IDDF2024-ABS-0025 A novel 2-step approach combining the SAFE score and the MR-MASH score for predicting at-risk NASH

<h3>Background</h3> Identifying at-risk nonalcoholic steatohepatitis (defined as NASH and ≥ stage 2 fibrosis) in patients with NAFLD is critical. The steatosis-associated fibrosis estimator (SAFE) is a novel non-invasive test designed to stage significant fibrosis (F ≥2) and the MR-MASH score is an image biomarker proposed for predicting NASH. We aim to evaluate a 2-step approach combining the SAFE and MR-MASH scores in predicting at-risk NASH. <h3>Methods</h3> We retrospectively analyzed 286 patients with NAFLD undergoing contemporaneous laboratory tests, MRI and liver biopsy. The SAFE, FIB-4, and the MR-MASH scores were calculated. Based on the suggested cutoffs and their intended targets, algorithms combining the SAFE or FIB-4 and MR-MASH were developed from a training set and tested in a validation set. <h3>Results</h3> In the entire cohort, SAFE outperformed FIB-4 in staging significant fibrosis (AUCs =0.798 <i>vs.</i> 0.744, <i>P</i> &lt;0.01). The MR-MASH score predicted NASH with an AUC =0.803. Two-step approaches with serial MRI examination performed in patients with indeterminate-high SAFE or FIB-4 showed 90.0% and 78.6% PPVs to identify NASH. In the training set with 172 patients, the percentages of misclassifications concerning the 2-step approaches applying the MR-MASH score for patients with indeterminate-high SAFE or FIB-4 to predict at-risk NASH were 8.7% and 11.6%. The percentages of uncertain areas were 5.8% and 3.5%, respectively. In the validation cohort with 114 patients, the percentages of misclassifications were 7.9% and 14.9%, while the percentages of uncertain areas were 6.1% and 2.6%, respectively. In the entire cohort, the accuracy, wrong classification and uncertain area of the serial combination of the SAFE and MR-MASH scores for predicting at-risk NASH were 85.7%, 5.9% and 8.4%, respectively (IDDF2024-ABS-0025 Figure 1 (A)). The accuracy, wrong classification and uncertain area of the serial combination of the FIB-4 and MR-MASH scores for predicting at-risk NASH were 83.9%, 3.1% and 12.9%, respectively (IDDF2024-ABS-0025 Figure 1 (B)). <h3>Conclusions</h3> The serial combination of SAFE and MR-MASH performs well in predicting at-risk NASH.