Performance of Noninvasive Tests of Fibrosis Among Asians, Hispanic, and non-Hispanic Whites in the STELLAR Trials

Performance of Noninvasive Tests of Fibrosis Among Asians, Hispanic, and non-Hispanic Whites in the STELLAR Trials Vincent Wai-Sun Wong,*Won Young Tak, ‡ George Boon Bee Goh, § Pin-Nan Cheng, jj Eric J. Lawitz, ¶ Zobair M. Younossi, # Raj Vuppalanchi,** Ziad Younes, ‡‡ Naim Alkhouri, §§ Lulu Wang, jjjj Jialuo Liu, jjjj Kathryn Kersey, jjjj Robert P. Myers, jjjj Stephen A. Harrison, ¶¶ Zachary Goodman, # Michael Trauner, ## Manuel Romero-Gomez,*** Quentin M. Anstee, ‡‡‡ Mindie H. Nguyen, §§§ and Takeshi Okanoue jjjjjj *Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong; ‡ School of Medicine Kyungpook National University, Kyungpook National University Hospital, Daegu, Republic of Korea; § Department of Gastroenterology and Hepatology, Singapore General Hospital, Singapore; jj Department of Internal Medicine, National Cheng Kung University Hospital, Tainan, Taiwan; ¶ Texas Liver Institute, University of Texas Health San Antonio, San Antonio, Texas; # Inova Fairfax Hospital, Falls Church, Virginia; **Indiana University Medical Center, Indianapolis, Indiana; ‡‡ Gastro One, Germantown, Tennessee; §§ Arizona Liver Health, Chandler, Arizona; jjjj Gilead Sciences, Inc, Foster City, California; ¶¶ Pinnacle Clinical Research, San Antonio, Texas; ## Division of Gastroenterology and Hepatology, Department of Internal Medicine III, Medical University of Vienna, Vienna, Austria; ***Hospital Universitario Virgen del Rocio, Sevilla, Spain; ‡‡‡ Translational and Clinical Research Institute, Faculty of Medical Sciences, Newcastle University, Newcastle-upon-Tyne, UK; Newcastle NIHR Biomedical Research Centre, Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle upon Tyne, UK; §§§ Division of Gastroenterology and Hepatology, Stanford University Medical Center, Palo Alto, California; and jjjjjj Saiseikai Suita Hospital, Suita City, Osaka, Japan BACKGROUND & AIMS: The effect of race on routinely available noninvasive tests of fibrosis is incompletely understood. This study evaluated the performance of noninvasive tests among white and Asian patients in the STELLAR trials (NCT03053050 and NCT03053063), which evaluated selonsertib in patients with advanced (F3-F4) fibrosis due to nonalcoholic steatohepatitis (NASH). METHODS: Baseline liver biopsies were centrally read using the NASH Clinical Research Network system, and 4 noninvasive tests (Nonalcoholic fatty liver disease fibrosis score [NFS], Fibrosis-4 index [FIB-4], Enhanced Liver Fibrosis test [ELF], and liver stiffness by vibration-controlled transient elastography) were measured. The performance of these tests to discriminate advanced fibrosis was evaluated using areas under the receiver operating characteristics curves with 5-fold crossvalidation repeated 100 times. RESULTS: Among 3207 patients screened with evaluable liver histology, 2281 were whites and 762 were Asians. Seventy-two percent of whites and 67% of Asians had advanced fibrosis. The areas under the receiver operating characteristics curves of the noninvasive tests for advanced fibrosis were similar in whites and Asians: 0.73 and 0.75 for NFS, 0.78 and 0.80 for FIB-4, 0.79 and 0.81 for ELF, and 0.80 and 0.83 for liver stiffness, respectively. At the published cutoffs, the tests had similar sensitivities and specificities in the 2 groups. However, the sensitivities of NFS, FIB-4, and ELF were low in both white and Asian patients younger than 40 years. CONCLUSIONS: In the global phase III STELLAR trials, the diagnostic performance of routinely available noninvasive tests for the detection of advanced fibrosis due to NASH was acceptable and similar between white and Asian patients. Keywords: Cirrhosis; Liver Histology; NIT. Abbreviations used in this paper: AUROC, area under the receiver operating characteristics curve; BMI, body mass index; CI, confidence interval; CV, cross-validation; ELF, Enhanced Liver Fibrosis panel; FIB-4, Fibrosis-4 index; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; VCTE, vibration-controlled transient elastography. © 2022 by the AGA Institute. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons. org/licenses/by-nc-nd/4.0/). 1542-3565 https://doi.org/10.1016/j.cgh.2022.01.015 Clinical Gastroenterology and Hepatology 2022;-:-–- Nonalcoholic fatty liver disease (NAFLD) is the most common cause of chronic liver disease, affecting at least a quarter of the global adult population. 1 NAFLD has also become increasingly common among Asians. 2 Systematic reviews have reported NAFLD in 33.9% of Asians in 2012 to 2017 and 32.9% of Chinese in 2018. 3–5 Although Asian patients with NAFLD tend to have less severe disease, 6 Asia is expected to harbor the biggest number of patients with cirrhosis and hepatocellular carcinoma secondary to nonalcoholic steatohepatitis (NASH) by 2030. 7,8 Because only a fraction of patients with NAFLD will progress to liver-related morbidity and mortality, noninvasive tests as initial assessment are preferable. 9 Additionally, most phase IIb and III studies for NASH currently utilize histology in enrollment criteria and primary endpoints. 10 The high screen failure rate typical of these studies means patients are subject to an unnecessary invasive procedure; it also adds costs to drug development. Therefore, it is important to understand the performance of noninvasive tests not only in routine care but also in clinical trial settings. Historically, most noninvasive tests were derived and validated in whites. There are reasons to suspect that the tests may perform differently in Asians. Asians tend to develop NAFLD and metabolic complications at a lower body mass index (BMI), and lean or nonobese NAFLD is more often described in Asians. 11–13 This may particularly affect scores that include BMI as a component (eg, the NAFLD fibrosis score). 14,15 This is highly relevant because nonobese NAFLD was reported in about 40% of the global NAFLD population. 16 Likewise, some studies showed that liver stiffness measurement by vibrationcontrolled transient elastography (VCTE) might be less accurate in patients with extreme BMI. 17–19 Other factors such as ethnicity and age may also impact noninvasive tests. 20,21 With this background, we aimed to compare the accuracyofnoninvasivetestsinalargecohortofwhiteandAsian patients. We also compared Hispanic and non-Hispanic whites, as well as the potential confounding effects of BMI and age on the performance of noninvasive tests. Methods Study Design Overview This was a post hoc analysis of the screening data of the STELLAR-3 (NCT03053050) and STELLAR-4 (NCT03053063) trials evaluating the efficacy of selonsertib, an apoptosis signal-regulating kinase 1 inhibitor, in patients with NASH with bridging fibrosis (F3) and compensated cirrhosis (F4), respectively (Supplementary Methods). 22,23 The protocols conformed to ethical guidelines and were approved by the appropriate national and institutional review committees. All patients provided written informed consent. Clinical Assessments Investigators obtained history from and performed physical examination on the patients. Race was selfreported as white, Asian, black, American Indian, Pacific Islander, or other. BMI was calculated as body weight (kg) divided by height (m) squared. Blood was drawn after overnight fasting for liver biochemistry, metabolic parameters, and complete blood count. This study evaluated 4 noninvasive tests, namely the NAFLD fibrosis score, Fibrosis-4 index (FIB-4), Enhanced Liver Fibrosis panel (ELF), and liver stiffness measurement by VCTE (Supplementary Methods). Liver Histology Liver histology served as the reference standard in this study. All biopsy samples were centrally read and scored (by Z.G.) using the NASH Clinical Research Network system. 24 The primary endpoint was to identify advanced fibrosis, which included patients with bridging fibrosis (F3) and cirrhosis (F4). Statistical Analysis Continuous variables were expressed as median (interquartile range). The performance of noninvasive tests was evaluated using areas under the receiver What You Need to Know Background Clinicians need to use noninvasive tests to select patients with nonalcoholic fatty liver disease (NAFLD) for treatment and clinical trials. Most noninvasive tests were derived and validated in white patients. Findings Using data of 2281 white and 762 Asian patients being screened for the phase III STELLAR trials, we showed that the overall accuracies of the NAFLD fibrosis score, Fibrosis-4 index, Enhanced Liver Fibrosis test, and liver stiffness measurement by vibration—controlled transient elastography were similar between the 2 racial groups, as well as between Hispanic and non-Hispanic whites. The same cutoffs could also be used with similar sensitivities and specificities. Implications for patient care Clinicians can use these routinely available noninvasive tests in both white and Asian patients with NAFLD without adopting different cutoffs. This would simplify patient management and allow comparison of test results across studies. 2 Wong et al Clinical Gastroenterology and Hepatology Vol. -, No. - operating characteristics curves (AUROC) from univariate logistic regressions through 5-fold cross-validation (CV) repeated 100 times. Mean AUROC and 95% confidence interval (CI) are reported in the tables. 25 Specifically, we first calculate averaged AUROC over 5-fold CV (AUROC CV ) and variance of AUROC (s2 CV ) in each replicate, then mean AUROC was averaged AUROC CV and 95% CI was based on normal approximation with pooled variance over 100 replicates. AUROC comparisons were based on the c 2 distribution. We evaluated approaches using a single test with 1 cutoff and a single test with 2 cutoffs and calculated the corresponding sensitivities, specificities, and positive and negative predictive values. All test performance analyses were evaluated by racial subgroups. To address controversies about test confounders, we also performed subgroup analysis by ethnicity (Hispanic vs non-Hispanic), BMI (<30 and 30 kg/m 2 ) or age (<40, 40–64, and 65 years) within each racial subgroup. The Cochran-Mantel-Haenszel test stratified by diabetes status was used to evaluate the association between categorical variables. The Wilcoxon rank-sum test stratified by diabetes was used to compare baseline laboratory tests in Asian vs white and Hispanic vs non-Hispanic. Due to the exploratory nature of these analyses, there were no multiplicity adjustment for P- values. Results Analysis Population A total of 4024 patients underwent screening (2155 for STELLAR-3 and 1869 for STELLAR-4), of whom 3207 patients had evaluable histology (1001 [31%] were based on historical biopsies). This included 2281 whites, 762 Asians and 164 patients of other races. Because the vast majority were whites and Asians, we focused on these 2 racial groups. Among 2281 white patients, 539 (24%) were Hispanic. Supplementary Table 1 shows the breakdown of racial groups from each recruiting country or region. The at-risk PNPLA3 GG genotype was found in 57.6% of Hispanic whites, 25.6% of non-Hispanic whites, and 38.9% of Asians (Supplementary Figure 1). Compared with white patients, Asians were male predominant and fewer had hypertension, and had lower BMI, higher alanine aminotransferase, aspartate aminotransferase, and bilirubin levels but lower gammaglutamyl transferase level and platelet count (Table 1). The groups had similar age and hemoglobin A 1c . Overall, 76% of the entire cohort had NAFLD activity score 4, and 71% had advanced fibrosis (72% in whites and 67% in Asians; P¼.044). Noninvasive Tests in White and Asian Patients Overall and in each racial subgroup, the results of all 4 noninvasive tests increased with fibrosis stage (Table 2,Figure 1). At each fibrosis stage, the median NAFLD fibrosis score was lower in Asian than in white patients but only statistically significant at F4 (P¼.042); the median FIB-4 score was higher in Asian than in white patients at F2 (P¼.015), F3 (P<.001), and F4 (P< .001); the median ELF score was higher in Asian than in white patients at F3 (P¼.002) and F4 (P¼.023). Median liver stiffness was similar in the 2 groups. The AUROC of the 4 noninvasive tests for F3–F4 fibrosis in white patients were 0.73 to 0.80 and 0.75 to 0.83 for Asian patients (Table 3,Supplementary Table 2). Reliable liver stiffness measurements were obtained in 92% of whites and 81% of Asians. When dual published cutoffs were used (low cutoffs to rule out and high cutoffs to rule in F3–F4 fibrosis), the NAFLD fibrosis score, FIB-4, and ELF had similar sensitivities and specificities in white and Asian patients. Although liver stiffness had similarly high sensitivity in both white (84%) and Asian (83%) patients, the test was more specific in Asians (68% vs 78%). In addition, the positive predictive values of all 4 tests were over 90% in both racial groups. However, the negative predictive values of all 4 tests were lower in white patients than in Asian patients, with the biggest difference observed for liver stiffness (40% vs 58%). Based on dual cutoffs, 40% to 50% of white and Asian patients were in the gray zone (indeterminate results where tests are nondiagnostic) by the NAFLD fibrosis score, FIB-4, and ELF (Table 3). By contrast, only 8% to 10% of white and Asian patients were in the gray zone by liver stiffness measurement. Among whites, the AUROC of the 4 noninvasive tests for F3–F4 fibrosis were 0.73 to 0.87 for Hispanics and 0.72 to 0.77 for non-Hispanics (Table 3,Supplementary Table 2). Using dual cutoffs, the NAFLD fibrosis score, FIB-4, and ELF had similar sensitivities and specificities in both groups. Although liver stiffness measurement had similarly high sensitivity (84%) in both groups, the specificity was higher in Hispanics than non-Hispanics (75% vs 65%). Although the positive predictive values of all 4 tests were over 90% in both groups, the negative predictive values were particularly low in non-Hispanics, ranging from 33% to 58%. Noninvasive Tests in White and Asian Patients With Different BMI The test performance was similar in white and Asian patients with BMI <30 or 30 kg/m 2 with AUROCs ranging from 0.73 to 0.85 (Table 4,Supplementary Table 3). The sensitivities and specificities of the NAFLD fibrosis score, FIB-4, and ELF were similar in patients with different BMI in both groups. However, the specificity of liver stiffness measurement was lower in whites with BMI 30 kg/m 2 than those with BMI <30 kg/m 2 (61% vs 78%), but the same was not observed in Asians (86% vs 79%). As in the main analysis, the -2022 NITs of Fibrosis in Whites and Asians 3 Table 1. Clinical Characteristics of White and Asian Patients Hispanic (n ¼539) White All white (n ¼2281) Asian (n ¼762) White and Asian (N ¼3043) White vs Asian Pvalue Hispanic vs non-Hispanic Pvalue Non-Hispanic (n ¼1734) Age, y57 (50–63) 59 (52–64) 58 (52–64) 58 (49–64) 58 (51–64) .3043 .0111 Female sex 318 (59) 983 (57) 1308 (57) 359 (47) 1667 (55) <.0001 .2714 Body mass index, kg/m 2 32.82 (29.1–37.23) 34.18 (30.36–38.76) 33.94 (30.04–38.48) 28.21 (25.3–31.52) 32.41 (28.26–37.14) <.0001 2e-04 Diabetes 299 (55) 1070 (62) 1375 (60) 440 (59) 1815 (60) .4289 .0084 Hypertension 182 (34) 875 (50) 1060 (46) 268 (35) 1328 (44) <.0001 <.0001 ALT, U/L 38 (26–63.5) 43 (30–62.5) 42 (29–63) 50 (32–77) 43.5 (30–67) <.0001 .0299 AST, U/L 36 (24–57) 39.33 (29.5–55.67) 38.83 (28–56) 46 (32–65.5) 40 (29–58.5) <.0001 .0018 GGT, U/L 50 (32–92) 64 (38–118) 61 (36–112) 53 (34–87) 58 (35–106) <.0001 <.0001 Bilirubin, mg/dL 0.5 (0.4–0.7) 0.55 (0.4–0.8) 0.55 (0.4–0.8) 0.7 (0.5–0.9) 0.6 (0.4–0.8) <.0001 5e-04 Direct bilirubin, mg/dL 0.15 (0.1–0.2) 0.20 (0.1–0.2) 0.2 (0.1–0.2) 0.2 (0.1–0.2) 0.2 (0.1–0.2) 7e-04 <.0001 Platelets, 10 3 /mL201 (157–256.0) 191 (144–245.5) 194 (147–249) 181 (135–227) 190 (143–243) <.0001 .0051 Hemoglobin A 1c ,%6.2 (5.6–7.4) 6.2 (5.6–7.2) 6.2 (5.6–7.3) 6.2 (5.6–7.1) 6.2 (5.6–7.2) .8439 <.0001 Histology Biopsy length 2 cm 319 (59) 994 (57) 1316 (58) 281 (37) 1597 (52) <.0001 .4837 NAS 4 384 (71) 1353 (78) 1743 (76) 556 (73) 2299 (76) .0969 .0091 Steatosis grade 2 35 (6) 114 (7) 150 (7) 37 (5) 187 (6) .0946 1 Lobular inflammation grade 3 198 (37) 705 (41) 907 (40) 283 (37) 1190 (39) .2968 .1857 Hepatocellular ballooning grade 2 314 (58) 1080 (62) 1399 (61) 456 (60) 1855 (61) .7224 .2976 Fibrosis stage F0 85 (16) 86 (5) 171 (7) 52 (7) 223 (7) .0022 <.0001 F1 57 (11) 122 (7) 180 (8) 83 (11) 263 (9) .0022 <.0001 F2 85 (16) 194 (11) 281 (12) 113 (15) 394 (13) .0022 <.0001 F3 137 (25) 551 (32) 691 (30) 254 (33) 945 (31) .0022 <.0001 F4 175 (32) 781 (45) 958 (42) 260 (34) 1218 (40) .0022 <.0001 Note: Data are presented as median (interquartile range) or number (%). Note: The Cochran-Mantel-Haenszel test with diabetes as stratification factor was used to evaluate association between categorical variables. The Wilcoxon rank-sum test with diabetes as stratification factor was used compare baseline laboratory tests in Asian vs white and Hispanic vs non-Hispanic. Among the 3043 white and Asian patients with evaluable histology, there were 20 with missing diabetes status who were removed from the P- value calculation. ALT, Alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma-glutamyl transpeptidase; NAFLD, nonalcoholic fatty liver disease; NAS, NAFLD activity score. 4 Wong et al Clinical Gastroenterology and Hepatology Vol. -, No. - Table 2. Results of Noninvasive Tests in White and Asian Patients Hispanic (n ¼539) White All white (n ¼2281) Asian (n ¼762) White and Asian (N ¼3043) White vs Asian Pvalue Hispanic vs non-Hispanic Pvalue Non-Hispanic (n ¼1734) NAFLD fibrosis score 0.21 (1.23 to 0.69) 0.26 (0.72 to 1.15) 0.18 (0.86 to 1.09) 0.15 (1.04 to 0.76) 0.1 (0.92 to 0.98) <.0001 <.0001 F0 1.04 (2.28 to 0.4) 1.12 (1.52 to 0.29) 1.07 (1.86 to 0.29) 1.87 (2.62 to 0.93) 1.22 (2.11 to 0.45) .129 .1201 F1 1.05 (2.08 to 0.29) 0.71 (1.65 to 0.09) 0.75 (1.82 to 0.09) 0.99 (2.23 to 0.33) 0.88 (2.01 to 0.02) .0532 .2632 F2 0.73 (1.77 to 0.23) 0.47 (1.7 to 0.39) 0.61 (1.73 to 0.3) 0.98 (1.79 to 0.1) 0.73 (1.74 to 0.26) .5721 .357 F3 0.44 (1.18 to 0.34) 0.06 (1 to 0.77) 0.11 (1.04 to 0.69) 0.3 (1 to 0.43) 0.18 (1.01 to 0.6) .0521 .0337 F4 0.7 (0.13 to 1.69) 0.72 (0.16 to 1.61) 0.72 (0.15 to 1.63) 0.57 (0.3 to 1.3) 0.68 (0.18 to 1.53) .0417 .7883 FIB-4 1.58 (1.03–2.44) 1.84 (1.23–2.78) 1.78 (1.19–2.69) 2.11 (1.33–3.31) 1.85 (1.22–2.86) <.0001 <.0001 F0 1.03 (0.82–1.34) 1 (0.75–1.34) 1.02 (0.8–1.34) 0.97 (0.68–1.47) 1.01 (0.77–1.36) .8487 .3447 F1 1.05 (0.8–1.49) 1.18 (0.85–1.56) 1.12 (0.82–1.56) 1.25 (0.8–1.89) 1.18 (0.82–1.59) .2873 .2666 F2 1.24 (0.92–1.73) 1.31 (0.96–1.94) 1.29 (0.94–1.87) 1.46 (1.11–1.99) 1.37 (0.99–1.91) .0151 .0983 F3 1.66 (1.22–2.35) 1.65 (1.23–2.41) 1.65 (1.23–2.41) 2.26 (1.47–3.08) 1.76 (1.29–2.61) <.0001 .9003 F4 2.54 (1.82–3.83) 2.41 (1.73–3.61) 2.43 (1.75–3.66) 3.17 (2.07–4.42) 2.55 (1.8–3.84) <.0001 .1955 ELF 9.92 (9.07–10.8) 10.11 (9.36–10.88) 10.07 (9.3–10.86) 10.13 (9.36–10.89) 10.08 (9.32–10.87) .1306 .0037 F0 9.07 (8.39–9.51) 8.95 (8.32–9.45) 8.99 (8.37–9.5) 8.91 (8.35–9.34) 8.97 (8.37–9.48) .5805 .5725 F1 9.23 (8.77–9.62) 9.14 (8.68–9.79) 9.15 (8.7–9.73) 9.18 (8.76–9.83) 9.16 (8.72–9.75) .7284 .7307 F2 9.35 (8.7–9.85) 9.5 (8.94–10.15) 9.49 (8.86–10.11) 9.6 (8.94–10.18) 9.51 (8.89–10.13) .1422 .0588 F3 10.05 (9.62–10.63) 9.93 (9.33–10.53) 9.97 (9.36–10.56) 10.18 (9.57–10.84) 10.02 (9.42–10.63) .0022 .1227 F4 10.8 (10.17–11.41) 10.6 (10–11.3) 10.64 (10.04–11.34) 10.72 (10.21–11.68) 10.66 (10.07–11.39) .0234 .0361 Transient elastography Patients with available liver stiffness by transient elastography, n 238 1006 1247 431 1678 Liver stiffness by transient elastography, kPa 14.4 (8.9–20.7) 15.4 (10.5–23.9) 15.3 (10.4–23.2) 13.6 (9.4–20.8) 14.8 (10.1–22.3) 4e-04 .0346 Interquartile range- to-median ratio of liver stiffness measurement 0.13 (0.09–0.18) 0.15 (0.1–0.21) 0.14 (0.1–0.2) 0.13 (0.08–0.18) 0.14 (0.1–0.2) <.0001 .0033 XL probe 150 (63) 683 (68) 835 (67) 109 (25) 944 (56) <.0001 .3546 -2022 NITs of Fibrosis in Whites and Asians 5 negative predictive value of liver stiffness measurement was higher in Asians, particularly in those with BMI 30 kg/m 2 (51%, compared with 37% in Asians with BMI <30 kg/m 2 and 26% in whites). Using dual cutoffs, the NAFLD fibrosis score included more patients with BMI <30 kg/m 2 in the gray zone (48%–58% among racial and BMI groups) (Table 4). Only 6% to 12% of patients were in the gray zone by liver stiffness measurement. Noninvasive Tests in White and Asian Patients by Age Sensitivities were low in white patients <40 years for the NAFLD fibrosis score (53%), FIB-4 (34%), and ELF (40%) (Table 5,Supplementary Table 4). In Asian patients <40 years, the sensitivities were 57%, 42%, and 50%, respectively. The sensitivities of these tests were 71% to 98% in patients aged 40 to 64 and 65 years. The sensitivity of VCTE was 80% in all age groups in both racial groups, with the highest sensitivity of 94% in Asian patients <40 years. By contrast, the NAFLD fibrosis score, FIB-4, and ELF had high specificities of 91% to 100% in both white and Asian patients <65 years (Table 5). However, the specificity of the NAFLD fibrosis score was low at 69% in white patients aged 65 years, although the specificities remained high for FIB-4 (88%) and ELF (96%). Although the specificities were 81% for the NAFLD fibrosis score and 94% for ELF in Asian patients aged 65 years, the specificity of FIB-4 dropped to 56%. McPherson and colleagues proposed higher cutoffs for the NAFLD fibrosis score (0.12) and FIB-4 (2.0) for patients aged 65 years. 21 Using these cutoffs, the sensitivity and specificity of the NAFLD fibrosis score were 71% and 55% in white patients aged 65 years, and 67% and 56% in Asian patients aged 65 years. The sensitivity and specificity of FIB-4 were 70% and 73% in white patients aged 65 years, and 83% and 38% in Asian patients aged 65 years. Two-step Approach Supplementary Table 5 shows the 2-step approach of different noninvasive tests, with the second test performed when the first (FIB-4 or NAFLD fibrosis score) yielded indiscriminate results. In general, this approach allowed a drastic reduction in the proportion of patients classified in the gray zone with preserved positive predictive values and modest reduction in negative predictive values in all racial groups evaluated. Discussion In this large international study using high-quality clinical trial data, routinely available noninvasive tests including NAFLD fibrosis score, FIB-4, ELF, and liver Table 2.Continued Hispanic (n ¼539) White All white (n ¼2281) Asian (n ¼762) White and Asian (N ¼3043) White vs Asian Pvalue Hispanic vs non-Hispanic Pvalue Non-Hispanic (n ¼1734) Reliable liver stiffness measurement a 223 (94) 918 (91) 1143 (92) 351 (81) 1494 (89) .232 .3116 F0 5.9 (4.55–7.2) 7.85 (6.2–11.3) 6.7 (4.9–9.2) 6.5 (4.8–8.8) 6.6 (4.9–9.2) .8148 .0572 F1 6.8 (5.6–9.4) 9.8 (7.3–11.6) 9.1 (6.8–11.4) 9.05 (7.4–10.4) 9.05 (6.85–10.65) .486 .0327 F2 8.94 (6.05–12.9) 10.25 (7.2–14.15) 9.8 (6.85–13.75) 8.95 (6.9–12) 9.5 (6.9–13.4) .1767 .2243 F3 13.3 (9.9–18) 12.8 (9.7–17.6) 12.9 (9.7–17.7) 12.45 (9.6–17.1) 12.8 (9.7–17.3) .4404 .3277 F4 20.7 (15.3–31.6) 21 (14.1–29.7) 21.05 (14.3–29.9) 21.3 (15.9–29.9) 21.25 (14.4–29.9) .4629 .6375 Note: Data are presented as median (interquartile range) or number (%). Note: Percentage of patients with XL probe and reliable liver stiffness measurement are based on patients with non-missing values of liver stiffness by transient elastography (kPa) in each sub-population. ELF, Enhanced liver fibrosis panel; FIB-4, Fibrosis-4 index; NAFLD, nonalcoholic fatty liver disease. a Reliable liver stiffness measurement was defined as 10 valid acquisitions and an interquartile range-to-median ratio of 0.3 or less among 1678 patients who underwent transient elastography. 6 Wong et al Clinical Gastroenterology and Hepatology Vol. -, No. - stiffness measurement had similar performance in white and Asian patients with NAFLD. Importantly, the sensitivities and specificities of the tests were reasonable when the original published cutoffs were applied to Asian patients, suggesting that currently available noninvasive tests can be used in Asians without further calibration or adjustment of cutoffs. In addition, the noninvasive tests also performed similarly in Hispanic and non-Hispanic whites. Among the noninvasive tests evaluated in this study, VCTE is the most extensively evaluated in both white and Asian patients. 26–29 Previous studies suggested that liver stiffness measurement might lead to false-positive diagnosis of advanced fibrosis in patients with extreme BMI, likely due to technical difficulties. 17–19 Obesity also increases the risk of failure to obtain liver stiffness measurements. 30 Nonetheless, the availability of the XL probe allows reliable liver stiffness measurement in the majority of obese patients. 31 In the current study, VCTE had similarly high accuracy in whites and Asians with BMI below and above 30 kg/m 2 . Compared with other noninvasive tests, VCTE classified the fewest patients in the gray zone. However, the lower specificity of VCTE in whites than in Asians, especially among whites with BMI above 30 kg/m 2 , suggests that the confounding effect of obesity on liver stiffness measurement was still at play. 32 Among the blood tests, ELF had the highest AUROC in white and Asian patients, although its superiority over FIB-4 was marginal. Although BMI is a component of the NAFLD fibrosis score, the diagnostic accuracy of this score was similar in the racial groups (Table 3). Likewise, the score performed similarly in patients with BMI below and above 30 kg/m 2 . This is probably due to the relative low weighting assigned to BMI in the NAFLD fibrosis score. 14 Recently, a cross-sectional study showed that the NAFLD fibrosis score and FIB-4 might perform even better in patients with BMI below 25 kg/m 2 . 33 Because the noninvasive tests are imperfect, investigators usually propose dual cutoffs with a low cutoff to rule out significant fibrosis and a high cutoff to rule in advanced fibrosis. 9 Although it is not difficult to determine cutoffs with high sensitivities and specificities, the important issue is the proportion of patients left in the gray zone with indeterminate results. Based on published cutoffs, around 40% to 50% of white and Asian patients in this study had indeterminate results with the blood biomarkers (Table 3). Although VCTE classified much fewer patients in the gray zone (8%–10%), the tradeoff was a slightly lower overall accuracy. McPherson and colleagues first reported lower diagnostic accuracy of the aspartate aminotransferase/ alanine aminotransferase ratio, NAFLD fibrosis score, Figure 1. Distribution of noninvasive tests in white and Asian patients. NFS, Nonalcoholic fatty liver disease fibrosis score. -2022 NITs of Fibrosis in Whites and Asians 7 Table 3. Performance of Noninvasive Tests in Diagnosing Advanced Fibrosis (F3–F4) in White and Asian Patients (2-cutoff Model) Percentage of patients with F3–F4 AUROC (95% CI) Cutoff % (95% CI) Sensitivity Specificity PPV NPV Gray zone a Accuracy White Hispanic patients NAFLD fibrosis score (n ¼372) 66 0.73 (0.68–0.78) <1.455 to 0.676 90 (86–94) 90 (84–95) 95 (91–97) 83 (75–89) 55 (50–60) 90 (87–93) FIB-4 (n ¼528) 58 0.81 (0.78–0.85) <1.3 to 2.67 81 (76–85) 97 (94–99) 97 (94–99) 78 (73–83) 42 (37–46) 88 (84–90) ELF (n ¼539) 58 0.83 (0.79–0.86) <9.8 to 11.3 78 (73–83) 99 (97–100) 99 (97–100) 77 (72–82) 41 (37–45) 87 (84–90) Liver stiffness measurement (n ¼238) 76 0.87 (0.82–0.92) <9.9 to 11.4 kPa 84 (78–89) 75 (62–86) 92 (86–95) 59 (47–71) 5 (2–8) 82 (76–87) White non-Hispanic patients NAFLD fibrosis score (n ¼1348) 86 0.72 (0.68–0.76) <1.455 to 0.676 89 (87–91) 85 (79–90) 97 (96–98) 56 (50–62) 49 (46–52) 88 (87–90) FIB-4 (n ¼1697) 77 0.76 (0.73–0.78) <1.3 to 2.67 81 (78–83) 91 (88–94) 97 (96–98) 58 (54–62) 45 (43–48) 83 (81–85) ELF (n ¼1720) 77 0.77 (0.75–0.8) <9.8 to 11.3 71 (69–74) 97 (95–99) 99 (98–99) 50 (47–54) 47 (45–50) 77 (75–79) Liver stiffness measurement (n ¼1006) 89 0.76 (0.72–0.81) <9.9 to 11.4 kPa 84 (82–87) 65 (55–74) 95 (93–97) 33 (27–40) 8 (7–10) 82 (80–84) All white patients NAFLD fibrosis score (n ¼1725) 82 0.73 (0.7–0.76) <1.455 to 0.676 89 (87–91) 87 (83–91) 97 (96–98) 65 (60–69) 50 (48–53) 89 (87–90) FIB-4 (n ¼2233) 73 0.78 (0.76–0.8) <1.3 to 2.67 81 (79–83) 93 (91–95) 97 (96–98) 64 (61–68) 44 (42–47) 84 (83–86) ELF (n ¼2267) 72 0.79 (0.77–0.81) <9.8 to 11.3 73 (70–75) 98 (96–99) 99 (98–99) 58 (55–61) 46 (44–48) 80 (78–81) Liver stiffness measurement (n ¼1247) 87 0.8 (0.76–0.84) <9.9 to 11.4 kPa 84 (82–86) 68 (61–75) 95 (93–96) 40 (34–46) 8 (6–9) 82 (80–84) Asian patients NAFLD fibrosis score (n ¼586) 78 0.75 (0.7–0.8) <1.455 to 0.676 88 (85–91) 92 (86–96) 98 (96–99) 68 (61–75) 54 (50–58) 89 (86–91) FIB-4 (n ¼735) 69 0.8 (0.76–0.83) <1.3 to 2.67 88 (85–90) 90 (85–93) 95 (92–97) 77 (71–82) 40 (37–44) 88 (86–91) ELF (n ¼748) 68 0.81 (0.77–0.84) <9.8 to 11.3 76 (72–80) 96 (93–98) 98 (96–99) 66 (60–70) 44 (40–47) 83 (80–85) Liver stiffness measurement (n ¼431) 76 0.83 (0.78–0.87) <9.9 to 11.4 kPa 83 (78–87) 78 (68–85) 92 (88–95) 58 (50–67) 10 (7–13) 81 (77–85) AUROC, Area under the receiver operating characteristics curve; CI, confidence interval; ELF, enhanced liver fibrosis panel; FIB-4, Fibrosis-4 index; NAFLD, nonalcoholic fatty liver disease; NPV, negative predictive value; PPV, positive predictive value. a “Gray zone”introduced by utilizing dual cutoffs is defined as indeterminate results where the tests are nondiagnostic. 8 Wong et al Clinical Gastroenterology and Hepatology Vol. -, No. - Table 4. Performance of Noninvasive Tests in Diagnosing Advanced Fibrosis (F3-F4) in White and Asian Patients by BMI (2-cutoff Model) Percentage of patients with F3–F4 AUROC (95% CI) Cutoff % (95% CI) Sensitivity Specificity PPV NPV Gray zone a Accuracy White patients, BMI <30 kg/m 2 NAFLD fibrosis score (n ¼428) 81 0.76 (0.71–0.81) <1.455 to 0.676 80 (76–84) 99 (93–100) 100 (98–100) 55 (46–63) 58 (53–62) 84 (80–87) FIB-4 (n ¼429) 81 0.75 (0.69–0.8) <1.3 to 2.67 82 (78–86) 87 (78–93) 96 (93–98) 54 (45–63) 44 (39–49) 83 (79–87) ELF (n ¼436) 80 0.80 (0.75–0.85) <9.8 to 11.3 70 (65–75) 100 (96–100) 100 (98–100) 45 (38–53) 48 (43–53) 76 (72–80) Liver stiffness measurement (n ¼263) 91 0.79 (0.68–0.89) <9.9 to 11.4 kPa 78 (73–83) 78 (56–93) 97 (94–99) 26 (16–38) 12 (8–16) 78 (73–83) White patients, BMI 30 kg/m 2 NAFLD fibrosis score (n ¼1297) 82 0.73 (0.69–0.76) <1.455 to 0.676 92 (90–94) 83 (78–88) 96 (95–97) 70 (64–75) 48 (45–50) 91 (89–92) FIB-4 (n ¼1300) 82 0.78 (0.75–0.81) <1.3 to 2.67 80 (77–82) 94 (90–97) 98 (97–99) 50 (45–55) 46 (44–49) 82 (80–84) ELF (n ¼1310) 82 0.8 (0.76–0.83) <9.8 to 11.3 73 (70–76) 97 (95–99) 99 (98–100) 45 (40–49) 49 (46–52) 78 (75–80) Liver stiffness measurement (n ¼846) 93 0.77 (0.71–0.83) <9.9 to 11.4 kPa 86 (83–88) 61 (48–73) 97 (95–98) 26 (19–33) 6 (5–8) 84 (81–86) Asian patients, BMI <30 kg/m 2 NAFLD fibrosis score (n ¼376) 78 0.75 (0.69–0.81) <1.455 to 0.676 87 (83–91) 93 (85–97) 98 (95–99) 68 (58–76) 57 (51–62) 89 (85–92) FIB-4 (n ¼377) 78 0.75 (0.69–0.81) <1.3 to 2.67 93 (89–95) 77 (67–86) 93 (90–96) 74 (64–83) 37 (32–42) 89 (86–92) ELF (n ¼381) 77 0.78 (0.72–0.83) <9.8 to 11.3 78 (73–83) 97 (90–99) 99 (96–100) 56 (48–64) 49 (44–54) 82 (78–86) Liver stiffness measurement (n ¼220) 87 0.8 (0.72–0.88) <9.9 to 11.4 kPa 80 (74–86) 79 (59–92) 96 (92–99) 37 (25–50) 10 (7–15) 80 (74–85) Asian patients, BMI 30 kg/m 2 NAFLD fibrosis score (n ¼210) 79 0.75 (0.68–0.83) <1.455 to 0.676 90 (84–94) 91 (78–97) 97 (93–99) 70 (57–82) 50 (43–56) 90 (85–94) FIB-4 (n ¼210) 79 0.75 (0.68–0.83) <1.3 to 2.67 78 (71–84) 93 (81–99) 98 (94–100) 53 (42–65) 45 (38–52) 81 (75–86) ELF (n ¼211) 79 0.78 (0.71–0.86) <9.8 to 11.3 71 (64–78) 93 (81–99) 98 (93–99) 46 (35–57) 44 (37–51) 76 (69–81) Liver stiffness measurement (n ¼137) 85 0.85 (0.76–0.94) <9.9 to 11.4 kPa 85 (78–91) 86 (64–97) 97 (92–99) 51 (34–69) 10 (6–17) 85 (78–91) AUROC, Area under the receiver operating characteristics curve; BMI, body mass index; CI, confidence interval; ELF, enhanced liver fibrosis panel; FIB-4, Fibrosis-4 index; NAFLD, nonalcoholic fatty liver disease; NPV, negative predictive value; PPV, positive predictive value. a “Gray zone”introduced by utilizing dual cutoffs is defined as indeterminate results where the tests are non-diagnostic. -2022 NITs of Fibrosis in Whites and Asians 9 Supplementary Table 2. Performance of Noninvasive Tests in Diagnosing Advanced Fibrosis (F3–F4) in White and Asian Patients (1-cutoff Model) Percentage of patients with F3–F4 AUROC (95% CI) Cutoff % (95% CI) Sensitivity Specificity PPV NPV White Hispanic patients NAFLD fibrosis score (n ¼372) 66 0.73 (0.68–0.78) 1.455 90 (86–94) 40 (31–49) 74 (69–79) 68 (56–78) 0.676 33 (27–40) 90 (84–95) 87 (79–93) 41 (35–47) FIB-4 (n ¼528) 58 0.81 (0.78–0.85) 1.3 81 (76–85) 62 (56–69) 75 (70–80) 70 (63–76) 2.67 34 (29–39) 97 (94–99) 94 (87–97) 51 (46–56) ELF (n ¼539) 58 0.83 (0.79–0.86) 9.8 78 (73–83) 78 (72–84) 83 (79–87) 72 (66–78) 11.3 22 (18–27) 99 (97–100) 97 (90–100) 48 (44–53) Liver stiffness measurement (n ¼238) 76 0.87 (0.82–0.92) 9.9 kPa 84 (78–89) 71 (58–83) 91 (85–94) 58 (45–70) 11.4 kPa 79 (72–85) 75 (62–86) 91 (86–95) 52 (41–64) White non-Hispanic patients NAFLD fibrosis score (n ¼1348) 86 0.72 (0.68–0.76) 1.455 89 (87–91) 30 (24–37) 89 (87–90) 31 (24–38) 0.676 41 (38–44) 85 (79–90) 94 (92–96) 19 (17–22) FIB-4 (n ¼1697) 77 0.76 (0.73–0.78) 1.3 81 (78–83) 57 (52–62) 87 (84–88) 46 (42–51) 2.67 32 (30–35) 91 (88–94) 93 (90–95) 28 (26–31) ELF (n ¼1720) 77 0.77 (0.75–0.8) 9.8 71 (69–74) 71 (66–75) 89 (87–91) 42 (39–46) 11.3 18 (16–20) 97 (95–99) 96 (92–98) 26 (24–28) Liver stiffness measurement (n ¼1006) 89 0.76 (0.72–0.81) 9.9 kPa 84 (82–87) 52 (42–62) 94 (92–95) 28 (22–35) 11.4 kPa 76 (73–79) 65 (55–74) 95 (93–96) 25 (20–30) All white patients NAFLD fibrosis score (n ¼1725) 82 0.73 (0.7–0.76) 1.455 89 (87–91) 34 (29–39) 86 (84–87) 41 (35–48) 0.676 40 (37–42) 87 (83–91) 93 (91–95) 25 (22–27) FIB-4 (n ¼2233) 73 0.78 (0.76–0.8) 1.3 81 (79–83) 59 (55–63) 84 (82–86) 53 (49–57) 2.67 33 (30–35) 93 (91–95) 93 (91–95) 34 (32–36) ELF (n ¼2267) 72 0.79 (0.77–0.81) 9.8 73 (70–75) 73 (70–77) 88 (86–89) 51 (47–54) 11.3 19 (17–21) 98 (96–99) 96 (93–98) 32 (29–34) Liver stiffness measurement (n ¼1247) 87 0.8 (0.76–0.84) 9.9 kPa 84 (82–86) 59 (51–66) 93 (91–95) 36 (30–42) 11.4 kPa 77 (74, 79) 68 (61, 75) 94 (92, 96) 31 (26, 36) Asian patients NAFLD fibrosis score (n ¼586) 78 0.75 (0.7–0.8) 1.455 88 (85–91) 43 (34–52) 85 (81–88) 50 (40–60) 0.676 33 (29–37) 92 (86–96) 94 (89–97) 28 (23–32) FIB-4 (n ¼735) 69 0.8 (0.76–0.83) 1.3 88 (85–90) 48 (41–55) 79 (75–82) 64 (56–71) 2.67 48 (43–52) 90 (85–93) 91 (87–94) 44 (39–48) ELF (n ¼748) 68 0.81 (0.77–0.84) 9.8 76 (72–80) 71 (65–77) 85 (81–88) 58 (53–64) 11.3 24 (20–28) 96 (93–98) 93 (87–97) 37 (33–41) Liver stiffness measurement (n ¼431) 76 0.83 (0.78–0.87) 9.9 kPa 83 (78–87) 65 (55–74) 88 (84–92) 54 (45–63) 11.4 kPa 73 (68–78) 78 (68–85) 91 (87–94) 48 (40–56) AUROC, Area under the receiver operating characteristics curve; CI, confidence interval; ELF, Enhanced Liver Fibrosis panel; FIB-4, Fibrosis-4 index; NAFLD, nonalcoholic fatty liver disease; NPV, negative predictive value; PPV, positive predictive value. 13.e3 Wong et al Clinical Gastroenterology and Hepatology Vol. -, No. - Supplementary Table 3. Performance of Noninvasive Tests in Diagnosing Advanced Fibrosis (F3–F4) in White and Asian Patients With Different BMI (1-cutoff Model) Percentage of patients with F3-F4 AUROC (95% CI) Cutoff % (95% CI) Sensitivity Specificity PPV NPV White patients, BMI <30 kg/m 2 NAFLD fibrosis score (n ¼428) 81 0.76 (0.71–0.81) 1.455 80 (76–84) 52 (41–63) 87 (83–91) 39 (30–48) 0.676 20 (16–25) 99 (93–100) 99 (92–100) 23 (19–28) FIB-4 (n ¼429) 81 0.75 (0.69–0.8) 1.3 82 (78–86) 52 (41–63) 88 (84–91) 41 (32–51) 2.67 36 (31–41) 87 (78–93) 92 (86–96) 25 (20–30) ELF (n ¼436) 80 0.8 (0.75–0.85) 9.8 70 (65–75) 77 (67–85) 92 (89–95) 39 (32–47) 11.3 16 (12–20) 100 (96–100) 100 (94–100) 23 (19–27) Liver stiffness measurement (n ¼263) 91 0.79 (0.68–0.89) 9.9 kPa 78 (73–83) 74 (52–90) 97 (93–99) 25 (15–36) 11.4 kPa 66 (59–72) 78 (56–93) 97 (93–99) 18 (11–27) White patients, BMI 30 kg/m 2 NAFLD fibrosis score (n ¼1297) 82 0.73 (0.69–0.76) 1.455 92 (90–94) 27 (22–34) 85 (83–87) 43 (35–52) 0.676 46 (43–49) 83 (78–88) 93 (90–95) 25 (22–29) FIB-4 (n ¼1300) 82 0.78 (0.75–0.81) 1.3 80 (77–82) 61 (54–67) 90 (88–92) 39 (34–45) 2.67 30 (27–33) 94 (90–97) 96 (93–98) 23 (20–26) ELF (n ¼1310) 82 0.8 (0.76–0.83) 9.8 73 (70–76) 75 (69–80) 93 (91–95) 38 (34–43) 11.3 19 (16–21) 97 (95–99) 97 (94–99) 21 (19–24) Liver stiffness measurement (n ¼846) 93 0.77 (0.71–0.83) 9.9 kPa 86 (83–88) 50 (37–63) 96 (94–97) 22 (15–30) 11.4 kPa 80 (77–83) 61 (48–73) 96 (95–98) 19 (14–26) Asian patients, BMI <30 kg/m 2 NAFLD fibrosis score (n ¼376) 78 0.75 (0.69–0.81) 1.455 87 (83–91) 43 (33–55) 84 (80–88) 49 (37–61) 0.676 29 (24–34) 93 (85–97) 93 (86–98) 27 (22–32) FIB-4 (n ¼377) 78 0.75 (0.69–0.81) 1.3 93 (89–95) 39 (28–50) 84 (80–88) 59 (45–72) 2.67 56 (50–62) 77 (67–86) 90 (84–94) 33 (27–40) ELF (n ¼381) 77 0.78 (0.72–0.83) 9.8 78 (73–83) 67 (56–77) 89 (85–93) 47 (38–56) 11.3 23 (19–29) 97 (90–99) 96 (88–99) 27 (22–32) Liver stiffness measurement (n ¼220) 87 0.8 (0.72–0.88) 9.9 kPa 80 (74–86) 64 (44–81) 94 (89–97) 32 (20–46) 11.4 kPa 70 (63–77) 79 (59–92) 96 (91–98) 28 (18–39) Asian patients, BMI 30 kg/m 2 NAFLD fibrosis score (n ¼210) 79 0.75 (0.68–0.83) 1.455 90 (84–94) 41 (26–57) 85 (79–90) 51 (34–69) 0.676 40 (33–48) 91 (78–97) 94 (86–98) 29 (21–37) FIB-4 (n ¼210) 79 0.75 (0.68–0.83) 1.3 78 (71–84) 50 (35–65) 86 (79–91) 38 (26–52) 2.67 33 (26–41) 93 (81–99) 95 (86–99) 27 (20–35) ELF (n ¼211) 79 0.78 (0.71–0.86) 9.8 71 (64–78) 70 (55–83) 90 (84–95) 39 (28–51) 11.3 22 (16–29) 93 (81–99) 92 (79–98) 24 (18–31) Liver stiffness measurement (n ¼137) 85 0.85 (0.76–0.94) 9.9 kPa 85 (78–91) 71 (48–89) 94 (88–98) 47 (29–65) 11.4 kPa 76 (67–83) 86 (64–97) 97 (91–99) 39 (25–55) AUROC, Area under the receiver operating characteristics curve; BMI, body mass index; CI, confidence interval; ELF, Enhanced Liver Fibrosis panel; FIB-4, Fibrosis-4 index; NAFLD, nonalcoholic fatty liver disease; NPV, negative predictive value; PPV, positive predictive value. -2022 NITs of Fibrosis in Whites and Asians 13.e4 Supplementary Table 4. Performance of Noninvasive Tests in Diagnosing Advanced Fibrosis (F3–F4) in White and Asian Patients by Age (1-cutoff Model) Percentage of patients with F3–F4 AUROC (95% CI) Cutoff % (95% CI) Sensitivity Specificity PPV NPV White patients, age <40 years NAFLD fibrosis score (n ¼68) 69 0.7 (0.57–0.82) 1.455 53 (38–68) 76 (53–92) 83 (65–94) 42 (26–59) 0.676 13 (5–26) 95 (76–100) 86 (42–100) 33 (21–46) FIB-4 (n ¼105) 53 0.71 (0.61–0.81) 1.3 34 (22–48) 94 (83–99) 86 (65–97) 55 (44–66) 2.67 5 (1–15) 100 (93–100) 100 (29–100) 48 (38–58) ELF (n ¼111) 51 0.75 (0.66–0.84) 9.8 40 (28–54) 91 (80–97) 82 (63–94) 59 (48–70) 11.3 4 (0–12) 100 (93–100) 100 (16–100) 50 (40–59) Liver stiffness measurement (n ¼51) 78 0.71 (0.55–0.86) 9.9 kPa 80 (64–91) 55 (23–83) 86 (71–95) 43 (18–71) 11.4 kPa 75 (59–87) 64 (31–89) 88 (73–97) 41 (18–67) White patients, age 40–64 years NAFLD fibrosis score (n ¼1261) 81 0.74 (0.7–0.77) 1.455 89 (86–90) 36 (30–43) 85 (83–88) 43 (36–50) 0.676 37 (34–40) 91 (87–94) 94 (92–97) 25 (23–28) FIB-4 (n ¼1637) 72 0.77 (0.75–0.8) 1.3 79 (77–81) 60 (56–65) 84 (81–86) 53 (49–57) 2.67 30 (27–32) 94 (91–96) 93 (89–95) 34 (32–37) ELF (n ¼1655) 72 0.79 (0.76–0.81) 9.8 71 (69–74) 74 (70–78) 87 (85–89) 51 (47–54) 11.3 18 (16–20) 98 (96–99) 96 (92–98) 32 (30–35) Liver stiffness measurement (n ¼923) 86 0.82 (0.78–0.86) 9.9 kPa 85 (82–87) 61 (52–70) 93 (91–95) 40 (34–48) 11.4 kPa 77 (74–80) 70 (62–78) 94 (92–96) 34 (28–40) White patients, age 65 years NAFLD fibrosis score (n ¼396) 86 0.68 (0.6–0.75) 1.455 96 (94–98) 7 (2–18) 87 (83–90) 24 (7–50) 0.676 52 (46–57) 69 (55–81) 91 (86–95) 19 (14–25) 0.12 71 (66–75) 55 (41–68) 91 (86–94) 23 (16–31) FIB-4 (n ¼491) 80 0.78 (0.73–0.83) 1.3 92 (89–94) 35 (26–45) 85 (81–88) 52 (39–64) 2.67 45 (40–50) 88 (79–93) 94 (89–97) 28 (23–33) 2 70 (66–75) 73 (63–82) 91 (88–94) 38 (31–45) ELF (n ¼501) 80 0.77 (0.72–0.82) 9.8 81 (77–85) 62 (52–71) 89 (86–92) 46 (37–55) 11.3 24 (19–28) 96 (90–99) 96 (90–99) 24 (20–29) Liver stiffness measurement (n ¼273) 92 0.72 (0.62–0.83) 9.9 kPa 83 (77–87) 43 (22–66) 95 (91–97) 17 (8–30) 11.4 kPa 76 (70–81) 57 (34–78) 96 (92–98) 17 (9–27) Asian patients, age <40 years NAFLD fibrosis score (n ¼39) 59 0.83 (0.69–0.97) 1.455 57 (34–77) 88 (62–98) 87 (60–98) 58 (37–78) 0.676 13 (3–34) 94 (70–100) 75 (19–99) 43 (26–61) FIB-4 (n ¼64) 38 0.76 (0.64–0.88) 1.3 42 (22–63) 88 (73–96) 67 (38–88) 71 (57–83) 2.67 12 (3–32) 100 (91–100) 100 (29–100) 66 (52–77) ELF (n ¼66) 36 0.76 (0.64–0.88) 9.8 50 (29–71) 90 (77–97) 75 (48–93) 76 (62–87) 11.3 4 (0–21) 100 (92–100) 100 (3–100) 65 (52–76) Liver stiffness measurement (n ¼35) 51 0.82 (0.69–0.96) 9.9 kPa 94 (73–100) 47 (23–72) 65 (44–83) 89 (52–100) 11.4 kPa 78 (52–94) 65 (38–86) 70 (46–88) 73 (45–92) Asian patients, age 40–64 years NAFLD fibrosis score (n ¼392) 79 0.76 (0.71–0.82) 1.455 87 (82–90) 42 (31–53) 84 (80–88) 46 (35–58) 0.676 28 (23–33) 95 (88–99) 96 (89–99) 26 (22–32) FIB-4 (n ¼489) 68 0.8 (0.76–0.84) 1.3 87 (83–91) 46 (38–55) 78 (73–82) 63 (53–71) 2.67 43 (37–48) 94 (89–97) 94 (89–97) 43 (38–49) ELF (n ¼500) 67 0.8 (0.76–0.84) 9.8 74 (69–78) 69 (62–76) 83 (78–87) 56 (49–63) 11.3 23 (19–28) 96 (91–98) 92 (84–97) 38 (33–42) 13.e5 Wong et al Clinical Gastroenterology and Hepatology Vol. -, No. - Supplementary Table 4.Continued Percentage of patients with F3–F4 AUROC (95% CI) Cutoff % (95% CI) Sensitivity Specificity PPV NPV Liver stiffness measurement (n ¼291) 75 0.85 (0.79–0.9) 9.9 kPa 82 (76–87) 69 (57–80) 89 (84–93) 56 (45–66) 11.4 kPa 73 (66–78) 82 (71–90) 92 (87–96) 50 (40–59) Asian patients, age 65 years NAFLD fibrosis score (n ¼155) 83 0.68 (0.57–0.79) 1.455 98 (93–100) 19 (6–38) 85 (78–90) 62 (24–91) 0.676 48 (40–57) 81 (62–94) 93 (83–98) 25 (16–35) 0.12 67 (58–75) 56 (35–75) 88 (80–94) 26 (16–40) FIB-4 (n ¼182) 81 0.64 (0.54–0.74) 1.3 97 (92–99) 9 (2–24) 82 (76–88) 38 (9–76) 2.67 65 (57–73) 56 (38–73) 86 (79–92) 27 (17–39) 2 83 (76–89) 38 (22–56) 85 (79–91) 34 (20–51) ELF (n ¼182) 81 0.76 (0.67–0.84) 9.8 86 (80–92) 56 (38–73) 90 (83–94) 49 (32–65) 11.3 28 (21–36) 94 (80–99) 95 (85–99) 23 (16–31) Liver stiffness measurement (n ¼105) 87 0.78 (0.68–0.89) 9.9 kPa 82 (73–90) 64 (35–87) 94 (86–98) 36 (18–57) 11.4 kPa 75 (65–83) 71 (42–92) 94 (86–98) 30 (16–49) AUROC, Area under the receiver operating characteristics curve; CI, confidence interval; ELF, Enhanced Liver Fibrosis panel; FIB-4, Fibrosis-4 index; NAFLD, nonalcoholic fatty liver disease; NPV, negative predictive value; PPV, positive predictive value. Supplementary Table 5. Two-step Approach for the Diagnosis of Advanced Fibrosis (F3–4) in White and Asian Patients Test sequence % (95% CI) Sensitivity Specificity PPV NPV Gray zone Accuracy White patients FIB-4 /ELF (n ¼2273) 67 (64–69) 93 (91–95) 96 (95–97) 52 (49–55) 25 (24–27) 74 (72–76) FIB-4 /LSM (n ¼2245) 75 (73–77) 90 (87–92) 95 (94–96) 57 (54–60) 21 (19–22) 79 (77–81) NFS /ELF (n ¼2271) 72 (70–74) 92 (90–94) 96 (95–97) 56 (53–59) 28 (26–30) 78 (76–79) NFS /LSM (n ¼1872) 82 (80–84) 80 (76–84) 94 (92–95) 55 (51–59) 19 (17–21) 82 (80–84) Asian patients FIB-4 /ELF (n ¼748) 76 (72–79) 89 (84–92) 93 (91–96) 63 (58–68) 22 (19–25) 80 (77–83) FIB-4 /LSM (n ¼740) 83 (80–87) 86 (81–90) 93 (90–95) 70 (65–76) 19 (17–22) 84 (81–87) NFS /ELF (n ¼748) 74 (70–78) 94 (90–96) 96 (94–98) 63 (58–68) 31 (28–35) 80 (77–83) NFS /LSM (n ¼662) 81 (77–84) 84 (78–89) 93 (90–95) 62 (56–68) 21 (18–24) 82 (78–84) Note: In the first step, patients with results below the lower cutoff and above the higher cutoff were considered to have advanced fibrosis excluded and confirmed, respectively. Those with results between the 2 cutoffs (ie, gray zone) would undergo the second test. The cutoffs of the noninvasive tests are as follows: FIB-4, 1.3–2.67; NFS, 1.455 to 0.676; ELF, 9.8–11.3; LSM, 9.9–11.4. ELF, Enhanced Liver Fibrosis panel; FIB-4, Fibrosis-4 index; LSM, liver stiffness measurement; NFS, nonalcoholic fatty liver disease fibrosis score; NPV, negative predictive value; PPV, positive predictive value. -2022 NITs of Fibrosis in Whites and Asians 13.e6