MR Elastography versus

Document Sample
Liver Fibrosis Noninvasive Assessment with MR Elastography versus Powered By Docstoc
					                                            Note: This copy is for your personal non-commercial use only. To order presentation-ready
                                            copies for distribution to your colleagues or clients, contact us at

                                                                                         Liver Fibrosis: Noninvasive
                                                                                         Assessment with MR Elastography
                                                                                         versus Aspartate Aminotransferase–
                                                                                         to-Platelet Ratio Index1
                           Laurent Huwart, MD

                                                                                              Purpose:    To prospectively compare the sensitivity and specificity of
                           Christine Sempoux, MD, PhD
                                                                                                          magnetic resonance (MR) elastography with those of the
                           Najat Salameh, MSc
                                                                                                          routinely available aspartate aminotransferase–to-platelet
                           Jacques Jamart, MD                                                             ratio index (APRI) test for staging hepatic fibrosis in pa-
                           Laurence Annet, MD, PhD                                                        tients who have undergone liver biopsy for suspicion of
                           Ralph Sinkus, PhD                                                              chronic liver disease, with histopathologic examination as
                           Frank Peeters, PhD                                                             the reference standard.
                           Leon C. ter Beek, PhD
                           Yves Horsmans, MD, PhD                                         Materials and   The study was approved by the ethics committee. All pa-
                           Bernard E. Van Beers, MD, PhD                                      Methods:    tients gave written informed consent. Eighty-eight patients
                                                                                                          (37 men, 51 women; mean age, 54.0 years         13.1 [stan-
                                                                                                          dard deviation]) who underwent liver biopsy for suspicion
                                                                                                          of chronic liver disease underwent MR elastography and
                                                                                                          APRI testing within 2 days after liver biopsy. At his-
                                                                                                          topathologic examination, the fibrosis stage was assessed
                                                                                                          according to METAVIR scores (fibrosis scores F0 [no fi-
                                                                                                          brosis] to F4 [cirrhosis]). MR elastography was performed
                                                                                                          by transmitting mechanical waves within the liver and
                                                                                                          measuring the small cyclic displacement of the liver spins
                                                                                                          with a phase-contrast spin-echo sequence. The perfor-
                                                                                                          mances of MR elastography and APRI testing were as-
                                                                                                          sessed, and the optimal cutoff values for fibrosis stage
                                                                                                          were determined with receiver operating characteristic
                                                                                                          (ROC) curve analysis.

                                                                                               Results:   At MR elastography, areas under the ROC curves (Az) for
                                                                                                          elasticity and viscosity, respectively, were 0.999 and 0.863
                                                                                                          at fibrosis scores greater than or equal to F2, 0.997 and
                                                                                                          0.962 at scores greater than or equal to F3, and 1.000 and
                                                                                                          0.986 at score F4. Az values for elasticity at MR were
                                                                                                          significantly larger than those for the APRI (0.854 at
                                                                                                          scores F2, P .001; 0.886 at scores F3, P .003; and
                                                                                                          0.851 at score F4, P        .004). Optimal cutoff values of
                                                                                                          elasticity were 2.5 kPa for fibrosis scores greater than or
                             From the Diagnostic Radiology Unit (L.H., N.S., L.A.,
                                                                                                          equal to F2, 3.1 kPa for scores greater than or equal to F3,
                           F.P., B.E.V.B.), Department of Pathology (C.S.), and Labo-
                           ratory of Gastroenterology (Y.H.), Universite Catholique de
                                                                       ´                                  and 4.3 kPa for score F4.
                           Louvain, St-Luc University Hospital, Avenue Hippocrate
                           10, B-1200 Brussels, Belgium; Center of Biostatistics and        Conclusion:   Large Az values for elasticity ( 0.990 for scores     F2,
                           Medical Documentation, Universite Catholique de Louvain,                          F3, and F4) show that MR elastography was accurate in
                           Mont-Godinne University Hospital, Yvoir, Belgium (J.J.);                       liver fibrosis staging and superior to biochemical testing
                           Laboratoire Ondes et Acoustique, Universite Paris 7,
                                                                                                          with APRIs.
                           Paris, France (R.S.); and Philips Medical Systems, Best,
                           the Netherlands (L.C.t.B.). Received September 9, 2006;
                           revision requested December 5; revision received January                        RSNA, 2007
                           10, 2007; accepted January 29; final version accepted
                           April 3. Supported by grants FRSM 3.4578.00 and
                           3.4580.06 from the Fonds National de la Recherche Sci-
                           entifique, Belgium. Address correspondence to L.H.

                               RSNA, 2007

                           458                                                                                                       Radiology: Volume 245: Number 2—November 2007
GASTROINTESTINAL IMAGING: MR Elastography of Liver Fibrosis                                                                                    Huwart et al

       ll chronic liver diseases may lead       rhosis need follow-up for detection of       which is triggered by the MR spectrom-
       to liver fibrosis. An abundance           esophageal varices and hepatocellular        eter. The time-dependent magnetic mo-
       of data now emphasize that fi-            carcinomas (9–11). Several noninvasive       ment of the coil couples to the static
brosis is evolving and, with effective          methods for staging liver fibrosis have       magnetic field, and the resulting torque
intervention, reversible (1–4). The             been proposed. These methods include         leads to the longitudinal oscillations of
stage of liver fibrosis is used to deter-        the use of biochemical scores, such          the piston. Low-frequency (65-Hz) lon-
mine the prognosis and the treatment            as the aspartate aminotransferase–to-        gitudinal mechanical waves are trans-
options. Although liver biopsy is the           platelet ratio index (APRI), and liver im-   mitted into the right liver lobe by the
current reference standard for deter-           aging techniques (12–16). However, the       piston, which is placed against the last
mining fibrosis stage, it is invasive and        value of these diagnostic methods, espe-     ribs in the back of the patient in the
is not well accepted by patients, espe-         cially for the diagnosis of intermediate     supine position.
cially when repeated examinations are           fibrosis, remains debatable (2).
needed. In addition, the accuracy of                 Magnetic resonance (MR) elastog-        Patients
liver biopsy in the assessment of fibro-         raphy is a noninvasive method of mea-        Ninety-six consecutive patients who
sis may be questioned because of sam-           suring the viscoelastic properties of the    had undergone liver biopsy in the gas-
pling variability and interobserver varia-      liver. Preliminary reports suggest that      troenterology department of St-Luc
tion in the interpretation of semiquantita-     MR elastography is a feasible method         University Hospital between Septem-
tive fibrosis scores (5–8).                      for staging liver fibrosis (17,18). Thus,     ber 2004 and April 2006 because they
     Thus, there is a clear need for non-       the purpose of our study was to pro-         were suspected of having chronic liver
invasive alternatives to liver biopsy. Ide-     spectively compare the sensitivity and       disease were prospectively included in
ally, these noninvasive tests would en-         specificity of MR elastography with           the study. The study protocol was ap-
able one to reliably distinguish, at mini-      those of the routinely available APRI        proved by the ethics committee of this
mum, three stages of fibrosis: no or             test for staging hepatic fibrosis in pa-      institution. Patients were enrolled af-
early fibrosis, intermediate fibrosis, and        tients who have undergone liver biopsy       ter giving written informed consent.
advanced fibrosis or cirrhosis (2). Many         for suspicion of chronic liver disease,      MR elastography and APRI measure-
hepatologists believe that among pa-            with histopathologic analysis as the ref-    ment were performed within 2 days
tients with hepatitis B and C, as among         erence standard.                             after liver biopsy. After study inclusion,
patients with nonalcoholic liver disease,                                                    three patients were removed from the
those with at least intermediate fibrosis                                                     investigation because of claustrophobia
should be treated (2,4). With early dis-         Materials and Methods                       during MR elastography and five were
ease, the toxicity and costs of treatment                                                    removed because their liver biopsy find-
may outweigh the benefits. In addition,          Experimental Setup                           ings were unsuitable for fibrosis staging.
patients with advanced fibrosis or cir-                                        ´
                                                The authors from the Universite Catholique   Thus, the final study group comprised
                                                de Louvain had sole control of the data      88 patients (37 men, 51 women; mean
                                                generated in this study. The elasto-
 Advances in Knowledge                          graphic method used in this study has
    Patients with no (stage F0), mini-          been described in detail previously (18–     Published online
    mal (stage F1), intermediate                20): Briefly, the transducer contains a       10.1148/radiol.2452061673
    (stage F2), or advanced (stage              piston with a diameter of 6.5 cm and a
                                                                                             Radiology 2007; 245:458 – 466
    F3) fibrosis and patients with cir-          coil driven by the pulse generator,
    rhosis (stage F4) can be differen-                                                       Abbreviations:
    tiated with MR elastography.                                                             APRI aspartate aminotransferase–to-platelet ratio index
    The diagnostic accuracy of MR                Implications for Patient Care               Az area under the ROC curve
                                                                                             ROC receiver operating characteristic
    elastography for staging liver fi-              The high accuracy of MR elastog-
                                                                                             ROI region of interest
    brosis is higher than that of bio-             raphy suggests that this noninva-
    chemical testing with the aspar-               sive method has the potential to          Author contributions:
    tate aminotransferase–to-platelet              replace liver biopsy for the diag-        Guarantors of integrity of entire study, L.H., C.S., B.E.V.B.;
                                                                                             study concepts/study design or data acquisition or data
    ratio index (APRI). In our study,              nosis of liver fibrosis.
                                                                                             analysis/interpretation, all authors; manuscript drafting or
    areas under the receiver operat-               More particularly, MR elastogra-          manuscript revision for important intellectual content, all
    ing characteristic curve for MR                phy might be useful in the selec-         authors; manuscript final version approval, all authors;
    elasticity and APRI, respectively,             tion of patients with liver fibrosis       literature research, L.H., F.P., L.C.t.B., B.E.V.B.; clinical
    were 0.999 and 0.854 for fibrosis               who should either be treated              studies, L.H., C.S., N.S., L.A., Y.H., B.E.V.B.; experimental
    scores greater than or equal to                (score F2) or undergo surveil-            studies, R.S., F.P., L.C.t.B.; statistical analysis, J.J., R.S.,
    F2, 0.997 and 0.886 for scores                 lance for portal hypertension and         F.P.; and manuscript editing, L.H., R.S., F.P., L.C.t.B.,
    greater than or equal to F3, and               hepatocellular carcinoma
    1.000 and 0.851 for score F4.                  (score F3).                               Authors stated no financial relationship to disclose.

Radiology: Volume 245: Number 2—November 2007                                                                                                           459
GASTROINTESTINAL IMAGING: MR Elastography of Liver Fibrosis                                                                        Huwart et al

age, 54.0 years 13.1 [standard devia-          ing the Voigt model to obtain shear elas-    APRI Measurement
tion]; mean body mass index, 25.9              ticity and viscosity maps. The described     Aspartate aminotransferase level (AST,
kg/m2 4.0). Ascites was diagnosed at           MR elastographic method has been vali-       in international units per liter, divided
imaging in 13 patients. The cause of           dated previously, and the reproducibil-      by upper limit of normal) (22) and
chronic liver disease was viral in 69 pa-      ity of this technique has been assessed      platelet count (PLC, in 109 cells per li-
tients (chronic hepatitis C in 66, chronic     in healthy volunteers (18–20).               ter) were measured, respectively, with
hepatitis B in three), alcohol abuse in             For each patient, a junior radiolo-     Synchron Clinical System LX20 (Beck-
10, autoimmune disease in two (autoim-         gist (L.H.) with 3 years experience in       man-Coulter, Fullerton, Calif) and Ad-
mune hepatitis in one, primary biliary         MR imaging placed the largest rectan-        via 120 Hematology System (Bayer,
cirrhosis in one), 1-antitrypsin defi-          gular region of interest (ROI) that fit in    Leverkusen, Germany) autoanalyzers.
ciency in one, and unclassified in six.         the liver on the central section. This       The APRI was calculated as follows:
                                               observer was blinded to the patients’        (AST 100)/PLC (23).
MR Elastography                                clinical and biochemical data and his-
Images were obtained with a 1.5-T whole-       topathologic results. The shear elasticity   Histopathologic Analysis
body MR unit (Gyroscan Intera; Philips         (in kilopascals) and viscosity (in pascals   In 73 patients, percutaneous liver bi-
Medical Systems, Best, the Netherlands)        times seconds) of the liver were mea-        opsy was performed by two senior
by using a four-element torso coil. The        sured as the mean values within the          hepatologists (including Y.H., 20 years
MR elastography sequence was a mo-             large ROI on the elasticity and viscosity    experience) by using the Menghini tech-
tion-sensitized spin-echo sequence with        maps. For each patient, the intrasubject     nique with a 1.4-mm diameter needle
sinusoidal displacement-encoding gradi-        heterogeneities of the liver elasticity      (Hepafix; Braun, Melsungen, Germany).
ents that were synchronized to the me-         and viscosity were measured as the           In 15 patients with ascites and/or
chanical excitation by using a trigger on      standard deviations of the mean mea-         blood coagulation difficulty, liver bi-
the MR spectrometer. Five sagittal sec-        surements within the large ROI on the        opsy was performed with a transjugu-
tions through the right liver were ac-         elasticity and viscosity maps. The inter-    lar approach by using a catheter nee-
quired with 431/61 (repetition time            subject heterogeneities of the liver elas-   dle (Cook, Bjaeverskov, Denmark).
msec/echo time msec), a 4-mm section           ticity and viscosity were measured as             After biopsy, the liver samples were
thickness, a 250-mm field of view, a            the standard deviations of the mean          fixed in formalin for 24 hours, embed-
64     64 matrix (2), two acquired sig-        elasticity and viscosity measurements at     ded in paraffin, and stained with hema-
nals, and respiratory gating with a navi-      each fibrosis stage. It should be noted       toxylin-eosin and Masson trichrome. All
gator in the right hemidiaphragm. The          that the shear elasticity modulus mea-       biopsy specimens were analyzed by
sections were placed distant from the          sured with this three-dimensional MR         a senior hepatopathologist (C.S., 15 years
hepatic hilum to avoid inclusion of large      elastographic method differs from the        experience) who was blinded to the MR
hepatic vessels in the field of view. Four      Young modulus reported in studies of         elastography results and the biologic and
time points were obtained by changing          one-dimensional transient ultrasono-         clinical data. The fibrosis stage was evalu-
the phase offset between the mechani-          graphic (US) elastography (Fibroscan;        ated semiquantitatively according to the
cal excitation and the MR sequence to          EchoSens, Paris, France) by a scaling        METAVIR scoring system (24,25). The
assess the amplitude and phase of the          factor of three: The Young modulus           METAVIR scoring system was originally
displacement in one direction. The mo-         equals three times the shear elasticity      used to stage fibrosis in patients with
tion-encoding gradients were applied           modulus in soft tissues (21).                chronic hepatitis C, but it has also been
successively in three orthogonal direc-
tions to capture all components of the
three-dimensional displacement vector.
The total acquisition time was about 20
                                                 Figure 1
minutes, depending on the efficiency of                                                                     Figure 1: Flow diagram of
the respiratory-gating navigator.                                                                          patients suspected of having
    Longitudinal mechanical waves                                                                          chronic liver disease and who
were used for excitation because they                                                                      underwent liver biopsy during the
are less attenuated than shear waves and                                                                   20-month study period.
therefore have good penetration through-
out the liver (17,18). The shear waves
were generated by means of mode con-
version at interfaces and were sepa-
rated from the longitudinal contribution
by applying the curl operator on the
total displacement vector field. The
phase images were then analyzed by us-

460                                                                                              Radiology: Volume 245: Number 2—November 2007
GASTROINTESTINAL IMAGING: MR Elastography of Liver Fibrosis                                                                                    Huwart et al

applied in patients with other chronic          then compared by using the Hotelling test         Results
liver diseases (26,27). METAVIR scores          for correlated correlations (28). The per-
range from F0 to F4: F0 means no fibro-          formances of the various parameters in
sis; F1, portal fibrosis without septa; F2,      discriminating the fibrosis stages were         Fibrosis Staging
portal fibrosis and a few septa; F3, nu-         studied by using nonparametric receiver        Performing MR elastography and liver
merous septa without cirrhosis; and F4,         operating characteristic (ROC) curves.         biopsy led to no objective adverse
cirrhosis. The liver biopsy specimens had       Cutoff values of elasticity were chosen by     events in the examined population.
to contain at least 10 portal tracts or obvi-   maximizing the Youden index on the esti-       The mean length of the liver biopsy
ous regenerating nodules to be included         mated curves. Sensitivity, specificity, and     specimens was 34.5 mm 10.5 (stan-
in the analysis. The length of each liver       positive and negative predictive values        dard deviation). Twenty-two (25%) of
biopsy specimen was measured in milli-          were computed with exact 95% confi-             the 88 patients had a METAVIR score
meters.                                         dence intervals based on F distribution        of F0; 13 (15%), a score of F1; 15
                                                (29). All tests were two tailed. Because       (17%), a score of F2; 14 (16%), a
Statistical Analyses                            10 comparisons were performed, a               score of F3; and 24 (27%), a score of F4
Pearson coefficients were used to assess         Bonferroni adjusted         value of .005      (Fig 1). Elasticity, viscosity, intrasubject
the correlations between viscoelastic pa-       (.05/10) was used to preserve the              heterogeneity, and APRI increased with
rameter and METAVIR score (expressed            overall .05 error rate. The analyses           increasing liver fibrosis stage (Figs 2, 3).
as the nontransformed METAVIR score             were performed by using SC (Lambda-            Elasticity (r 0.86, P .001), viscosity
and the exponential function of the             Plus, Gembloux, Belgium) and SPSS              (r    0.75, P     .001), intrasubject het-
METAVIR score) and between APRI and             (SPSS, Chicago, Ill) statistical soft-         erogeneity of elasticity (r       0.70, P
METAVIR score. These coefficients were           ware.                                          .001), and intrasubject heterogeneity of

 Figure 2

                                                                                             Figure 2: Box plots of (a) elasticity, (b) intrasubject
                                                                                             heterogeneity of elasticity, (c) viscosity, (d) intrasubject
                                                                                             heterogeneity of viscosity, and (e) APRI for each METAVIR
                                                                                             fibrosis stage. Boundary of boxes closest to zero indicates
                                                                                             25th percentile, line within boxes indicates median, and
                                                                                             boundary of boxes farthest from zero indicates 75th per-
                                                                                             centile. Error bars indicate smallest and largest values
                                                                                             within 1.5 box lengths of 25th and 75th percentiles. Outli-
                                                                                             ers are represented as individual points. In e, one outlier is
                                                                                             not represented in the F4 group to maintain the clarity of
                                                                                             the graph.

Radiology: Volume 245: Number 2—November 2007                                                                                                           461
GASTROINTESTINAL IMAGING: MR Elastography of Liver Fibrosis                                                                                               Huwart et al

 Figure 3

 Figure 3: MR elastography images and reconstructed MR images (431/61) in (a– d) 36-year-old man with F2 disease and (e– h) 59-year-old woman with F4
 disease. (a, e) Magnitude MR elastography images show the largest rectangular ROI within the liver. (b, f) Corresponding reconstructed (postprocessed) phase
 images (after application of the curl operator) show good penetration of the shear waves. Reconstructed (c) elasticity and (d) viscosity maps in patient with stage
 F2 fibrosis are relatively homogeneous: The mean elasticity and viscosity measured within the ROI on these sections are 2.9 kPa 0.7 and 2.3 Pa sec 1.5,
 respectively. Reconstructed (g) elasticity and (h) viscosity maps in patient with stage F4 disease are heterogeneous: Mean elasticity and viscosity measurements
 are 6.8 kPa 2.4 and 7.0 Pa sec 4.1, respectively.

viscosity (r 0.68, P .001) measure-                     greater than those computed with the                     tion between APRI and nontransformed
ments correlated with fibrosis stage.                    nontransformed METAVIR score (P                          METAVIR score (P .05).
Higher correlations were observed                       .002).
between the viscoelastic parameters                                                                              ROC Analysis
and the exponential function of the                     APRI Measurement                                         For each METAVIR fibrosis score thresh-
METAVIR score: r 0.94 and P .001                        APRI measurement was only weakly                         old, the area under the ROC curve (Az)
for elasticity, r 0.85 and P .001 for                   correlated to METAVIR score (r 0.42                      for elasticity was larger than that for the
viscosity, r     0.79 and P     .001 for                and P       .001 for linear correlation                  other measurements (Fig 4, Table 1). Az
intrasubject heterogeneity of elasticity,               between APRI and nontransformed                          values for elasticity were significantly
and r 0.77 and P .001 for intrasub-                     METAVIR score, r 0.48 and P .001                         larger than those for the APRI at all
ject heterogeneity of viscosity. All cor-               for correlation between APRI and expo-                   fibrosis score thresholds (P       .003 for
relation coefficients computed with the                  nential function of METAVIR score).                        F1, P      .001 for F2, P       .003 for
exponential function of the METAVIR                     The coefficient for the correlation be-                     F3, P      .004 for F4) and than those
score, except that for the heterogeneity                tween APRI and exponential function of                   for intrasubject heterogeneity of elastic-
of viscosity (P .006, greater than Bon-                 the METAVIR score was not signifi-                        ity, viscosity, and intrasubject heteroge-
ferroni      of .005), were significantly                cantly greater than that for the correla-                neity of viscosity at fibrosis scores

462                                                                                                                     Radiology: Volume 245: Number 2—November 2007
GASTROINTESTINAL IMAGING: MR Elastography of Liver Fibrosis                                                                                          Huwart et al

greater than or equal to F1 (P   .001,                         ues of elasticity were 2.4 kPa for                 4.3 kPa for score F4. Corresponding
P    .004, and P    .001, respectively)                        METAVIR fibrosis scores greater than                sensitivities, specificities, and positive
and greater than or equal to F2 (P                             or equal to F1, 2.5 kPa for scores                 and negative predictive values are given
.001).                                                         greater than or equal to F2, 3.1 kPa for           in Table 2.
   The most discriminating cutoff val-                         scores greater than or equal to F3, and

 Figure 4                                                                                                          Discussion
                                                                                                                  The results of our study show that MR
                                                                                                                  elastography is an accurate noninvasive
                                                                                                                  method of staging liver fibrosis and is
                                                                                                                  superior to biochemical testing with
                                                                                                                  APRI values. The current reference
                                                                                                                  standard for the diagnosis and staging of
                                                                                                                  liver fibrosis is histopathologic analysis
                                                                                                                  of biopsy samples (5). Although his-
                                                                                                                  topathologic analysis is limited because
                                                                                                                  it is invasive and owing to a potential
                                                                                                                  lack of reproducibility due to the het-
                                                                                                                  erogeneity of liver fibrosis, the small
                                                                                                                  size of hepatic samples (30–32), and the
                                                                                                                  inter- and intraobserver variability in
                                                                                                                  histopathologic examinations (33–35),
                                                                                                                  it remains the only validated and univer-
                                                                                                                  sally accepted clinical method that en-
                                                                                                                  ables differentiation of early or no fibro-
                                                                                                                  sis (F0 –F1), intermediate fibrosis (F2),
                                                                                                                  and advanced fibrosis or cirrhosis (F3–
                                                                                                                  F4), which is essential to the decision to
                                                                                                                  treat (2,36).
                                                                                                                       In our study, the elasticity measure-
                                                                                                                  ments enabled us to clearly separate the
                                                                                                                  intermediate fibrosis stages and to more
                                                                                                                  precisely differentiate between stage F2
                                                                                                                  and stages F0 –F1. With an optimized
                                                                                                                  cutoff value of 2.5 kPa for stages greater
                                                                                                                  than or equal to F2, the sensitivity of
                                                                                                                  elasticity was 98% at a specificity of
                                                                                                                  100%. This high accuracy is clinically
 Figure 4: ROC curves for elasticity (green), intrasubject heterogeneity of elasticity (blue), viscosity          important, because according to the
 (pink), intrasubject heterogeneity of viscosity (turquoise), and APRI (red) at METAVIR fibrosis score             American Association for the Study of
 thresholds of (a) F1, (b) F2, (c) F3, and (d) F4.                                                                Liver Diseases, patients with hepatitis C
                                                                                                                  genotype 1 infection should be treated

 Table 1

   Az Values for Elasticity, Viscosity, Intrasubject Heterogeneity of Elasticity and Viscosity, and APRI according to METAVIR Score
   Parameter                                    F1                             F2                            F3                         F4

    Elasticity                               0.955 (0.887, 0.987)            0.999 (0.956, 1.000)          0.997 (0.952, 1.000)         1.000 (0.958, 1.000)
    Heterogeneity of elasticity              0.822 (0.724, 0.896)            0.889 (0.803, 0.946)          0.969 (0.907, 0.994)         0.969 (0.907, 0.994)
    Viscosity                                0.844 (0.750, 0.913)            0.863 (0.771, 0.927)          0.962 (0.898, 0.991)         0.986 (0.932, 0.998)
    Heterogeneity of viscosity               0.810 (0.711, 0.886)            0.834 (0.738, 0.905)          0.945 (0.873, 0.982)         0.965 (0.901, 0.992)
    APRI                                     0.836 (0.740, 0.907)            0.854 (0.761, 0.921)          0.886 (0.799, 0.944)         0.851 (0.758, 0.919)

   Note.—Numbers in parentheses are 95% confidence intervals.

Radiology: Volume 245: Number 2—November 2007                                                                                                                  463
GASTROINTESTINAL IMAGING: MR Elastography of Liver Fibrosis                                                                                           Huwart et al

                                                 Table 2
only when substantial fibrosis ( F2) is
observed (4,37).                                  Most Discriminant Elasticity Cutoff Values at Different METAVIR Scores, with
     In our study, advanced fibrosis and           Corresponding Sensitivities, Specificities, and Positive and Negative Predictive Values
cirrhosis were also diagnosed accu-
                                                  Parameter                                F1                   F2               F3             F4
rately. The cutoff value of 3.1 kPa for
fibrosis stages greater than or equal to            Elasticity cutoff (kPa)               2.4                  2.5              3.1              4.3
F3 had a sensitivity of 95% and a speci-              Sensitivity (%)                     83 (72, 91)          98 (90, 100)     95 (82, 99)     100 (86, 100)
ficity of 100%. This high accuracy in the              Specificity (%)                     100 (85, 100)        100 (90, 100)    100 (93, 100)    100 (94, 100)
diagnosis of advanced fibrosis is also im-             Positive predictive value (%)      100 (94, 100)        100 (93, 100)    100 (90, 100)    100 (86, 100)
portant because patients with advanced                Negative predictive value (%)       67 (48, 92)          97 (85, 100)     96 (87, 100)    100 (94, 100)
fibrosis should be screened for portal             Note.—Numbers in parentheses are 95% confidence intervals.
hypertension and hepatocellular carci-
noma (9–11). Correlation coefficients
for the relationship between viscoelastic      within the liver (31). In particular, it                       only two stages of the fibrosis spec-
parameter and exponential function of          should be noted that the intrasubject                          trum—namely, minimal and advanced
the METAVIR score were significantly            heterogeneity was much higher with cir-                        fibrosis. These scoring systems are less
better than those for the correlation          rhosis than with the other fibrosis                             effective for differentiating intermedi-
between viscoelastic parameter and             stages (38). However, the clinical rele-                       ate fibrosis stages. Moreover, the main
nontransformed METAVIR score (P                vance of the intrasubject heterogene-                          drawback of these tests is that some
.002). This observation reflected the           ities of elasticity and viscosity appeared                     parameters can be influenced by extra-
exponential accumulation of fibrous tis-        to be limited because the mean elastic-                        hepatic diseases.
sue within the liver, as seen by Bedossa       ity measurements alone had high accu-                              Several liver imaging methods have
et al (31).                                    racy.                                                          also been used to diagnose and stage
     The intersubject heterogeneity of              Several other noninvasive alterna-                        liver fibrosis. Most of these methods,
elasticity for cirrhosis (stage F4) was        tives to biopsy have been proposed for                         including double contrast material– en-
large. This may be explained by the            staging liver fibrosis. Serum tests in-                         hanced MR imaging, perfusion MR
variable macroscopic patterns of cirrho-       clude measurement of the doses of spe-                         imaging, and diffusion MR imaging, are
sis, which may appear as micronodular,         cific markers of fibrosis, such as hyal-                         limited to the detection of advanced
macronodular, or incomplete septal cir-        uronic acid and N-terminal collagen III                        fibrosis (12,13,45). The effectiveness
rhosis (38). Moreover, the intersubject        propeptide, which are products of the                          of these methods in the detection of
heterogeneity of elasticity in cirrhosis       degradation or synthesis of the extracel-                      substantial fibrosis ( F2) remains un-
may be related to the severity of the          lular matrix (40,41). The usefulness of                        proved.
disease, as shown in studies involving         these markers is limited because fibro-                             Transient US elastography has been
US elastography (26,39).                       sis is not specific to the liver. The use of                    proposed as a more direct method of as-
     In our study, the viscosity mea-          more comprehensive biomarkers based                            sessing liver fibrosis (14–16,26,46,47). In
surements were less accurate com-              on proteome or glycome fingerprints                             patients with chronic hepatitis C, re-
pared with the elasticity measure-             has been proposed (42,43). However,                            ported Az values for transient US elastog-
ments in the staging of liver fibrosis, as      these methods are costly and are not                           raphy are 0.79 – 0.83 for substantial fibro-
shown by the correlation coefficients,          readily available.                                             sis ( F2) and 0.95– 0.97 for cirrhosis
box plots, and Az values—particularly               Scoring systems such as the Fi-                           (F4). These results appear to be some-
for stages greater than or equal to F1         broTest (Biopredictive, Paris, France)                         what lower than those that we obtained
and stages greater than or equal to F2.        and APRI methods have also been de-                            with MR elastography for several rea-
Even if living tissues have both elastic       veloped. These serum tests involve the                         sons: First, the volumes assessed with
and viscous properties—and, thus, a            use of biochemical markers that have no                        MR elastography were much larger than
viscoelastic model seems more appro-           direct relationship with fibrosis and a                         those evaluated with US elastography.
priate than a single elastic model—fur-        purely statistical approach to predicting                      Second, MR elastography enables as-
ther studies are needed to determine if        the fibrosis stage (23,44). The APRI was                        sessment of the entire three-dimen-
an improved model would yield more             measured in our study because of its                           sional displacement vector induced by
accurate results for viscosity than the        simplicity and because it is fairly effec-                     the mechanical waves. Third, the use of
Voigt model used in our study (18).            tive (14,23). However, APRI measure-                           compressional waves at MR elastogra-
     We also observed that measure-            ments were less reliable than elasticity                       phy permits good penetration through-
ments of the intrasubject heterogene-          measurements in the staging of liver fi-                        out the liver. However, the advantages
ities of elasticity and viscosity increased    brosis. These results are in agreement                         of MR elastography are balanced by
with increasing fibrosis stage. This find-       with previous study findings (2,3),                             higher cost and increased examination
ing is consistent with the known vari-         which showed that serum scoring sys-                           time compared with those of US elastog-
ability in the distribution of fibrosis         tems enable accurate differentiation of                        raphy. The clinical effectiveness and

464                                                                                                                  Radiology: Volume 245: Number 2—November 2007
GASTROINTESTINAL IMAGING: MR Elastography of Liver Fibrosis                                                                                      Huwart et al

cost-effectiveness of MR elastography           study show that MR elastography is a                        Sirlin CB. Liver fibrosis: noninvasive diagno-
and US elastography need to be com-             reliable imaging method and is supe-                        sis with double contrast material– enhanced
                                                                                                            MR imaging. Radiology 2006;239(2):425–
pared in the same patients in future            rior to APRI measurement for staging
studies.                                        liver fibrosis. These findings suggest
     Our study was limited by the lack of       that noninvasive MR elastography po-                    13. Annet L, Materne R, Danse E, Jamart J,
precise correlation between the hepatic         tentially has a role in determining the                     Horsmans Y, Van Beers BE. Hepatic flow pa-
                                                                                                            rameters measured with MR imaging and Dopp-
sample analyzed by the pathologist and          treatment and the prognosis of patients
                                                                                                            ler US: correlations with degree of cirrhosis and
the viscoelastic maps constructed at MR         with chronic liver disease because it en-                   portal hypertension. Radiology 2003;229(2):
elastography. Because of the heteroge-          ables substantial and advanced fibrosis                      409–414.
neity of liver fibrosis, the length of the       to be readily diagnosed.
                                                                                                        14. Castera L, Vergniol J, Foucher J, et al. Pro-
biopsy sample is known to influence the                                                                      spective comparison of transient elastogra-
accuracy of METAVIR score– based cat-           Acknowledgment: Philips Medical Systems pro-                phy, Fibrotest, APRI, and liver biopsy for the
egorization. Bedossa et al found that           vided the experimental setup—including the pulse            assessment of fibrosis in chronic hepatitis C.
correct categorization of liver fibrosis         generator, transducer, and software for data acquisi-       Gastroenterology 2005;128(2):343–350.
increased when the length of the biopsy         tion and image analysis—used in this study.
                                                                                                        15. Sandrin L, Fourquet B, Hasquenoph JM,
sample was increased to 25 mm, with-                                                                        et al. Transient elastography: a new noninva-
out any substantial benefit from longer          References                                                  sive method for assessment of hepatic fibro-
specimens (31). In our study, the mean           1. Rockey DC. Antifibrotic therapy in chronic liver         sis. Ultrasound Med Biol 2003;29(12):1705–
length of the biopsy specimens was 34.5             disease. Clin Gastroenterol Hepatol 2005;3(2):          1713.
mm       10.5 and thus adequate for cor-            95–107.                                             16. Ziol M, Handra-Luca A, Kettaneh A, et al.
rect categorization.                             2. Rockey DC, Bissell DM. Noninvasive mea-                 Noninvasive assessment of liver fibrosis by
     Like previous study investigators,             sures of liver fibrosis. Hepatology 2006;43(2            measurement of stiffness in patients with
we used a transcostal approach for liver            suppl 1):S113–S120.                                     chronic hepatitis C. Hepatology 2005;41(1):
                                                                                                            48 –54.
MR elastography (17,18). It has been             3. Bataller R, Brenner DA. Liver fibrosis. J Clin
shown that a transcostal approach                   Invest 2005;115(2):209 –218.                        17. Rouviere O, Yin M, Dresner MA, et al. MR
yields better shear waves in the liver                                                                      elastography of the liver: preliminary re-
                                                 4. Friedman SL. Liver fibrosis: from bench to               sults. Radiology 2006;240(2):440 – 448.
than does the subcostal approach and                bedside. J Hepatol 2003;38(suppl 1):S38 –
that both approaches yield similar elas-            S53.                                                18. Huwart L, Peeters F, Sinkus R, et al. Liver
ticity results (17).                                                                                        fibrosis: non-invasive assessment with MR
                                                 5. Bravo AA, Sheth SG, Chopra S. Liver biopsy.             elastography. NMR Biomed 2006;19(2):
     In our study, some hepatic vessels             N Engl J Med 2001;344(7):495–500.                       173–179.
that were included in the field of view
might have influenced results. However,           6. Cadranel JF, Rufat P, Degos F; for the Group        19. Sinkus R, Tanter M, Xydeas T, Catheline S,
                                                    of Epidemiology of the French Association               Bercoff J, Fink M. Viscoelastic shear proper-
we think that this influence was minimal
                                                    for the Study of the Liver (AFEF). Practices            ties of in vivo breast lesions measured by MR
because the ROIs were placed on a sag-              of liver biopsy in France: results of a pro-            elastography. Magn Reson Imaging 2005;
ittal section away from the large vessels           spective nationwide survey. Hepatology                  23(2):159 –165.
of the hepatic hilum. Therefore, the he-            2000;32(3):477– 481.
patic vessels represented only a small                                                                  20. Sinkus R, Tanter M, Catheline S, et al. Imag-
                                                 7. Castera L, Negre I, Samii K, Buffet C. Pain             ing anisotropic and viscous properties of
area on the viscoelastic maps. Although             experienced during percutaneous liver bi-               breast tissue by magnetic resonance-elastog-
we were able to use only rectangular                opsy. Hepatology 1999;30(6):1529 –1530.                 raphy. Magn Reson Med 2005;53(2):372–
ROIs, we are currently working to de-            8. Poynard T, Ratziu V, Bedossa P. Appropri-               387.
velop a software that permits the place-            ateness of liver biopsy. Can J Gastroenterol
                                                                                                        21. Manduca A, Oliphant TE, Dresner MA, et al.
ment of arbitrary ROIs.                             2000;14(6):543–548.
                                                                                                            Magnetic resonance elastography: non-inva-
     We did not study the influence of                                                                       sive mapping of tissue elasticity. Med Image
                                                 9. Bruix J, Sherman M. Management of hepato-
steatosis, edema, or iron overload on               cellular carcinoma. Hepatology 2005;42(5):              Anal 2001;5(4):237–254.
elasticity and viscosity measurements.              1208 –1236.
                                                                                                        22. Wilson LE, Torbenson M, Astemborski J,
In a study of US elastography, Ziol et al                                                                   et al. Progression of liver fibrosis among in-
                                                10. de Franchis R. Evolving consensus in portal
(16) reported that at multivariate analy-           hypertension: report of the Baveno IV Con-              jection drug users with chronic hepatitis C.
sis of fibrosis, necroticoinflammatory                sensus Workshop on Methodology of Diag-                 Hepatology 2006;43(4):788 –795.
activity, and steatosis, fibrosis was the            nosis and Therapy in Portal Hypertension.
                                                                                                        23. Wai CT, Greenson JK, Fontana RJ, et al. A
only parameter significantly correlated              J Hepatol 2005;43(1):167–176.
                                                                                                            simple noninvasive index can predict both
to liver elasticity. The influences of ste-      11. Jalan R, Hayes PC; for the British Society of           significant fibrosis and cirrhosis in patients
atosis, edema, and iron overload on he-             Gastroenterology. UK guidelines on the                  with chronic hepatitis C. Hepatology 2003;
patic viscoelastic parameters should be             management of variceal haemorrhage in cir-              38(2):518 –526.
further studied in larger series and ex-            rhotic patients. Gut 2000;46(suppl 3-4):III1–
                                                                                                        24. The French METAVIR Cooperative Study
perimental models.                                                                                          Group. Intraobserver and interobserver
     In conclusion, the results of our          12. Aguirre DA, Behling CA, Alpert E, Hassanein TI,         variations in liver biopsy interpretation in

Radiology: Volume 245: Number 2—November 2007                                                                                                           465
GASTROINTESTINAL IMAGING: MR Elastography of Liver Fibrosis                                                                                       Huwart et al

      patients with chronic hepatitis C. Hepatol-     34. Maharaj B, Maharaj RJ, Leary WP, et al.          41. Patel K, Lajoie A, Heaton S, et al. Clinical
      ogy 1994;20(1 pt 1):15–20.                          Sampling variability and its influence on the         use of hyaluronic acid as a predictor of fibro-
                                                          diagnostic yield of percutaneous needle bi-          sis change in hepatitis C. J Gastroenterol
25. Bedossa P. An algorithm for the grading of
                                                          opsy of the liver. Lancet 1986;1(8480):523–          Hepatol 2003;18(3):253–257.
    activity in chronic hepatitis C. Hepatology
    1996;24:289 –293.                                                                                      42. Callewaert N, Van Vlierberghe H, Van
                                                      35. Soloway RD, Baggenstoss AH, Schoenfield               Hecke A, Laroy W, Delanghe J, Contreras R.
26. Foucher J, Chanteloup E, Vergniol J, et al.           LJ, Summerskill WH. Observer error and
    Diagnosis of cirrhosis by transient elastogra-                                                             Noninvasive diagnosis of liver cirrhosis using
                                                          sampling variability tested in evaluation of         DNA sequencer-based total serum protein
    phy (FibroScan): a prospective study. Gut             hepatitis and cirrhosis by liver biopsy. Am J
    2006;55(3):403– 408.                                                                                       glycomics. Nat Med 2004;10(4):429 – 434.
                                                          Dig Dis 1971;16(12):1082–1086.
27. Brunt EM. Grading and staging the his-            36. Regev A, Berho M, Jeffers LJ, et al. Sampling    43. Poon TC, Hui AY, Chan HL, et al. Prediction
    topathological lesions of chronic hepatitis:          error and intraobserver variation in liver bi-       of liver fibrosis and cirrhosis in chronic hep-
    the Knodell histology activity index and be-          opsy in patients with chronic HCV infection.         atitis B infection by serum proteomic
    yond. Hepatology 2000;31(1):241–246.                  Am J Gastroenterol 2002;97(10):2614 –                fingerprinting: a pilot study. Clin Chem
                                                          2618.                                                2005;51(2):328 –335.
28. Cohen A. Comparison of correlated correla-
    tions. Stat Med 1989;8(12):1485–1495.             37. Strader DB, Wright T, Thomas DL, Seeff LB.       44. Imbert-Bismut F, Ratziu V, Pieroni L, Char-
29. Korn EL. Sample size tables for bounding              Diagnosis, management, and treatment of              lotte F, Benhamou Y, Poynard T. Biochemi-
    small proportions. Biometrics 1986;42(1):             hepatitis C. Hepatology 2004;39(4):1147–             cal markers of liver fibrosis in patients with
    213–216.                                              1171.                                                hepatitis C virus infection: a prospective
                                                                                                               study. Lancet 2001;357(9262):1069 –1075.
30. Colloredo G, Guido M, Sonzogni A, Leandro         38. Crawford J. Liver cirrhosis. In: MacSween
    G. Impact of liver biopsy size on histological        RN, Burt AD, Portmann BC, Ishak KG,              45. Koinuma M, Ohashi I, Hanafusa K, Shibuya
    evaluation of chronic viral hepatitis: the            Scheuer PJ, Anthony PP, eds. Pathology of            H. Apparent diffusion coefficient measure-
    smaller the sample, the milder the disease.           the liver. 4th ed. London, England: Churchill        ments with diffusion-weighted magnetic res-
    J Hepatol 2003;39(2):239 –244.                        Livingstone, 2002;575:– 620.                         onance imaging for evaluation of hepatic fi-
                                                      39. Kazemi F, Kettaneh A, N Kontchou G, et al.           brosis. J Magn Reson Imaging 2005;22(1):
31. Bedossa P, Dargere D, Paradis V. Sampling                                                                  80 – 85.
    variability of liver fibrosis in chronic hepati-       Liver stiffness measurement selects patients
    tis C. Hepatology 2003;38(6):1449 –1457.              with cirrhosis at risk of bearing large oe-
                                                                                                           46. Yeh WC, Li PC, Jeng YM, et al. Elastic mod-
                                                          sophageal varices. J Hepatol 2006;45(2):
32. Siddique I, El-Naga HA, Madda JP, Memon                                                                    ulus measurements of human liver and cor-
                                                          230 –235.
    A, Hasan F. Sampling variability on percuta-                                                               relation with pathology. Ultrasound Med Biol
                                                      40. Guechot J, Laudat A, Loria A, Serfaty L,             2002;28(4):467– 474.
    neous liver biopsy in patients with chronic
                                                          Poupon R, Giboudeau J. Diagnostic accuracy
    hepatitis C virus infection. Scand J Gastroen-
                                                          of hyaluronan and type III procollagen           47. Colletta C, Smirne C, Fabris C, et al.
    terol 2003;38(4):427– 432.
                                                          amino-terminal peptide serum assays as               Value of two noninvasive methods to detect
33. Abdi W, Millan JC, Mezey E. Sampling vari-            markers of liver fibrosis in chronic viral hep-       progression of fibrosis among HCV carriers
    ability on percutaneous liver biopsy. Arch            atitis C evaluated by ROC curve analysis.            with normal aminotransferases. Hepatology
    Intern Med 1979;139(6):667– 669.                      Clin Chem 1996;42(4):558 –563.                       2005;42(4):838 – 845.

466                                                                                                              Radiology: Volume 245: Number 2—November 2007