Letter: non‐invasive prediction models to exclude cirrhosis in NAFLD—not everyone fits the mould. Authors' reply

We thank Giri et al for their interest in our recent manuscript.1 They raised some interesting points regarding our study sample and the generalizability of our proposed cutoffs for FIB4 and the NAFLD Fibrosis Score (NFS) in different patient populations.2 We agree that nonalcoholic fatty liver disease (NAFLD) is a dynamic disease that can improve over time with dietary change, exercise and weight loss and that these changes may have occurred in the time interval between the liver biopsy and entry into the cohort. The amount of weight loss that has been associated with improvement in the fibrosis stage is 10%.3 Unfortunately, no more than 10%– 12% of patients will successfully experience this degree of weight loss, even in the setting of a structured weightloss program.3,4 We expect that very few patients would experience substantial weight loss, therefore, few would be reclassified from having cirrhosis to lower fibrosis stages in the year between liver biopsy and cohort enrollment. Giri et al are correct in noting that older age and diabetes may influence the performance of FIB4 and NFS. However, we observed that the AUROC, sensitivity and specificity in the UK cohort were almost identical to the US cohort for FIB4 (0.87 vs 0.86, 0.78 vs 0.80 and 0.76 vs 0.78) and NAFLD fibrosis score (0.89 vs 0.84, 0.68 vs 0.70, and 0.89 vs 0.86) despite the UK cohort having an older mean age and higher prevalence of diabetes. PPV were slightly higher for NAFLD fibrosis score (0.46 in UK vs 0.35 in US) and similar for FIB4 (0.31 vs 0.29). Whilst some studies have reported decreased accuracy of FIB4 in patients with diabetes,5 FIB4 has been associated with an increased risk of liverrelated morbidity and mortality in a cohort of patients with an even higher prevalence of diabetes than in our study.6 In our derivation and validation cohorts, the mean BMI was 34–­35 kg/m2,­ with­most­ patients­ having­ BMI > 30­ and­ ~15% with BMI >40, which is representative of the general NAFLD population. Our study was not powered to investigate the performance of our cutoffs in subgroups such as patients with diabetes or class III obesity. We recognise age is an important predictor of cirrhosis in patients with NAFLD, and that accurate identification of highrisk patients is imperative. Shah et al7 reported a FIB4 cutoff of 1.3 as having 90% sensitivity to rule out cirrhosis. Our newly identified cutoff of 1.67 is in fact higher, yet lower than the suggested 2.0 cutoff in patients who are older than 65.8 Further work must be done to identify the most appropriate FIB4 cutoff in older patients, as the cutoff of 2.0 has not yet been wellvalidated. The use of a single cutoff, as proposed in our study, allows us to confidently rule out cirrhosis, but emphasises the need for additional testing to diagnose cirrhosis. Giri et al's comments highlight the need for further validation of our newly proposed cutoffs, particularly in patients older than 65, as well as those with diabetes and/or class III obesity.