Can transient elastography be used for the management of chronic hepatitis B patients?

Transient elastography (TE) is a non-invasive method of quantifying liver fibrosis that has been extensively studied in chronic hepatitis C (CHC). Its utilization is increasingly accepted as an alternative to liver biopsy. In chronic hepatitis B (CHB) patients, a precise evaluation of liver fibrosis stage is also of crucial importance to take appropriate therapeutic decisions and evaluate the need for hepato-cellular carcinoma surveillance. In several centres world, TE has entered the clinical practice and many clinicians extrapolate to CHB what they usually do in CHC. However, TE has been so far less studied in CHB. Because of the complex natural history of CHB that frequently displays a fluctuating pattern in terms of liver inflammation and fibrosis progression, it would be conceivable that differences exist between CHB and CHC. In a recent study performed in 202 CHB patients, Marcellin et al. 1 reported that the global diagnostic accuracy of TE given by AUROC were 0.81 for METAVIR ≥F2, 0.93 for ≥F3 and 0.93 for F4, values that were close to that usually reported in CHC. A recent meta-analysis using indirect comparisons also showed that the diagnostic accuracy of TE in CHB was comparable to that observed in CHC 2. However, only few studies have conducted direct comparisons between CHC and CHB. In this issue of Liver International, Verveer et al. 3 and Cardoso et al. 4 present studies that directly compared the diagnostic performance of TE between the two settings. Authors must be congratulated for performing studies that have included a great number of patients with special attention paid to quality criteria of liver biopsies. The main result of these two studies is that the global diagnostic performance of TE was similar between the two diseases, with AUROC roughly around 0.80 for the diagnosis of significant fibrosis (METAVIR ≥F2) and 0.90 for the diagnosis of cirrhosis. One major limitation of using TE in clinical practice is the definition of cut-offs that should be used to reliably classify patients in a given fibrosis stage. In the meta-analysis by Tsochatzis et al. 2, optimal cut-offs could be retrieved in 4 CHB and 14 CHC studies. Although variability was observed, reported cut-offs were globally lower in CHB compared with CHC (on average 7.0 vs 7.6 for ≥F2, 8.2 vs 10.9 for ≥F3 and 11.3 vs 15.3 for F4). A logical explanation for these discrepancies would be that fibrous septa are usually thinner in CHB compared with CHC. Indeed, a recent morphometric study has suggested that fibrosis areas – reflecting the total amount of liver fibrosis – were significantly lower in CHB compared with CHC patients after stratification on METAVIR fibrosis stages 5. In addition, because CHB tends to result in cirrhosis with larger nodules – i.e. macronodular cirrhosis – than in CHC, TE pulse is more likely to pass the normal liver parenchyme between fibrotic bands leading to lower liver stiffness (LS). These histological features might explain why a recent study found that median LS was nearly twice greater in F3 compared with F4 CHB patients, with a poor reliability of TE for the diagnosis of cirrhosis in CHB 6. However, this study included only 70 CHB patients with a small number of F3. The two studies published in Liver International give interesting data on this issue. In the Cardoso study, LS values according to each fibrosis stage were globally similar between CHB and CHC except for cirrhosis (14 vs 20 kPa), although this difference was not significant. Authors did not calculate from their series optimal cut-offs, but used a priori cut-offs calculated in previous studies (≥F2: 7.2 vs 7.1; ≥F3: 8.1 vs 9.5; F4: 11.0 vs 12.5 kPa for CHB and CHC respectively). Agreement between liver biopsy and TE were close in CHB compared with CHC (for example 82% vs 78% for the diagnosis of significant fibrosis and 89% vs 93% for the diagnosis of cirrhosis). However, positive predictive value were lower in CHB compared with CHC (example of cirrhosis: 39% vs 58%) whereas negative predictive value were similar (98% in both groups). These data suggest that using lower cut-offs in CHB decreases the risk of being false negative while increasing the risk of being false positive. In the Verveer study, optimal cut-offs optimizing sensitivity were calculated in both groups. Comparing CHB and CHC, values were 6 vs 5 kPa for ≥F2, 9 vs 8 kPa for ≥F3, and 13 vs 13 kPa for F4 respectively. Therefore, optimal cut-offs were slightly greater in CHB compared with CHC. One explanation might be that the spectrum bias, linked to the different repartition of fibrosis stages across the study groups, was not taken into consideration. For example, the fact that more patients were F1 – while non-significant – in CHB compared with CHC could have increased LS in the F0F1 CHB group and subsequently increase the cut-off discriminating F0F1 vs F2F3F4. However, median LS according to each stage of fibrosis were globally similar between CHB and CHC and this potential bias is not likely to have played a major role. Another explanation for the discrepancies between studies is the impact of other histological parameters, such as necroinflammation and steatosis. Some reports have suggested that influence of steatosis on LS was negligible but this association may have been attenuated by the low prevalence of severe steatosis 7. In contrast, investigators have concluded that fatty liver on ultrasonographic examination can raise LS values 8. However, the insufficient biopsy data in this study weakens their conclusions. In the Cardoso study, steatosis was associated to LS in CHC but not in CHB, in which it was less frequent (11% vs 23%). Similar data were recently reported in another comparative study 9. Thus, the lower prevalence of steatosis combined to lower total content of fibrosis in CHB could again favour lower LS values and corresponding cut-offs in CHB. By contrast, liver inflammation that clearly appears to contribute to LS could be more pronounced in CHB patients especially during viral reactivation and could counterbalance the impact of other lesions 10. The Verveer study confirmed that necroinflammatory lesions were associated to greater LS in both CHB and CHC. However, biopsies from CHB patients showed less inflammation than in CHC. The reported higher LS cut-offs in CHB patients remain therefore hard to understand. Although detailed analytical methods were not provided, both Cardoso and Verveer studies found alanine aminotransferase (ALT) serum levels (an indirect marker of necroinflammation) to be significantly associated to LS in CHB and CHC. The formal demonstration that elevated ALT increase LS was brought by the fact that antiviral therapy rapidly decreases LS in parallel to ALT normalization 11. In that study, the diagnostic performance of TE was better after ALT normalization, suggesting that elevated ALT lead to overestimation of fibrosis. However, ALT levels were especially high, of more than two-fold in 60% and more than five-fold the upper limit of normal (ULN) in 26% of cases. One can thus consider that only greatest ALT values have an impact of LS. However, even small changes in ALT level have been shown to influence LS values and significant differences between patients with ALT < 1xULN and those with ALT 1–2 ULN have been reported 12. An unresolved question is whether ALT levels, which frequently fluctuate in CHB, should be taken into consideration to adapt the cut-offs used for fibrosis categorization. In the study by Myers et al. 13 including various liver diseases, ALT > 60 IU/L was a parameter associated to discordance between TE and biopsy. Chan et al. 14 have therefore proposed an algorithm using higher cut-offs in patients with elevated ALT to avoid overestimation of fibrosis because of inflammation. However, cut-offs were close to each other (for example for cirrhosis values were 12.0 vs 13.4 kPa for normal and elevated ALT respectively) and there was no statistical comparison with different cut-offs. In the Cardoso study, both strategies – using fixed or adapted to ALT cut-offs – were compared. Interestingly, adapting cut-offs on ALT did not improve the percentage of patients correctly classified. It is important to note that patients with ALT > 5xULN were excluded from that analysis. Furthermore, an ALT level of 1.5–2xULN seems to be a better cut-off value to enhance the accuracy of TE 15. Hence, instead of the arbitrary categorization in the Cardoso study (ALT<ULN vs ALT 1–5xULN), stratification using cut-off LS values calculated from their own study population with maximal statistical power might have revealed the potential influence of high ALT levels on LS values. Although some recent studies on TE did not include high ALT patients to increase the reliability of TE 15, the optimal ALT cut-off should be determined in future studies. From all these data, what conclusions can be drawn on the use of TE in CHB? Firstly, its diagnostic performance is globally similar to that observed in CHC. Secondly, although refinement could be made in selecting more appropriate cut-offs possibly adapted on ALT levels, the use of cut-offs already described and used in CHC does not alter considerably the accuracy of fibrosis estimation by TE. Third, obvious pitfalls such as false negativity because of macronodular cirrhosis or false positivity because of ALT flare-up should be known and TE results should always be interpreted with caution and along with the clinical context. Finally, the main question is whether that TE results are reliable enough to take therapeutic decisions in CHB without the need of a liver biopsy. Because LS is the consequence of multiple intra-hepatic events including steatosis and inflammation, it will never be a perfect marker of fibrosis. However, because these other histological features contribute to severity of the natural history of CHB, it would be logical to consider LS as an independent parameter that needs to be tested as a long-term prognostic factor. Encouraging results were recently published by Jung et al. 16 suggesting that the cumulative risk of hepato-cellular carcinoma at 3 years was very low in patients with LS below 8 kPa at inclusion. Further long-term longitudinal studies are required before TE can be routinely implemented in the management of CHB.