Non‐invasive assessment of liver fibrosis: Time to move from cross‐sectional studies to longitudinal ones

Liver fibrosis occurs in response to almost all causes of chronic liver insults, and the initiation of its deposition imposes an important phase in chronic liver disease. Eventually, without appropriate interventions, liver fibrosis progresses, leading to changes in liver morphology, deterioration of liver function and hemodynamics, complications due to portal hypertension, and an increased inclination for hepatocarcinogenesis. Thus, accurately determining the presence and degree of liver fibrosis is of paramount importance in identifying treatment strategies, responses to treatment, and the risks for liver-related complications and prognosis in patients with chronic liver disease. Liver biopsy remains the ‘gold standard’ for assessing the severity of liver fibrosis, but is invasive and sometimes associated with rare but serious complications, including bleeding, pneumothorax, and procedure-related death.1 Furthermore, performing repeated biopsies within a short time frame is impractical to assess changes in the degree of liver fibrosis. In addition to sampling error inherent to the percutaneous approach, there is both intra- and interobserver variability in histological interpretation.2 An ideal non-invasive method for evaluating liver fibrosis should accurately determine the presence of significant fibrosis. In addition, it should be readily available, highly reproducible, and widely applicable to liver diseases with various causes. Although liver biopsy does not fulfill many of these criteria, it has remained the gold standard mainly because of the absence of better alternatives. During the last 2 decades, many studies have proposed non-invasive tests to replace liver biopsy. Serum hyaluronic acid level, the aspartate aminotransferase (AST) to alanine aminotransferase (ALT) ratio, the AST to platelet ratio index (APRI), the age–spleen–platelet ratio index (ASPRI), Forn's test,3 Fibrotest, Fibrospect, Hepascore,4 and the European Liver Fibrosis panel (ELF) score are examples.5,6 Recently, transient elastography (TE) was introduced as a promising non-invasive method to assess liver fibrosis and it has considerable accuracy for predicting cirrhosis in patients suffering from chronic liver disease with diverse causes.7–9 The studies by Malik et al.10 and Kun et al.11 in this issue of the Journal of Gastroenterology and Hepatology compared the performance of several non-invasive methods for liver fibrosis, including TE, in patients with chronic liver disease. Malik et al. studied 404 patients.10 TE showed the highest performance for predicting liver cirrhosis regardless of cause. The diagnostic performance of TE, hyaluronic acid, clinical signs, APRI score, and the AST/ALT ratio for all patients with mixed causes, as reflected by the area under the receiver operating characteristics curves (AUROC), was 0.90, 0.81, 0.74, 0.71, and 0.66, respectively. Corresponding AUROC for patients with chronic hepatitis C (CHC) were 0.90, 0.76, 0.73, 0.70, and 0.61, respectively. Although TE showed the best performance for predicting liver cirrhosis in patients with CHC, the AUROC of TE seemed to be lower than those from European studies. Considering the paucity of TE data from the USA, where TE is not yet approved, further studies are needed to demonstrate the diagnostic performance of TE in that country for its widespread use. Importantly, the study by Malik et al.10 clearly shows the pros and cons of TE and liver biopsy, and how cross-sectional studies should be designed in the future by analyzing discrepancies in diagnosing cirrhosis using TE versus biopsy. When Hepascore was selected as a reference standard, this was correlated with liver biopsy results in cases with a body mass index (BMI) > 30 kg/m2 and a biopsy length > 2 cm, indicating that biopsy was more reliable than TE. Conversely, Hepascore was correlated with TE in cases with a BMI < 30 kg/m2 and biopsy length < 2 cm, indicating that TE performed better than liver biopsy. Accordingly, liver biopsy quality and use of TE only after removing cases with known confounders (i.e. high BMI) are important for future, high-quality cross-sectional studies. Indeed, current cross-sectional studies are trying to increase biopsy quality, while a new TE probe for obese patients is now under investigation.12 In another JGH article, Kun et al.11 proposed a new fibrosis prediction model (S-index) for chronic hepatitis B (CHB) in a training cohort of 386 patients, which they confirmed among 146 patients in a validation cohort. The authors constructed an S-index (1000 × gamma-glutamyl transpeptidase [GGT]/ [platelet count × serum albumin2]), which showed an AUROC of 0.812 for significant fibrosis and 0.890 for cirrhosis in the validation cohort. The AUROC of the S-index were higher than those of the Shanghai Liver Fibrosis Group model,13 fibrometer, Forn's index, Hui model,14 Hepascore, and APRI. Using this S-index, biopsy could be avoided in 48% of patients. Although this study showed the superior performance of the S-index for predicting significant fibrosis in CHB and the authors proposed an algorithm for antiviral treatment according to the S-index and ALT level, whether the S-index can also be used as a non-invasive tool to assess treatment response after initiating antiviral treatment in patients with CHB should be further investigated, as the authors acknowledged. Until now, most studies have focused on assessing the performance of non-invasive methods in comparison with histological fibrosis. However, the continuum in development of non-invasive models or devices, including the S-index and TE, for predicting liver fibrosis will be restricted if we rely solely on cross-sectional studies with histology as the reference standard. This is partly because biopsy is an imperfect gold standard. Indeed, comparing AUROC among non-invasive methods in cross-sectional studies based on liver biopsy as a reference is meaningless. The small differences in AUROC do not necessarily mean that one non-invasive model has an inferior performance to that of the other models because whether this difference in the AUROC is due to non-invasive models, liver biopsy, or both is unknown. Furthermore, trying to enhance AUROC up to 1 (perfect concordance with liver biopsy) is pointless, because the inaccuracy of liver biopsy may be responsible for the diagnostic imperfection of a given non-invasive method. Because the perfect gold standard has yet to be determined and a way for improving the accuracy of liver biopsy does not appear to exist, the validation of non-invasive methods through cross-sectional studies is limited. Thus, the performance of non-invasive methods should ultimately be judged and compared by long-term follow-up longitudinal studies using clinical end-points related to liver fibrosis, such as decompensation events, HCC development, or liver-related death.15 However, because these longitudinal studies will take a long time, a new model or device should be tested initially in high-quality cross-sectional studies. Finally, liver fibrosis is a dynamic process. If we can accurately measure it in a non-invasive, serial manner, management strategies for chronic liver disease could be improved and the efficacy of future therapies specifically aimed at reversing liver fibrosis could be validated conveniently. We cannot avoid the heterogeneity among studies due to different prevalence rates in each fibrotic stage resulting in spectrum bias and inapplicability of hospital-based data to a general community. However, longitudinal studies together with higher quality cross-sectional studies may reveal the real performance of a given non-invasive method for predicting liver fibrosis.