Cirrhosis of the liver is the end stage of a complex process—resulting from hepatocyte injury and the response of the liver—that leads to partial regeneration and fibrosis of the liver.
Cirrhosis poses a difficult challenge for management, while the disease's prevention, detection, and therapy engender major health costs. Diagnostic imaging offers diverse modalities for use in the noninvasive evaluation of the liver, as well as in interventional techniques; the latter may be used to treat such complications as portal hypertension and neoplasia. The diagnosis, management, and treatment of cirrhosis are reviewed in this article.
Anatomy
Hepatic morphologic changes
Regardless of etiology, gross morphologic changes of cirrhosis are recognized by a variety of image techniques. Enlargement of the left lobe and caudate lobe, believed to be the result of lobar-relative regeneration rather than fibrosis, secondary to an accident of vascular supply, is recognized by any cross-sectional technique, such as computed tomography (CT) scanning, magnetic resonance imaging (MRI), or ultrasonography (US), each depicted in the images below.
Using MRI, Okazaki and colleagues determined that alcoholic cirrhosis is associated more frequently with caudate lobe enlargement and the presence of a right posterior hepatic notch than is virus-induced cirrhosis.[1] Harbin, Hess, Giorgio, Torres, and their coauthors have described a number of indices, including the ratio of transverse caudate lobe width to right lobe width, multidimensional caudate lobe indices that can be obtained by US or CT scanning, and volume analysis of each liver segment, based on cross-sectional area by CT scanning or MRI.[2, 3, 4, 5] Lafortune and colleagues suggested that a reduction in the medial segment of the left hepatic lobe diameter is a helpful adjunct finding of cirrhosis in ultrasonographic investigation.[6]
Another sign of cirrhosis, the expanded gallbladder fossa sign as depicted in the image below, has been described on MRI examination, based on an evaluation by Ito and coauthors of 190 patients with cirrhosis and of 123 control patients.[7] The authors' criterion was enlargement of the pericholecystic space (ie, gallbladder fossa)—which had to be demarcated laterally by the edge of the right hepatic lobe, medially by the edge of the lateral segment of the left hepatic lobe, or posteriorly by the anterior edge of the caudate lobe—in conjunction with nonvisualization of the medial segment of the left hepatic lobe on the same axial image. This achieved a sensitivity, specificity, accuracy, and positive predictive value for the MRI diagnosis of cirrhosis of 68%, 98%, 80%, and 98%, respectively.
On ultrasonographic examination, the liver contour may appear nodular, as in the first image below, although Ladenheim and colleagues have questioned the specificity of this sign. Similar contour deformities, depicted in the second image below, are evident on examination by CT scanning or MRI. The echo texture appears coarsened. Increase in echogenicity, shown in the third image below, is caused by fatty infiltration, which may be diffuse in hepatitis or focal in hepatitis or cirrhosis.
Intrahepatic vascular changes in cirrhosis
In cirrhosis, the dynamics of the hepatic arterial and portal venous circulation change as the degree of fibrosis progresses, as depicted in the image below.
In addition, the vessels appear to elongate and become more tortuous because of the underlying parenchymal architectural distortion. This is recognized classically in angiography as "corkscrewing" of vessels and can be appreciated on cross-sectional imaging, depicted below.
Secondary manifestations of cirrhosis may be seen as morphologic or physiologic evidence of the disease. The development of spontaneous shunts has been described in advanced cirrhosis and was initially demonstrated by angiography, although it is now demonstrable by noninvasive techniques, such as Doppler US as in the image below, at an incidence of up to 7%.
Dual-phase CT scanning can demonstrate these shunts as early opacification of the intrahepatic veins during the early arterial phase-injection, portrayed below. The shunts are often accompanied by geographic, wedge-shaped perfusion abnormalities.
Extrahepatic manifestations of cirrhosis detectable by imaging techniques
Marshak, Karahan, and coauthors reported a higher frequency in the alteration in the thickness of the wall of the GI tract, depicted below, in patients with cirrhosis than in controls (64% vs 7%).[8, 9]
The development of splenomegaly and collaterals from portal hypertension is readily evident when any cross-sectional technique is used. Nodular iron deposition within the spleen, as seen on seen on MRI scans (Gamma-Gandy bodies), is highly suggestive of portal hypertension.
Functional imaging techniques, such as the use of technetium-99m (99m Tc)–labeled sulfur colloid, which is taken up by reticuloepithelial cells, and the presence of "colloid shift" to the bone marrow in cirrhosis, in addition to the recognition of hepatic morphologic changes and splenomegaly, have been helpful in confirming the presence and severity of cirrhosis, as in the image below.
Portal hypertension
Portal hypertension occurs once portal pressures reach 5-10 mm Hg above normal as a complication of cirrhosis. The well-recognized effect of increasing portal pressures is the development of splenomegaly, depicted below, and collateral portal-venous anastomoses, which occur at numerous sites, including gastroesophageal, paraumbilical, perirectal, and retroperitoneal locations.
Varices are not found where the portal vein pressure (indirectly measured as the hepatic vein pressure gradient [HVPG]) is less than 12 mm Hg. However, not all patients with elevated portal pressures develop variceal bleeding. Noninvasive diagnostic imaging methods, such as color flow Doppler US, contrast-enhanced CT scanning, and MRI, can be used to identify the presence of collaterals, as in the images below, but a major limitation is an inability to employ them in evaluating variceal pressures, which correlate more directly with the risk of hemorrhage.
Endoscopic evaluation provides a visual window on esophageal varices, which can be graded for prognosis. However, noninvasive techniques are useful in demonstrating collateral vessels beyond the reach of the endoscope. The images below depict 3-dimensional CT scanning, which is particularly helpful in demonstrating the development and pattern of collateral flow in portal venous hypertension.
Slow portal flow can mimic portal vein occlusion on cross-sectional imaging, and care must be taken when interpreting images in which not all diagnostic criteria are met. Long-standing thrombosis may be associated with cavernous transformation in which periportal collaterals re-establish flow to the liver, even in the setting of cirrhosis and elevated sinusoidal pressures. Neoplastic invasion of the portal veins must be differentiated from bland thrombus, as in the images below.
Preferred examination
Ultrasonography is the most widely used worldwide imaging modality and is used in combination with serum alpha fetoprotein (AFP) screening, based on evidence that increased frequency of examination leads to detection of hepatocellular carcinoma (HCCA) at an earlier stage. It is common practice to screen patients with chronic hepatitis and/or biopsy-documented cirrhosis annually or semiannually with these techniques, although in the United States the American Gastroenterological Association (AGA) does not officially endorse this practice. The accuracy of US, as with CT scanning and MRI, is more limited in the advanced stages of cirrhosis. A US study from Korea, with transplant correlation in 52 patients, demonstrated a sensitivity for the detection of HCCA of only 33% (6 of 18) lesions.
CT scanning is believed to be equivalent in sensitivity to, and more specific than, US. However, there are disadvantages related to contrast risk and radiation exposure, particularly if the modality is used over a lifetime for screening. Thus, CT scanning should be reserved for equivocal cases or for patients in whom disparate results are observed. For example, a heterogeneous appearance on US evaluation of the liver, as shown below, may mask malignant lesions and justify additional imaging.
Conversely, persistent lesions that are noted on US, even if not confirmed on a CT scan, probably should be biopsied, particularly in the setting of serologic abnormalities, as depicted in the image below. In end-stage patients who are destined for transplantation, the sensitivity of CT scanning is reduced (to as low as 37% in one series).[10] .
MRI with gadolinium (or other contrast agents) can be used as an alternative (although more costly) study. Similar to US and CT scanning however, a reduction in sensitivity to below 50% has been reported, particularly for lesions under 2 cm, in patients with end-stage disease.[11, 12] Even more invasive and sophisticated techniques, such as CT scanning performed with a catheter in the hepatic artery, as well as angiography, are usually reserved for use in patients undergoing evaluation for transplant at regional centers, where the goal is to exclude or establish the presence and multiplicity of malignant lesions for pretransplant assessment. These techniques are not routinely employed in the United States but are used extensively in Asia.
Unfortunately, for cultural reasons, orthotopic liver transplantation is not routinely performed in these countries; thus, pathologic correlation is limited to hepatic resections and biopsies. However, with the rapid increase in right lobe liver donation surgery, the pathologic correlation should be excellent because the entire explanted liver will be available.
Limitations of techniques
Real-time US is used extensively for screening, but biopsy or additional imaging modalities are required for confirmation. US is a nonspecific test and identifies many nodules, ranging from regenerative nodules, dysplastic nodules, and focal fat to benign neoplasms, such as hemangioma, many of which have no uniquely discriminating features on US.
Because these occur with significant frequency, they pose a diagnostic challenge. For example, in a study of screened patients with cirrhosis, the authors discovered that although combined assessment with US and AFP was accurate in identifying patients with HCCA (who formed 24% of the study's population), 25% of patients had benign focal masses, such as hemangioma or focal fat, requiring further imaging evaluation, and another 20% had focal lesions that could not be corroborated on other imaging studies or on subsequent US. This relatively high prevalence of benign lesions in patients with cirrhosis appears to be corroborated by a study by Horigome and colleagues.[13]
Therefore, it is necessary to commit either to the biopsy of all persistent lesions or to the corroboration of them prior to biopsy with other techniques, such as CT scanning (helical or multislice) or MRI, using dynamic imaging with contrast to obtain multiple vascular-phase images. Clinician preferences in the United States, as surveyed by Chalasani and coauthors, suggest an empirical trend toward routinely incorporating CT scanning in screening.[14]
Some cause for optimism is warranted in terms of reduction in the incidence of HCCA in patients with hepatitis C following interferon therapy. A meta-analysis by Papatheodoridis and colleagues of 11 studies involving more than 2000 patients determined that the incidence of HCCA in patients who underwent interferon therapy was reduced to 8.2%, compared with 21.5% in untreated patients, and was even lower in sustained responders (0.9%).[15]
A major unresolved problem is the evaluation of the efficacy of screening and the economic consequences of aggressive screening. Bolondi and coauthors, in Italy, and Larcos and colleagues, in the United States, estimated that each case of HCCA that is detected costs $8000-$24,000.[16, 17]
Despite the best efforts of the worldwide medical community in screening for HCCA, no evidence exists that mortality has been affected, because therapeutic options, although expanding, remain relatively limited. Survival in patients undergoing liver transplant who have unsuspected HCCA is adversely affected by tumor recurrence (reported in a French series by Adam and colleagues as reaching 5%).[18] The presence of neoplasm is not a contraindication to transplant, although survival in patients with tumors larger than 3 cm, with multiple nodules or portal invasion, is sufficiently impacted to preclude consideration in this subgroup.
Intervention
Of those patients with cirrhosis and varices, 25-40% experience bleeding. The management of portal hypertension and upper tract GI bleeding has been revolutionized by endoscopic and angiographic treatment. The use of the intravascularly placed transjugular intrahepatic portosystemic shunt (TIPS) has provided a second-line therapy for the management of portal hypertension, with reduced mortality and morbidity compared with that associated with the open surgical procedure.
Fischer, Kimura, and coauthors have used noninvasive modalities, such as Doppler US, portrayed below, to monitor shunt patency (with a reasonably high accuracy of greater than 90%).[19, 20] Occasionally, however, conventional Doppler techniques fail to image signals. Thus, enhancement by US contrast agents appears promising in improving visualization, as shown below.
Patient education
For patient education information, see the Hepatitis Center and the Liver, Gallbladder, and Pancreas Center, as well as Hepatitis A, Hepatitis B,Hepatitis C, and Cirrhosis.
Tidak ada komentar:
Posting Komentar