Transient elastography in acute hepatitis: All that's stiff is not fibrosis


  • See Articles on Page 380 and Page 592

  • Potential conflict of interest: Nothing to report.

Two papers in this issue of HEPATOLOGY examine the effect of inflammation on liver stiffness measurement (LSM), as assessed by transient elastography (Arena and Vizzutti et al.,1 Sagir et al.2). These papers challenge the commonly held assumption that liver stiffness is determined solely by the severity of hepatic fibrosis. Such a view is underpinned by the clinical experience that, on palpation, a cirrhotic liver is hard, and experimentally, by the correlation of stiffness (elastic modulus) with the histopathological fibrosis score in ex vivo hepatic samples.3 However, in the proof-of-principle study by Yeh and colleagues,3 it was stressed that “biological tissue is a composite material and it is difficult to separate the influence of each component of the tissue on the total of modulus estimates”. Demonstration that biological variables other than fibrosis contribute to LSM, therefore comes as no surprise.


ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUROC, area under the receiver operator characteristic; HPVG, hepatic venous pressure gradient; LSM, liver stiffness measurement; NAFLD, nonalcoholic fatty liver disease.

The development of transient elastography was a leap forward, enabling the rapid acquisition of objective liver stiffness measurements in vivo.4 In the seminal work by Sandrin and colleagues,4 a cohort of patients with chronic hepatitis C was studied. This patient group is appropriate for study as the natural history of the disease and its histological classification is well established, the prevalence of the disease is high in many countries and effective treatment options are available.5–7 Multiple regression analysis in early studies demonstrated a relationship between elasticity measurements and fibrosis stage, but not to the disease activity (or necroinflammatory score), nor to the degree of steatosis.4, 8 However, the observation that the range of liver stiffness values obtained in patients with cirrhosis was very large led to speculation that there may be a relationship between portal hypertension and liver stiffness.4

Subsequently, a number of well-conducted studies confirmed the correlation between liver stiffness and hepatic fibrosis in hepatitis C,8, 9 and other chronic hepatic conditions.10–12 However, studies have also indicated the presence of co-existing factors that contribute to liver stiffness.

Area under the receiver operator characteristic (AUROC) curves plot sensitivity and specificity for all possible cutoff values between two given states, giving a measure of diagnostic accuracy. Accordingly, quoted cutoff values vary between patient cohorts with the same disease, between cohorts of differing diseases and according to whether the operator requires a cutoff value which gives maximal sensitivity or specificity.10, 11 While the prevalence of a disease in the population studied affects the diagnostic accuracy, the presence of disease-specific factors other than fibrosis stage alone may also contribute to differences in LSM.

Substantial differences in cutoff values for cirrhosis have been observed between those with chronic hepatitis and those with alcohol-related chronic liver disease and non-alcoholic fatty liver disease (NAFLD).11, 12 While the inter-test and inter-rater reliability for LSM was found to be excellent (intraclass correlation of 0.98) in a cohort of 200 patients with chronic liver disease in Italy, the reliability of measurements in overweight individuals and those with NAFLD as assessed by the intraclass correlation coefficient is reduced.13 However, in that study cohort, the mean BMI was just 24.8 kg/m2, which is lower than that found in many populations. As such, it was under-powered to examine the effects of obesity and NAFLD on the reliability of LSM. On the basis of clinical experience, LSM is difficult to perform successfully and reliably in obese individuals. Current evidence does not demonstrate an independent effect of hepatic steatosis on LSM, but it remains to be seen whether steatosis per se systematically alters LSM in a study where patients are stratified primarily by a validated method for lipid quantitation, such as proton magnetic resonance spectroscopy.14

A number of studies have pointed to a relationship between portal hypertension and liver stiffness measurement. Foucher and colleagues demonstrated a correlation between liver stiffness measurement and splenomegaly, the presence of oesophageal varices and a history of bleeding varices.15 A relationship between LSM and presence of varices has also been described, although evidence for the relationship between LSM and size of varices is mixed.16, 17 Vizzutti and colleagues went on to demonstrate correlation between LSM and the hepatic venous pressure gradient (HVPG), particularly at lower HVPG values (<10-12 mmHg). This represents a complex relationship, which was less apparent at higher HVPG values.16 It has been stated that the “progressive rise in portal pressure…[is] due mainly to an increase in intrahepatic vascular resistance from the accumulation of fibrillar extracellular matrix”.18 However, it might also be argued that increased portal blood inflow contributes directly to the liver stiffness, although haemodynamic changes characteristic of advanced portal hypertension, including extrahepatic haemodynamic changes, may not be detected by changes in LSM. Accordingly, the utility of LSM for the assessment of portal hypertension is limited.

The association between liver stiffness and disease activity or necroinflammatory score has been observed by Fraquelli and colleagues, who showed a step-wise increase of LSM with necroinflammatory activity in a cohort of patients with disease of varied aetiology.13 There was a closer temporal relationship between biopsy and LSM in this cohort (within 3 days) than in Sandrin's cohort (within 1 year), which may explain why the relationship between necroinflammation was seen in the former, but not the latter study. Coco and colleagues examined the relationship between LSM and biochemical activity in patients with chronic viral hepatitis. The LSM was lower, stage-for-stage in those with biochemical remission (as assessed by ALT) than those with a higher ALT. They observed that two patients with minimal fibrosis and biochemical flares had high LSM, which resolved with improvement of the ALT.19

The two papers in this issue of HEPATOLOGY specifically address the effect of hepatic inflammation on LSM. Pinzani's group devised an elegant study in which 18 patients without a past history of liver disease, but with acute viral hepatitis were studied.1 Patients were included if the aminotransferase levels were >10 times the upper limit of normal, in the context of serological evidence of viral hepatitis. LSM was performed on the same day as venesection at the peak increase in aminotransferases, at a reduction of aminotransferases to 50% or less than the peak and then nearing resolution, when aminotransferases were less than or equal to twice the upper limit of normal. These authors found that in all patients, the LSM at the peak aminotransferase level exceeded 12kPa (quoted here as a cutoff for prediction of cirrhosis) and furthermore, that in all but one subject, the LSM returned to within normal range (quoted here as 7kPa). In addition, the LSM correlated with the aminotransferases at onset and with the AST at follow-up. Particular strengths of this study include the use of tight exclusion criteria to minimise the possibility of pre-existing liver disease, the close temporal relationship between aminotransferase and liver stiffness assessment, the follow-up of all subjects to resolution of their hepatitis as assessed by the aminotransferase level, and the high success rate of LSM. While it is known that acute hepatitis is associated with an inflammatory infiltrate, tissue oedema and hepatocyte swelling, all of which are likely to affect LSM, there is no histological confirmation of these features in this study, as liver biopsy was not clinically indicated.

The next paper, from Haussinger's group complements that by Arena and colleagues.2 Here, 20 patients with an acute hepatitis were studied. Eight of these cases were of drug-induced hepatitis due to a variety of agents and three of autoimmune hepatitis. Seven of the eight drug-induced cases, all those with autoimmune hepatitis and one of those with viral hepatitis were biopsied, although only three patients were biopsied during the acute phase. The histological features of acute hepatitis in these individuals have not been described and so cannot be directly related to LSM. It is interesting to note the presence of fibrosis in all those biopsied, suggesting a degree of chronicity, nevertheless no patient exceeded a fibrosis score of F2 and the LSM of all of those biopsied exceeded 13kPa, putting them all within the commonly quoted “cirrhotic” range.11 In those followed up longitudinally, the aminotransferases returned to a level commensurate with the fibrosis stage at biopsy. In this paper, no correlation between aminotransferase levels and LSM was observed, but this may reflect the diverse aetiologies of the patients studied and the lack of close temporal association between venesection and LSM. Nevertheless, in one patient with nitrofurantoin-induced hepatitis, both bilirubin and ALT fell in close association with LSM. The strengths of this study include the availability of histology in a number of patients to exclude definitively cirrhosis in patients with LSM within the cirrhotic range, and the inclusion of patients with acute drug-induced hepatitis as a patient group not previously studied. Such studies would be strengthened further by a detailed examination of the histological features of acute hepatitis compared to LSM.

The conclusions of these two papers merit reiteration: 1. LSM in the cirrhotic range does not predict advanced fibrosis and cirrhosis in those with acute hepatitis. 2. A high LSM in the context of chronic liver disease may be due to an acute (aminotransferase) flare. 3. Necroinflammatory activity should be considered in studies of the performance of transient elastography.

Finally, it is worth considering how far we have come in addressing the conundrum posed in 2002: to separate the influence of each component of the tissue on the total modulus estimates. The two studies in this issue of HEPATOLOGY bring forward our understanding of the extent that inflammation, in the absence of fibrosis, can increase hugely the liver stiffness to within the range seen in patients with cirrhosis, and that this phenomenon is reversible.1,2 The relative contributions of each component cannot be determined by a single physical characteristic such as liver stiffness, and use of more than one modality for the assessment of disease severity is appropriate. Toward the clinic, these studies reinforce the necessity that when interpreting LSMs, the clinical question should be defined and LSM should be interpreted in the context of a defined diagnosis, biochemical data and appropriate cutoff points.