Study subjects
A total of 284 patients who underwent liver biopsy between January 2014 and July 2021 at Yokohama City University Hospital and Saga University Hospital were enrolled in the study. All patients were diagnosed with NAFLD and underwent ELF, T4C7S, and VCTE tests within 6 months before or after liver biopsy. Patients with drinking habits (ethanol consumption <30 g/day in men and <20 g/day in women), viral hepatitis (hepatitis B or hepatitis C), drug-induced hepatitis, autoimmune hepatitis, primary biliary cholangitis, sclerosing cholangitis, hemochromatosis, α1-antitrypsin deficiency, or Wilson’s disease were excluded.
Ethics declarations
This study was approved by Yokohama City University Institutional Review Board and Saga University Clinical Research Review Board, and it was conducted according to the ethical guidelines of the Declaration of Helsinki. Written informed consent for participation in the study was obtained from each of the patients.
Clinical parameters
Demographic, anthropometric, clinical, and laboratory data were collected using standard protocols. Insulin resistance was assessed using the homeostasis model assessment of insulin resistance [immunoreactive insulin (IU/mL) × fasting blood sugar (mg/dL)/405] [22]. The ELF score was determined by analyzing serum samples (Siemens Health Care Diagnostic, Tokyo, Japan) and calculated as follows: 2.278 + 0.851 log (hyaluronic acid) + 0.751 log (N-terminal peptide of procollagen III) + 0.394 log (tissue inhibitor of metalloproteinase-1) [8]. The FIB-4 index was calculated as FIB-4=[age (years) × AST (IU/L)]/[platelets (109/L) × √ALT (IU/L)] [23]; NAFLD fibrosis score=[1.675 + 0.037] × [age (years) + 0.094] × [BMI (kg/m2) + 1.13] × [impaired fasting glucose/diabetes (yes=1, no=0) + 0.99] × AST/ALT ratio × 0.013 × platelet count (×109/L) × 0.66 × albumin (g/dL) [24]; AST/ALT ratio=AST (IU/L)/ALT(IU/L); and AST/platelet ratio=[(AST/upper limit of normal range of AST) × 100]/platelet (109/L).
VCTE
VCTE (Fibroscan: EchoSens, Paris, France) was performed by experienced operators for liver stiffness measurement. An M-probe was used for patients with a skin-liver capsule distance of less than 25 mm; otherwise, an XL probe was used [6]. Hepatic steatosis was concurrently evaluated with the controlled attenuation parameter using the same device. Data of liver stiffness measurements and controlled attenuation parameters were obtained within 6 months of the liver biopsy. Details of the measurement method have been described previously [25]. The median of ten valid measurements was defined as liver stiffness measurement and controlled attenuation parameters. Cases with interquartile range/median (M) ≥30% were excluded.
Histological assessment
Ultrasound-guided biopsies were performed with 16-gauge needles. Pathology specimens were stained with hematoxylin-eosin and Masson-trichrome or Azan. Histological features were assessed by expert liver pathologists. The fibrosis stage was determined according to the methods proposed by Kleiner et al. [26] and Brunt et al. [27]. Hepatic steatosis, hepatocyte ballooning, and lobular inflammation were evaluated using the NASH Clinical Research Network scoring system [26]. Advanced fibrosis was defined as fibrosis stage ≥3.
Statistical analysis
Binomial test was used to compare the differences between the 2-step and 3-step approaches for sensitivity, specificity, and proportion diagnosed as high risk. The exact test was used to compare differences between the 2-step and 3-step approaches for PPV and NPV [28]. To simulate the use of each algorithm in a primary care setting with few cases of advanced fibrosis, we constructed a new subpopulation with as few as 3.7% of patients with advanced fibrosis. To maximize the number of patients, all patients without advanced fibrosis were included, and randomly selected patients with advanced fibrosis were added to achieve 3.7%. This is equivalent to the prevalence of advanced fibrosis in the general population of East Asia [5]. In addition, the performance of each approach was compared using the bootstrap method for sensitivity, specificity, PPV, NPV, percentage of patients diagnosed as high risk, and referral rate. All statistical analyses were performed using R version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria) and JMP 9.0.2 (SAS Institute Inc., Cary, NC, USA).
Approaches for predicting advanced fibrosis
In the 2-step approach, first, FIB-4 was assessed in all patients, and VCTE was performed as the second step for patients with values higher than the low cutoff, i.e., intermediate-to-high FIB-4. The 3-step approach was performed in two ways. First, FIB-4 was performed on all patients. Second, ELF or T4C7S was measured for cases with values higher than the low cutoff for FIB-4. Finally, VCTE was performed for cases with values above the cutoff in the second step. Patients with results below the cutoff in each test were considered at low risk, and patients with results above the cutoff in the final step, VCTE, were considered at high risk. The diagnostic performances of these approaches for advanced fibrosis were evaluated by referring to the liver biopsy results.
Medical cost estimation
Medical costs were calculated assuming that the 2-step and 3-step approaches were applied in primary care. The mean medical cost per session for each approach was determined using the bootstrap method. ELF is not covered by the Ministry of Health, Labor and Welfare’s medical fee scale in Japan; therefore, the 3-step approach with ELF was compared with the 2-step approach based on the UK medical costs. In addition, T4C7S is applicable to the Japanese medical fee points; therefore, the 3-step approach with T4C7S was compared with the 2-step approach based on the Japanese medical costs. Supplementary Tables S2 show the costs for each item used in the calculations, and Supplementary Table S3 shows the cost per route for each approach..