1 INTRODUCTION

Liver fibrosis is the most important prognostic factor in patients with non-alcoholic fatty liver disease (NAFLD).1 Since NAFLD is rapidly increasing and is already the most common cause of chronic liver disease worldwide2, 3 and one of the main indications for liver transplantation,4, 5 reliable methods to stage liver fibrosis in NAFLD are an urgent need in clinical practice. Liver biopsy remains the reference standard to characterize non-alcoholic steatohepatitis (NASH).6 Nevertheless, liver biopsy is an invasive and expensive procedure associated with patient discomfort and sampling variability.7, 8 Elastography techniques, such as transient elastography (TE) and two-dimensional shear wave elastography (2D-SWE), provide a physical measure of liver stiffness which is closely related to fibrosis in chronic liver disease,9 and have emerged as an alternative to liver biopsy.

Although TE has been validated in NAFLD as a method to identify and stage fibrosis, it has been shown that steatosis and inflammatory activity might influence the accuracy of liver stiffness measurements (LSM) to predict fibrosis.9 However, the data regarding the impact of steatosis on LSM are still controversial. Some studies show no association,10-12 one study showed that severe steatosis leads to an overestimation of liver fibrosis by LSM assessed using TE,13 and another one showed that steatosis leads to an underestimation of liver fibrosis by LSM assessed by TE.14 To date, only one study evaluated the influence of steatosis and inflammation on LSM using 2D-SWE in NAFLD,15 and no head-to-head study compared TE and 2D-SWE vs histology with the aim of addressing which of the two methods is mostly influenced by confounders (steatosis and inflammation).

This study aimed to assess whether histological steatosis and inflammatory activity in NAFLD patients affects the accuracy of LSM by two different ultrasound elastography methods (TE and 2D-SWE) in predicting fibrosis.

2 PATIENTS AND METHODS 2.1 Study population

We prospectively included consecutive adult patients with NAFLD who underwent liver biopsy from August 2018 through September 2020 who had LSM performed using TE and 2D-SWE within 2 months from liver biopsy at an academic tertiary centre. Exclusion criteria were as follows: liver disease of other aetiology (chronic hepatitis B or C, autoimmune hepatitis, primary biliary cholangitis, primary sclerosing cholangitis, genetic hemochromatosis, drug-induced hepatotoxicity, a-1 antitrypsin deficiency, Wilson's disease, etc); exposure to drugs that can cause secondary NAFLD (corticosteroids, amiodarone and tamoxifen); significant alcohol consumption (>3 standard drinks/day in men and >2 drinks/day in women, or binge drinking defined as >5 standard drinks in men and >4 in women over a 2-hour period)3; ALT >5 times the upper limit of normality; and refusal of consent to further use of personal health-related data for research.

The study was performed according to the principles of the Declaration of Helsinki, and approval was obtained from the local ethics committee (KEK BE 2018-00487).

2.2 Clinical and laboratory assessment

Demographic, clinical, anthropometric and laboratory data were collected at the time of the biopsy. Obesity was defined as body mass index (BMI) ≥30 kg/m2. Diabetes mellitus was defined as a fasting glycaemia ≥7 mmol/L or an HbA1c ≥6.5%, or current antidiabetic treatment. The presence of arterial hypertension and dyslipidaemia was recorded from the clinical charts.

2.3 LSM by transient elastography (TE)

Transient elastography (TE 502 Touch; Echosens, Paris, France) provided with M and XL probe was used to assess LSM, following the EASL-ALEH clinical practice guidelines (M probe in patients with BMI <30 kg/m2 and XL probe in obese patients and/or skin-to-capsule distance ≥25 mm). Measurements were performed in a fasting state during a routine visit at our hepatology outpatient facility. Only patients with 10 valid measurements were included in the study. IQR/M values ≥.30 were considered unreliable; no valid shots were considered as failures of the technique.

2.4 LSM by two-dimensional shear wave elastography

LSM by 2D-SWE was performed using the Aixplorer ultrasound system (SuperSonic Imagine SA, Aix-en-Provence, France). Patients were placed in a supine position, with the right arm in extension. The operator selected a region of the right lobe of the liver with good spatial resolution for B-mode ultrasound imaging, free of large vascular structures and at least 15 mm below the capsule through a right intercostal space, and during breath hold activated 2D-SWE. Once a colour map with complete and homogeneous filling was obtained in the assessment area, a region of interest 15 mm in diameter was positioned in the centre of the colour map to measure stiffness using the Q box tool. We obtained three successful and valid measurements for each patient and used the mean value and the standard deviation of these measurements as liver stiffness measure.16 Variability (SD) over 30% of the average liver stiffness value was considered as unreliable measurement.17 When the operator obtained little or no signal in the region of interest for all acquisitions, the measurements were defined as failures.

2.5 Histological assessment

Liver biopsies were assessed by one experienced pathologist. NAFLD severity was scored according to the SAF scoring system6 that evaluates individually the grade of steatosis, activity and fibrosis. The steatosis grade was classified by the percentage of hepatocytes containing large- and medium-sized intracytoplasmic lipid droplets, on a scale of 0-3 (0: <5%; 1: 5%-33%; 2: 34%-66%; and 3: >67%). The grade of inflammatory activity was rated from A0 to A4 by addition of grades of ballooning and lobular inflammation, each graded from 0 to 2. Ballooning of hepatocytes was defined as the presence of hepatocyte clusters with a round shaped and pale cytoplasm (0: normal hepatocytes; 1: ballooning but normal size; and 2: ballooning with at least one enlarged ballooned hepatocyte). Lobular inflammation was defined as a focus of two or more inflammatory cells within the lobule organized either as microgranulomas or located within the sinusoids. Foci were counted at 20× magnification (grade 0: none; 1: ≤2 foci per lobule; and 2: >2 foci per lobule). The stage of fibrosis (F, from F0 to F4), was assessed according to the NASH Clinical Research Network staging system, with the single modification of pooling the three substages (1a, 1b and 1c) into a single F1 score. The diagnosis of NASH had >5% steatosis in hepatocytes and a grade of activity A ≥ 2.6, 18 Moreover, the NAFLD activity score (NAS) was calculated. NAS is the unweighted sum of steatosis grade, lobular inflammation and ballooning, ranging from 0 to 8 according to the grades of steatosis (0-3), lobular inflammation (0-3) and hepatocellular ballooning (0-2).19 Interpretability for liver biopsy was based on the standard criteria of length, number of portal tracts (>10) and lack of major fragmentation.

2.6 Statistical analysis

Continuous variables are described as mean ± standard deviation, categorical variables as a number of cases (percentage). These descriptive statistics are provided for the complete group (104 patients) and the subgroups with valid LSM by TE and by 2D-SWE. The subgroups of patients with and without valid SWE were compared using Welch's t-test where parametric assumptions are adequate. For ordinal variables, Wilcoxon's rank-sum test was used. Categorical variables were compared using the chi-squared test. Missing values on all predictor variables were handled using Multiple Imputation by Chained Equations with M = 20 imputations. The two non-invasive tests, LSM by TE and 2D-SWE, were not used as predictors when creating imputations for other variables and were not imputed themselves. Bivariate associations between patient demographic characteristics, laboratory and liver histology parameter, and the two non-invasive tests, LSM by TE and LSM by 2D-SWE, were calculated using pairwise Spearman's correlations on multiply imputed data.

We performed a model comparison against the baseline model with fibrosis stage on Natural log (ln) LSM by TE and ln LSM by 2D-SWE. Covariates as well as their interactions with fibrosis stage were added to the baseline model one-by-one. The extended models were compared to the baseline models using an F-test to determine whether the addition of the covariate would add explained variance to the baseline model. All model comparisons were adjusted for multiple testing using the method by Holm. Ordinal variables were coded as treatment contrasts. Possible main effects or interactions of the covariate were then only further examined if the model explained significantly more variance than the baseline model. In addition, one more complex model was calculated for LSM by TE, which included patient characteristics as control variables as well as all significant covariates from the prior analyses. All reported P-values are two-sided and values <.05 were considered statistically significant. All confidence intervals are at the 95% level. Statistical analyses were conducted using R open-source statistical software, version 4.0.3.

3 RESULTS 3.1 Characteristics of the study population

One hundred and four patients undergoing liver biopsy to grade and stage NAFLD were evaluated using liver histology, relevant laboratory parameters and LSM by TE and 2D-SWE. One hundred and two had reliable LSM by TE and 88 had reliable LSM by 2D-SWE. The differences between the 88 patients with reliable LSM by 2D-SWE and the 16 patients without reliable LSM by 2D-SWE are shown in Table S1. Eighty-six patients had reliable LSM using both TE and 2D-SWE methods; a flowchart of patient inclusion and exclusion is provided in Figure S1. Baseline characteristics of the whole population and a comparison of the cohorts using TE and 2D-SWE are shown in Table 1.

TABLE 1. Baseline features of these 104 patients and the comparison of the cohorts using transient elastography (TE) and two-dimensional shear wave elastography (2D-SWE) Characteristics Overall (n = 104) LSM by TE cohort (n = 102) LSM by 2D-SWE cohort (n = 88) Age – y 53.4 ± 12.6 53.4 ± 12.6 53.2 ± 12.7 Sex, female, n (%) 43 (41.3) 43 (42.2) 33 (57.5) BMI – kg/m2 30.9 ± 7.2 32.3 ± 7.2 31.6 ± 7.1 BMI – kg/m2, n (%) <25 8 (7.6) 8 (9.3) 8 (9.4) 25-29.9 37 (35.5) 32 (37.2) 32 (37.6) ≥30 59 (56.7) 47 (53.5) 45 (52.9) Diabetes mellitus, n (%) 48 (47.1) 47 (46.1) 39 (44.3) Arterial hypertension, n (%) 54 (52.0) 52 (51.0) 45 (51.0) Dyslipidaemia, n (%) 54 (52.5) 53 (52.0) 45 (51.0) ALT – IU/L 75.7 ± 45 80.7 ± 57.2 84.4 ± 58.4 AST – IU/L 63.7 ± 41.1 62.6 ± 40.2 65.0 ± 41.4 AST/ALT ratio 0.9 ± 0.4 0.9 ± 0.4 0.9 ± 0.4 Bilirubin – µmol/L 10.8 ± 6.9 10.6 ± 6.8 10.8 ± 6.4 GGT – IU/L 166 ± 240 164 ± 242 175 ± 258 Cholesterol – mmol/L 4.8 ± 1.21 4.8 ± 1.2 4.8 ± 1.20 HDL – mmol/L 1.1 ± 0.3 1.1 ± 0.3 1.1 ± 0.3 Triglycerides – mmol/L 2.2 ± 1.6 2.2 ± 1.6 2.2 ± 1.7 Glucose – mmol/L 6.3 ± 2.2 6.3 ± 2.2 6.2 ± 2.1 Insulin – mU/L 31.7 ± 29.1 31.7 ± 29.6 31.9 ± 31 Platelet count – g/L 220.8 ± 70.8 220 ± 70 223.4 ± 70.6 Albumin – g/L 38.8 ± 5.0 38 ± 4.9 38.7 ± 3.5 CAP value – (dB/m) 321.0 ± 47.6 320.3 ± 47.6 316.7 ± 46.8 LSM (TE) – kPa 11.9 ± 8.0 11.9 ± 7.9 11.1 ± 5.4 LSM (2D-SWE) – kPa 9.5 ± 4.4 9.5 ± 4.4 9.4 ± 4.3 Histology at biopsy Steatosis grade S0 (<5%) 0 (0) 0 (0) 0 (0) S1 (5%-33%) 27 (26.0) 27 (26.5) 19 (22.6) S2 (34%-66%) 31 (29.8) 29 (28.4) 29 (33.0) S3 (>66%) 46 (44.2) 46 (45.1) 40 (45.5) Steatosis % 55.3 ± 26.6 55.2 ± 26.8 57.3 ± 25.4 Activity A0 9 (8.7) 9 (8.8) 8 (9.1) A1 9 (8.7) 9 (8.8) 7 (8.0) A2 63 (60.6) 61 (59.8) 55 (62.5) A3 20 (19.2) 20 (19.6) 15 (17.0) A4 3 (2.9) 3 (2.9) 3 (3.4) Fibrosis stage F0 11 (10.6) 11 (10.8) 10 (11.4) F1 13 (12.5) 12 (11.8) 13 (14.8) F2 35 (33.7) 35 (34.3) 29 (33.0) F3 37 (35.6) 36 (35.4) 30 (34.1) F4 8 (7.7) 8 (7.8) 6 (6.8) Ballooning stage 0 14 (13.5) 14 (13.7) 12 (13.6) 1 71 (68.3) 69 (67.6) 62 (70.5) 2 18 (17.3) 18 (17.6) 13 (14.8) 3 1 (1.0) 1 (1.0) 1 (1.1) Lobular infiltration 0 12 (11.5) 12 (11.8) 10 (11.4) 1 82 (78.8) 80 (78.4) 71 (80.7) 2 10 (9.6) 10 (9.8) 7 (8.0) NAS score 0-2 13 (12.6) 13 (12.9) 10 (11.5) 3-4 78 (75.7) 76 (75.2) 67 (77.0) 5-8 12 (11.7) 12 (11.9) 10 (11.5) Note Data are given as mean ± standard deviation or as number of cases (percentage). Abbreviations: ALT, alanine aminotransferase; BMI, body mass index; GGT, gamma-glutamyl transpeptidase; HDL, high-density lipoprotein; HOMA, homeostasis model assessment; IU, international units; kPa, kilopascal; y, years.

The M probe and XL probe were used in 38.4% and 61.5% of patients respectively. Among XL probe patients, the BMI >30 kg/m2 criteria was used in 92%, and the skin-to-capsule distance criteria ≥25 mm was used in 8%. Technical failure to measure liver stiffness occurred in none of the cases with TE and in 15 of the cases (13%) with 2D-SWE because of the inability to obtain an adequate signal for the acquisitions. Unreliable LSM was observed in two cases (1.9%) with TE and in one case (0.9%) with 2D-SWE. Failure or unreliable results of LSM with 2D-SWE were associated with higher BMI (Spearman's correlation coefficient [ρ] = 0.32; 95%CI, 0.14-0.49; P < .001), higher values of LSM by TE (ρ = 0.22; 95%CI, 0.03-0.40; P = .02) and higher values of CAP (ρ = 0.21; 95%CI, 0.02-0.39; P = .03).

LSM by TE and 2D-SWE was correlated with the histological fibrosis stage (Figure 1). Notably, the correlation of histological fibrosis stage with liver stiffness was stronger for 2D-SWE (ρ = 0.71; 95%CI, 0.59-0.79; P < .001) than for TE (ρ = 0.52; 95%CI, 0.37-0.65; P < .001; Z = 2.21; P = .02). LSM by TE and 2D-SWE was strongly correlated with each other (ρ = 0.64; 95%CI, 0.50-0.74; P < .001).

Spearman's correlations for patient characteristics, liver histology, laboratory parameters and liver stiffness by two non-invasive tests (2D-SWE and TE). Of note, the correlation of histological fibrosis stage with liver stiffness was stronger for 2D-SWE than for TE. *P < .05; **P < .01

Table 2 shows the diagnostic performance of LSM by TE cut-off values for fibrosis described recently by Eddows et al for NAFLD patients,11 namely 8.2, 9.7 and 13.6 kPa for each fibrosis stage F ≥ F2, F ≥ F3 and F = F4 respectively. LSM by TE had good sensitivity and specificity with a good PPV (0.88) for ≥F2 and an excellent NPV (0.98) for F4. The discriminative capacity between F0-F2 vs F3 was lower with an AUROC of 0.72 (0.63-0.82) at a threshold of 9.7 kPa. False negative rate of LSM by TE for significant fibrosis (≥F2) was 16%. The diagnostic performance of LSM by 2D-SWE using the cut-off proposed for NAFLD patients20 is also detailed in Table 2. It showed good diagnostic performance for fibrosis stages F ≥ F2 and F = F4. The discriminative ability of LSM by 2D-SWE was numerically higher than by TE for distinguishing F0-F2 vs F3 with an AUROC of 0.84 (0.76-0.92) at a threshold of 9.2 kPa.

TABLE 2. Diagnostic performance of LSM by TE and 2D-SWE for each fibrosis stage LSM by TE F ≥ F2 F ≥ F3 F = F4 AUROC (95%CI) 0.76 (0.64-0.88) 0.72 (0.63-0.82) 0.89 (0.78-1.00) Cut-off (kPa) 8.2 9.7 13.6 Se (95%CI) 0.83 (0.75-0.89) 0.73 (0.64-0.81) 0.87 (0.79-0.92) Sp (95%CI) 0.62 (0.52-0.71) 0.53 (0.43-0.62) 0.77 (0.68-0.84) PPV (95%CI) 0.88 (0.80-0.93) 0.55 (0.46-0.64) 0.24 (0.16-0.33) NPV (95%CI) 0.53 (0.44-0.62) 0.72 (0.62-0.79) 0.98 (0.94-0.99) FPR (95%CI) 0.37 (0.28-0.47) 0.46 (0.37-0.56) 0.22 (0.15-0.31) FNR (95%CI) 0.16 (0.10-0.24) 0.26 (0.18-0.35) 0.12 (0.07-0.20) LSM by 2D-SWE AUROC (95%CI) 0.83 (0.72-0.93) 0.84 (0.76-0.92) 0.94 (0.89-0.99) Cut-off (kPa) 7.1 9.2 13 Se (95%CI) 0.86 (0.79-0.91) 0.65 (0.64-0.81) 0.83 (0.75-0.89) Sp (95%CI) 0.73 (0.64-0.81) 0.86 (0.78-0.91) 0.87 (0.80-0.92) PPV (95%CI) 0.90 (0.83-0.94) 0.78 (0.69-0.84) 0.33 (0.25-0.42) NPV (95%CI) 0.65 (0.55-0.73) 0.77 (0.62-0.79) 0.98 (0.94-0.99) FPR (95%CI) 0.26 (0.18-0.35) 0.13 (0.08-0.21) 0.33 (0.25-0.42) FNR (95%CI) 0.13 (0.08-0.21) 0.34 (0.25-0.43) 0.16 (0.10-0.24) Abbreviations: FNR, false negative rate; FPR, false positive rate; NPV, negative predictive value; PPV, positive predictive value; Se, sensitivity; Sp, specificity.

The AUROC and false positive rate for LSM by TE and 2D-SWE for diagnosing histological fibrosis stage according to the histological inflammatory activity stage are summarized in Table S2.

3.2 Potential confounders of the association between fibrosis and liver stiffness measured by TE and 2D-SWE

We calculated a baseline linear regression model of the fibrosis stage on the LSM. To assess whether the histological steatosis, inflammatory activity, NAS score and laboratory parameters would influence the association between fibrosis and LSM by TE or 2D-SWE, we compared all further models against this baseline model.

3.2.1 Influence of inflammation and steatosis on LSM by TE

The fibrosis stage on histology explained 35% of the variance of LSM by TE (P < .001). The variance explained by the model did not increase significantly compared to the baseline model after including the main effect of steatosis percentage and the interaction of steatosis percentage and fibrosis stage ( = .04; P = 1) (Table 3). When including inflammatory activity and the interaction between inflammatory activity and histological fibrosis stage, the model explained 25% more variance in LSM by TE (P < .01). After controlling for fibrosis, age, sex and BMI, both the main effect of inflammatory activity (P < .001) and the interaction effect between inflammatory activity and fibrosis (P = .01) remained significantly associated with LSM by TE, and independently explained 11% and 13% of variance in LSM by TE respectively (Table 4). This indicates that the strength of the association between fibrosis on histology and LSM by TE changes according to the severity of the inflammatory activity. This effect is visualized in Figure 2. Regression results for the individual combinations of fibrosis stage and grade of inflammatory activity are provided in Tables S3 and S4.

TABLE 3. Model comparisons against the baseline model with fibrosis stage on the liver stiffness measurement (LSM) by transient elastography (TE) and by two-dimensional shear wave elastography (2D-SWE) Outcome LSM by TE LSM by 2D-SWE Model F (df) P-value F (df) P-value Baseline: Fibrosis stage 12.81 (4, 85) .35 <.001 22.15 (4, 81) .52 <.001 +Steatosis (%), interaction 1.25 (5, 88) .04 1 0.15 (5, 74) .00 1 +Inflammatory activity, interaction 3.56 (20, 81) .25 <.01 0.78 (20, 68) .06 1 Note P-values Holm-adjusted for multiple comparisons. F is the ratio of explained variance between the baseline model and those with added predictors. TABLE 4. Logistic regression models with liver stiffness measurement (LSM) by transient elastography (TE) as the dependent variable Predictor (df) Model 1 Model 2 Model 3 F η2 P-value F η2 P-value F η2 P-value Fibrosis stage (4) 12.81 0.37 <.001 17.54 0.35 <.001 18.36 0.35 <.001 Inflammatory activity (4) 5.76 0.11 <.001 6.04 0.11 <.001 Inflammatory activity 2.68 0.13 <.01 2.70 0.13 <.01 Fibrosis score (10) Age (1) 4.32 0.02 .03 BMI (1) 3.69 0.02 .05 Gender (1)