Non-invasive tests and advanced chronic liver disease in NAFLD: two steps forward and one step back?

We read with great interest the study by Mózes et al on noninvasive tests (NITs) in nonalcoholic fatty liver disease (NAFLD) and commend the authors on their comprehensive individual patient data metaanalysis. The authors have confirmed that a twostep algorithm of fibrosis-4 index (FIB-4) and liver stiffness measurement with vibration controlled transient elastography (LSMVCTE) has superior diagnostic performance for ruling in and ruling out advanced fibrosis and/ or cirrhosis. Indeed, this concept has been demonstrated previously, including in our own work. However, we would like to highlight a few points for clarification. First, the authors suggest a reduction in liver biopsies from 33% to as low as 19% when using the proposed rulein cutoffs for cirrhosis, as summarised in Sankey diagrams. In our opinion, this comparative reduction in biopsies is not justified as the first diagnostic algorithm examines advanced fibrosis, whereas the others are optimised for cirrhosis. Furthermore, this comparison appears misleading as the assumption in the first algorithm is that all patients who are not classified as low risk for advanced fibrosis require biopsy. This is not reflective of realworld clinical practice, even in tertiary care where the prevalence of advanced fibrosis is high. Moreover, the algorithm does not account for the proportion of patients who have overt clinical, laboratory or radiological signs of cirrhosis (with or without portal hypertension), where liver biopsy is not indicated. Our group has demonstrated that only subjecting those classified as indeterminate risk of advanced fibrosis to biopsy in similar algorithms can be an acceptable and costeffective strategy. 4 5 Similarly, the LSMVCTE cutoffs selected by the authors do not take into account the paradigm of compensated advanced chronic liver disease (cACLD) established in the Baveno VI consensus. An LSMVCTE cutoff of ≥15 kPa is highly suggestive of cACLD and identifies those at risk of decompensation and hepatocellular carcinoma. Our group has recently showed that lower cutoffs of ≥12 kPa can in fact rulein cACLD with 92% specificity. Likewise, the cutoffs to rule in cirrhosis proposed by the authors are the same or higher than the wellestablished and validated cutoff of ≥20 kPa for clinically significant portal hypertension (CSPH). This cutoff forms the basis of the Baveno VI variceal screening criteria, which have been validated in NAFLD. This suggests that the patients ruled in as cirrhotic in this study are likely to have CSPH and that the proposed cirrhosis cutoffs have no additional value in terms of patient management. Finally, the clinical utility of the authors’ findings should be appropriately framed. In the diagnostic algorithms presented, the false negative rate is 10%. This is suboptimal in terms of missed diagnoses of cirrhosis at a prevalence of 11%. Furthermore, there is a high rate of indeterminate results. The goal in terms of the clinical utility of NITs is to accurately identify those at risk of liverrelated complications, while reconciling an acceptable number of false negative diagnoses with false positives and/or indeterminate results. This may require additional clinical data to be analysed with histology. Furthermore, decision curve analysis may aid this process, as we have demonstrated. 3 9 Although the study byMózes et al reaffirms that twostep algorithms are better than one, there has been limited progress forward for NITs in NAFLD in the current paradigm of cACLD. We believe that the twostep algorithm is optimal for patients in primary care where the prevalence of advanced fibrosis is relatively low. In populations with a higher prevalence of advanced fibrosis/cirrhosis encountered in secondary care, FIB-4 and NAFLD Fibrosis Score most likely have lower utility, and the dual cutoff LSMVCTE for cACLD should be used. The authors have a unique opportunity to test this using their data.