Statistical analysis

The primary variable of interest was the LIF score measured using LMS. Patients were stratified according to LIF severity as described above. Secondary variables were liver iron measured by T₂∗ and liver fat measured by ¹H-MRS. Cut-off values derived from our previous study [22] were used to define two severity categories for iron (high liver iron: T₂∗ ⩽12.5 ms, low liver iron: T₂∗ >12.5 ms) and 3 categories for fat fraction (0: fat <1.5%, 1: fat 1.5– <7.5%, 2: fat ⩾7.5%). The cut-off for defining the iron severity categories represents the best threshold for identifying those with any grade of histological iron deposition and corresponds to a dry weight iron 1.7 mg/g. The cut-offs for defining the liver fat categories represent the best thresholds for identifying those with no steatosis (fat fraction <1.5%) and those with steatosis in more than 66% of hepatocytes (steatosis grade 3; fat fraction ⩾7.5%). The primary outcome was the index liver related event as described above (liver related death, HCC, clinically evident ascites, encephalopathy, variceal bleeding). Subgroup analysis including only patients with compensated disease at baseline was also carried out.

Descriptive statistics were used to summarise the subjects’ baseline characteristics for different LIF categories. The Shapiro–Wilks test was used to test for normality of the data across the whole cohort. Analysis of variance was used to test for differences between multiple groups. Kaplan–Meier analysis was used to establish the proportions of patients in each LIF, fat and iron severity category that remained free of liver related events in the follow-up period. Differences between the curves were compared using the log-rank test. Cox regression analysis was used to examine the additive effects of LIF, liver fat and liver iron (measured by T₂∗) on predicting liver related events. The level of statistical significance was set at p <0.05. The analysis was carried out using SPSS v22 (Armonk, NY; IBM Corp) software. The GraphPad Prism software (v6.04) was used to generate the Kaplan–Meier curves for the illustration in the figures.