Abstract
Background: The current definition of metabolic dysfunction-associated steatotic liver disease (MASLD) relies on a classical assessment of steatosis via liver biopsy, with grades S0-S3 (5-100% fat) potentially underestimating low-grade steatosis. We propose a new, more sensitive classification based on magnetic resonance imaging-proton-density fat fraction (MRI-PDFF), splitting the existing S0 and S1 grades into three classes: new-S0, very early S1 (S1A), and later S1 (S1B). We aimed to determine whether these early S1A/S1B phenotypes differed clinically or biologically from the new-S0 grade using large population cohorts. Methods: We assessed the prevalence of the new MRI-PDFF-based grades in 29,252 healthy participants from the UK Biobank discovery cohort, 286 outpatients with type 2 diabetes, and in six previously published databases (N = 149,212) using SteatoTest-2 or a proxy. We performed a multimodal assessment of steatosis using longitudinal MRI-PDFF and liver biopsy data (N = 286). Models were used to adjust for phenotypes and overall mortality, controlling for age, sex, and cardiometabolic factors. Results: In the UK Biobank cohort, the prevalences of the new-S0, S1A, and S1B grades were 54%, 26%, and 17%, respectively. Grades S1A and new-S0 were most discriminated by triglycerides (odds ratio [OR]: 2.40, 95% confidence interval [CI]: 2.07-2.77, p < 0.00001) and body mass index (BMI; OR: 1.30, 95% CI: 1.27-1.33, p < 0.00001), and grades S1A and S1B were most discriminated by triglycerides, BMI, systolic blood pressure (SBP), and glycated hemoglobin (HbA1c). Adjusting for age, sex, SBP, BMI, HbA1c, triglycerides, and high-density lipoprotein-cholesterol) revealed significantly lower 15-year survival in the high-risk group (97.2%, 95% CI: 96.9-97.7) versus the low-risk (99.4%, 95% CI: 99.2-99.6) group (p < 0.00001). Conclusions: The early trajectory of liver steatosis is undetectable in 26% of middle-aged adults. This early steatosis phenotype differs clinically and biologically from the new-S0 grade in large population cohorts.</p>