This cross-sectional study was conducted in the Department of Medicine and Surgery of Queen Mary Hospital and the Department of Diagnostic and Interventional Radiology of Hong Kong Sanatorium and Hospital, Hong Kong. Adult patients who underwent LT, were regularly followed up, and had a prior valid VCTE assessment, as described in a previous study (19) were screened for eligibility. Patients were excluded if there were contraindications to MR imaging, such as the presence of metallic implants (e.g., non-MRI-compatible pacemaker or cerebral aneurysm clips) and claustrophobia or had been re-transplanted after the first VCTE.

Clinical data, including age, gender, donor type (deceased donor vs. living donor), indication for LT, concomitant medications, and medical comorbidities, were recorded. The study protocol conformed to the ethical guidelines of the 1975 Declarations of Helsinki as reflected in a priori approval by the Institutional Review Board/Ethics Committee of the University of Hong Kong and the Hospital Authority Hong Kong West Cluster (reference number: UW 20–689). Written informed consent was obtained from each patient.

Between October 2020 and September 2023, eligible patients who provided valid written informed consent were recruited. All patients had implant biopsies at the time of transplant, and the fat percentage in the implant biopsy was recorded. Steatotic liver graft (SLG) was defined as a percentage fat content ≥5% on implant biopsy, with 5–32.9% classified as mild steatosis, 33–66% as moderate steatosis, and > 66% as severe steatosis, in accordance with the NASH Clinical Network Scoring System definitions (20, 21). Following recruitment, the anthropometric assessment was performed, including body mass index (BMI) and waist circumference. Patients underwent paired MRI-PDFF, MRE, and VCTE assessments. Fasting glucose and lipid profiles, as well as liver biochemistry, were assessed. Abnormal liver biochemistry was defined as a persistent elevation of alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), or bilirubin at least twice the upper limit of normal. Metabolic risk factors were defined as the presence of at least one of the following: (1) central obesity as shown by waist circumference ≥ 90/80 cm in Asian men and women, (2) high blood pressure ≥ 130/85, (3) hypertriglyceridemia ≥1.7 mmol/L or on lipid-lowering treatment, (4) reduced high-density lipoprotein (HDL) cholesterol <1.0 or 1.3 mmol/L for men or women or on lipid-lowering treatment, and (5) fasting glucose ≥5.6 mmol/L or on anti-diabetic treatment (22). For BMI, Asian-specific cut-off values were adopted as follows: underweight: BMI <18 kg/m², normal: BMI 18–22.9 kg/m², overweight: BMI 23–24.9 kg/m², obese: BMI ≥25 kg/m² (23).

MRE and MRI-PDFF were conducted using a 1.5 Tesla Optima 450 W MR scanner (General Electric, Fairfield, Connecticut). MR operators were blinded to the clinical data of all study participants. The scanning protocol included an MR-sequence software product (IDEAL-IQ; GE Healthcare), which is a three-dimensional volumetric imaging sequence acquired through a single breath-hold. The acquired images were sent to the GE Advantage Window to acquire fat measurements (expressed in percentages) and MR elastography values (expressed in kPa). The arithmetic mean values for fat percentage and elastography readings were calculated using all available data points. PTHS and ≥ F3 were defined as MRI-PDFF ≥5% and MRE ≥3.64 kPa, respectively (24, 25). The total scanning time was approximately 10–15 min per patient (14, 26).

All recruited patients underwent VCTE assessment using the M probe unless otherwise specified. Hepatic fat quantification was assessed by CAP, and a median value (expressed in dB/m) was obtained after ≥10 reliable acquisition, defined as an interquartile range (IQR) <40 dB/m. Liver fibrosis was measured by liver stiffness (LS), and a median value (expressed in kPa) was obtained after ≥10 reliable acquisitions, defined as ≥60% success rate and IQR <30%. Two fully trained and certified operators performed VTCE.

The chi-squared test was used for categorical variables. Continuous variables were presented as median values, with IQR shown in brackets. Variables with skewed distribution were analyzed using the Mann–Whitney test. Those with three or more variables were analyzed using the Kruskal-Wallis test. Paired-related continuous variables were analyzed using the Wilcoxon paired test. Multivariate analysis was performed using bivariate logistic regression on significant univariate variables (defined as those with p-value <0.1). The correlation between two variables was assessed using the Pearson method. Subgroup analysis was performed among patients who were transplanted with a non-steatotic graft. All statistical analyses were performed using SPSS version 27.0 (IBM Corp., Armonk, NY). The Sankey diagram was constructed using Stata, version 17 (StataCorp LLC). A p-value of <0.05 was considered statistically significant.

A total of 292 patients (70.5% male, median age at LT: 51.9, IQR 45.9–56.8) were recruited. The majority of patients (73.6%) underwent transplantation due to hepatitis B virus (HBV)-related complications, which included acute HBV infection, acute flare of chronic hepatitis B, acute-on-chronic liver failure, decompensated cirrhosis, and HBV-related HCC.

Living donor liver transplantation (LDLT) was conducted in 53.8% of patients. The majority (77.4%) did not have excessive hepatic fat on implant biopsy, although one-quarter of patients received an SLG with varying degrees of steatosis at graft implantation. SLG was more common among deceased donor LT (DDLT) than LDLT (31.3% vs. 15.4%, p = 0.002). Diabetes mellitus, dyslipidemia, hypertension, and overweight were present in 11.3, 1.7, 9.9, and 32.9%, respectively, at the time of transplant (Table 1).

MRI was performed at a median time of 12.2 years post-LT (IQR 9.7–15.4 years) when the median age of patients was 64.3 (IQR 58.2–69.6) years old. The median body mass index increased from 21.3 kg/m² at LT to 23.7 kg/m² at recruitment (p < 0.001), with 59.8% being overweight/obese (p < 0.001); the breakdown of BMI groups at both timepoints is shown in Supplementary Figure 2. The proportion of patients having diabetes mellitus, dyslipidemia, and hypertension increased to 38.4, 22.6, and 65.1%, respectively (all p < 0.001). The majority of patients (94.2%) were on tacrolimus, while others were on various immunosuppressants; most of them were receiving single-agent therapy (Table 1). Abnormal liver biochemistry was observed in 19.5% of patients.

The median MRI PDFF was 2.51% (IQR 1.66–5.34%), showing significant inter-segmental differences: S2 2.85% (IQR 1.23–5.02%), S3 2.29% (IQR 1.23–5.02%), S5 2.55% (IQR 1.29–5.30%), S6 2.67% (IQR 1.62–5.33%), S7 2.49% (IQR 1.42–4.99%), and S8 2.90% (IQR 1.73–5.81%). The median MRE was 2.43 kPa (IQR 2.18–2.91 kPa), with similar median values between the upper and lower segments. According to MRI criteria, PTHS and ≥ F3 were present in 27.4 and 10.6%, respectively. The prevalence of PTHS increased with higher BMI categories (Supplementary Figure 3). The prevalence of PTHS increased proportionally with the number of metabolic risk factors (0: 9.1%, 1: 24.1%, 2: 20%, ≥3: 46.9%, p < 0.001; Figure 1).

In comparison, the prevalence of ≥F3 did not demonstrate a consistent trend across BMI categories (Supplementary Figure 3) or metabolic risk factors (0: 21.2%, 1: 4.8%, 2: 10.5%, ≥3: 12.3%, p = 0.069; Figure 1). Patients who were transplanted for HBV-related complications had higher MRI-PDFF than non-HBV-LT (p = 0.023) but were not associated with significant differences in MRE (Supplementary Figure 1). Sex and donor type were not associated with significant differences in MRI-PDFF and MRE (all p > 0.05).

Stratification by fat percentage on implant biopsy revealed that PTHS prevalence increased with the severity of graft steatosis (<5% implant biopsy fat: 23.7% PTHS, 5–32.9% implant biopsy fat: 37.5% PTHS, 33–66% implant biopsy fat: 50% PTHS, >66% implant biopsy fat: 50% PTHS, p = 0.035; Figure 2). However, the percentage of fat on implant biopsy was not associated with the prevalence of ≥F3 (<5% implant biopsy fat: 10.7% ≥ F3, 5–32.9% implant biopsy fat: 12.5% ≥ F3, 33–66% implant biopsy fat: 10% ≥ F3, >66% implant biopsy fat: 0% ≥ F3; p = 0.770).

Figure 3 demonstrates the relationship between implant biopsy fat percentage, PTHS, metabolic dysfunction, and ≥ F3. PTHS occurs not only in patients who received SLG but also in those with minimal fat in the implant biopsy. The presence of ≥F3 was not significantly associated with metabolic risk factors.

Upon multivariate binary logistic regression analysis, SLG (OR 2.067, 95% CI 1.082–3.951, p = 0.028), central obesity (OR 3.952, 95% CI 1.768–8.832, p = 0.001), and hypertension (OR 2.510, 95% CI 1.268–4.966, p = 0.008) were independently associated with excessive hepatic steatosis after LT on MRI-PDFF (Table 2). Similar findings were obtained when the subgroup of patients who did not receive SLG was considered, confirming the association of central obesity (OR 2.630, 95% CI 1.095–6.321, p = 0.031) and hypertension (OR 2.362, 95% CI 1.059–5.267, p = 0.036) as independent risk factors for PTHS (Supplementary Table 1).

For ≥F3, male sex (OR 0.326, 95% CI 0.108–0.981, p = 0.046), change in BMI (OR 0.848, 95% CI 0.742–0.970, p = 0.016), ALT (OR 0.944, 95% CI 0.900–0.990, p = 0.017), AST (OR 1.109, 95% CI 1.046–1.177, p = 0.001), and GGT (OR 1.016, 95% CI 1.006–1.026, p = 0.001) were independent variables in multivariate binary logistic regression (Table 3). In comparison, age, indication of LT, use of SLG (even with graft implant biopsy ≥33% fat), ALP, and metabolic risk factors were not associated with PTHS or ≥ F3 (Table 3). Similar findings were obtained when the subgroup of patients who did not receive SLG was considered, confirming the association of ALT (OR 0.922, 95% CI 0.870–0.978, p = 0.007) and AST (OR 1.153, 95% CI 1.068–1.246, p < 0.001) as independent risk factors for ≥F3 (Supplementary Table 2).

While steroids are known to have steatogenic effects, they were not associated with PTHS in our cohort. Upon sensitivity analysis by excluding long-term steroid users (n = 51), the prevalence of PTHS and ≥ F3 were 28.2 and 10.8%, respectively, which were similar to the main group results. Further analysis of the number of immunosuppressive drugs revealed no association with PTHS or ≥ F3.

Given the greater availability and broader applicability of VCTE compared to MRI, we examined the performance of VCTE in detecting PTHS and ≥ F3 using MRI metrics as the reference standard.

VCTE-CAP moderately correlated with MRI-PDFF (CAP: r = 0.61, p < 0.001). A breakdown of MRI-PDFF demonstrated a better correlation between VCTE and the right lobe (S5: 0.56, S6: 0.54, S7: 0.54, S8: 0.56) than the left lobe (S2: 0.47, S3: 0.52), all p < 0.001 (Figure 4). The area under the receiver-operating characteristic curve (AUROC) for VCTE-CAP in identifying hepatic steatosis was 0.812 (95% CI 0.757–0.868, p < 0.001). By maximizing Youden’s index, a cutoff value of 246 dB/m on VCTE was found to have a sensitivity of 78.5% and a specificity of 70% for identifying hepatic steatosis on MRI-PDFF.

VCTE-LS is highly correlated with MRE at recruitment (r = 0.77, p < 0.001). A breakdown of MRE demonstrated a better correlation between VCTE and the upper segment (r = 0.80) than the lower segment (r = 0.69); both p < 0.001 (Figure 4). The AUROC for VCTE-LS in identifying ≥F3 was 0.890 (95% CI 0.820–0.960, p < 0.001). By maximizing Youden’s index, a cutoff value of 7.75 kPa on VCTE achieved 83.3% sensitivity and 81.6% specificity for identifying ≥F3 on MRE.