Validation of the new 2021 EASL algorithm for the non-invasive diagnosis of advanced liver fibrosis in non-alcoholic fatty liver disease
Clémence M Canivet, Charlotte Costentin, Katharina M Irvine, Adèle Delamarre, Adrien Lannes, Nathalie Sturm, Frederic Oberti, Preya J Patel, Thomas Decaens, Marie Irles-Depé, Isabelle Fouchard, Paul Hermabessière, Marine Roux, Justine Barthelon, Paul Calès, Elizabeth E Powell, Victor de Ledinghen, Jérôme Boursier
Table of contents
Supplementary Figure 1: Methods for the estimation of EASL algorithm accuracy at different prevalences of advanced fibrosis...2 Supplementary Figure 2: Distribution of liver fibrosis stages according to the steps of the EASL algorithm in the ELF group...4 Supplementary Figure 3: Modified EASL algorithm using patented serum tests as the second-line procedure...5 Supplementary Figure 4: Estimated negative and positive predictive values of non-invasive fibrosis tests as a function of the prevalence of advanced fibrosis in the whole biopsy cohort...6 Supplementary Figure 5: Estimated negative and positive predictive values of non-invasive fibrosis tests as a function of the prevalence of advanced fibrosis in the ELF group...7 Supplementary Table 1 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with Fibrotest at different prevalences of advanced fibrosis in the whole biopsy cohort...8 Supplementary Table 2 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with FibroMeterV2G at different prevalences of advanced fibrosis in the whole biopsy cohort...8 Supplementary Table 3 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with Fibrotest at different prevalences of advanced fibrosis in the ELF group...8 Supplementary Table 4 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with FibroMeterV2G at different prevalence of advanced fibrosis in the ELF group...8 Supplementary Table 5 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with ELF at different prevalence of advanced fibrosis in the ELF group...9 Supplementary table 6: Comparison between patients with and without ELF available...10 Supplementary Table 7: Accuracy of non-invasive fibrosis tests, individually or in agreement-based combination (ELF group)...11 Supplementary Table 8: diagnostic accuracy of EASL algorithm in the biopsy cohort as a function of age, sex, diabetes and BMI...12 Supplementary Table 9: Diagnostic accuracy of “modified” EASL algorithm where patented serum tests are used in second-line testing...13 Supplementary Table 10: Patient characteristics of diabetes clinics/primary care cohort already
Supplementary Figure 1: Methods for the estimation of EASL algorithm accuracy at different prevalences of advanced fibrosis
To simulate the accuracy of the EASL algorithm at different prevalences, we calculated the negative (NPV) and positive (PPV) predictive values from pre-specified sensitivities and specificities. Indeed, sensitivity and specificity are intrinsic characteristics of a diagnostic test that are not modified by disease prevalence. The pre-specified sensitivities and specificities used were those published in the latest meta-analyses [21–24]. According to these works, the sensitivities and specificities of FIB4 / VCTE / FMV2G / ELF at the thresholds of the EASL algorithm are, respectively: 74% / 85% / 77% / 65%, and 64% / 72% / 72% / 86%. No robust meta-analysis has been conducted specifically for Fibrotest in NAFLD in the literature. Thus, for this test, we used the sensitivity and specificity obtained in the large biopsy cohort from the present work.
The method we applied is depicted in the Figure. We first used the pre-specified sensitivity and specificity of FIB4 to calculate NPV and PPV of the first-line FIB4 at the desired prevalence. Only patients with positive FIB4 moved to the second-line evaluation. By definition, the prevalence of advanced fibrosis for this second-line evaluation corresponded to the rate of patients with positive FIB4 and advanced fibrosis (true positives) among all patients with positive FIB4, which was the estimated PPV of FIB4. We thus further calculated NPV and PPV of VCTE using the pre-specified sensitivity and specificity of VCTE, and the FIB4 PPV as prevalence. Finally, results for the third-line testing were calculated in the same way using the pre-specified sensitivity and specificity of the concerned patented serum test, and VCTE PPV as prevalence. Details of the calculations and formulas are presented with the results in Supplementary Tables 1 to 5.
* Threshold used for the patented serum tests: ≥0.48 for Fibrotest; ≥0.48 for FibroMeterV2G; ≥9.8 for
Supplementary Figure 2: Distribution of liver fibrosis stages according to the steps of the EASL algorithm in the ELF group.
The size of the bars represents the rate of patients (%) and the numbers in bold the number of patients (n). ELF: enhanced liver fibrosis; F: Fibrosis stage; FMV2G: FibroMeterV2G; VCTE: vibration controlled transient elastography
Supplementary Figure 3: Modified EASL algorithm using patented serum tests as the second-line procedure.
ELF: enhanced liver fibrosis; NAFLD: non-alcoholic fatty liver disease; VCTE: vibration controlled transient elastography
Supplementary Figure 4: Estimated negative and positive predictive values of non-invasive fibrosis tests as a function of the prevalence of advanced fibrosis in the whole biopsy cohort.
The solid line corresponds to the positive predictive value and the dotted line to the negative predictive value. FMV2G: FibroMeterV2G; VCTE: vibration controlled transient elastography
Supplementary Figure 5: Estimated negative and positive predictive values of non-invasive fibrosis tests as a function of the prevalence of advanced fibrosis in the ELF group.
The solid line corresponds to the positive predictive value and the dotted line to the negative predictive value. ELF: enhanced liver fibrosis; FMV2G:
FibroMeterV2G; VCTE: vibration controlled transient elastography
Supplementary Table 1 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with Fibrotest at different prevalences of advanced fibrosis in the whole biopsy cohort.
TN: true negative; FP: false positive; FN: false negative; TP: true positive; DA: diagnostic accuracy;
PPV: positive predictive value; NPV: negative predictive value; +LR: positive likelihood ratio; -LR:
negative likelihood ratio; OR: odds ratio; VCTE: vibration controlled transient elastography (Fibroscan); 2nd test: rate of patients requiring the second-line fibrosis test; 3rd test: rate of patients requiring the third-line fibrosis test; biopsy: rate of patients requiring liver biopsy (%)
Supplementary Table 2 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with FibroMeterV2G at different prevalences of advanced fibrosis in the whole biopsy cohort.
FMV2G: FibroMeterV2G TN: true negative; FP: false positive; FN: false negative; TP: true positive; DA:
diagnostic accuracy; PPV: positive predictive value; NPV: negative predictive value; +LR: positive likelihood ratio; -LR: negative likelihood ratio; OR: odds ratio; VCTE: vibration controlled transient elastography (Fibroscan); 2nd test: rate of patients requiring the second-line fibrosis test; 3rd test: rate of patients requiring the third-line fibrosis test; biopsy: rate of patients requiring liver biopsy (%)
Supplementary Table 3 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with Fibrotest at different prevalences of advanced fibrosis in the ELF group.
TN: true negative; FP: false positive; FN: false negative; TP: true positive; DA: diagnostic accuracy;
PPV: positive predictive value; NPV: negative predictive value; +LR: positive likelihood ratio; -LR:
negative likelihood ratio; OR: odds ratio; VCTE: vibration controlled transient elastography (Fibroscan); 2nd test: rate of patients requiring the second-line fibrosis test; 3rd test: rate of patients requiring the third-line fibrosis test; biopsy: rate of patients requiring liver biopsy (%)
Supplementary Table 4 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with FibroMeterV2G at different prevalence of advanced fibrosis in the ELF group.
FMV2G: FibroMeterV2G; TN: true negative; FP: false positive; FN: false negative; TP: true positive; DA:
diagnostic accuracy; PPV: positive predictive value; NPV: negative predictive value; +LR: positive likelihood ratio; -LR: negative likelihood ratio; OR: odds ratio; VCTE: vibration controlled transient elastography (Fibroscan); 2nd test: rate of patients requiring the second-line fibrosis test; 3rd test: rate of patients requiring the third-line fibrosis test; biopsy: rate of patients requiring liver biopsy (%)
Supplementary Table 5 (see corresponding .xlsx file): Estimation of EASL algorithm accuracy with ELF at different prevalence of advanced fibrosis in the ELF group.
TN: true negative; FP: false positive; FN: false negative; TP: true positive; DA: diagnostic accuracy;
PPV: positive predictive value; NPV: negative predictive value; +LR: positive likelihood ratio; -LR:
negative likelihood ratio; OR: odds ratio; VCTE: vibration controlled transient elastography (Fibroscan); 2nd test: rate of patients requiring the second-line fibrosis test; 3rd test: rate of patients requiring the third-line fibrosis test; biopsy: rate of patients requiring liver biopsy (%)
Supplementary table 6: Comparison between patients with and without ELF available
Patients without ELF Patients with ELF p
(n=655) (n=396)
Age (years) 58.2 (49.9–65.5) 58.0 (49.0–65.3) 0.616
Male sex (%) 57.3 63.1 0.061
BMI (kg/m2) 30.8 (27.3–35.2) 31.6 (28.6–36.1) 0.003
T2DM (%) 51.0 47.7 0.309
Biopsy length (mm) 25 (19–33) 30 (24–35) <0.001
NAS 5.0 (3.0–6.0) 4.5 (2.5–5.0) <0.001
Fibrosis stage (%) <0.001
- F0 9.9 13.6
- F1 21.5 24.5
- F2 25.3 28.5
- F3 24.3 26.8
- F4 18.9 6.6
AST (IU/l) 41 (32–59) 38 (28–52) <0.001
ALT (IU/l) 57 (39–85) 54 (36–78) 0.040
GGT (IU/l) 81 (48–175) 74 (39–132) <0.001
Bilirubin (μmol/l) 10 (8–14) 10 (7–15) 0.090
Platelets (G/L) 215 (174–259) 218 (179–260) 0.198
Albumin (g/l) 43 (40–45) 43 (40–45) 0.574
Prothrombin time (%) 97 (89–101) 96 (89–104) 0.106
FIB4 1.50 (1.03–2.23) 1.38 (0.90–2.04) 0.028
VCTE (kPa) 8.8 (6.1–13.9) 8.3 (5.8–12.4) 0.080
FMV2G 0.46 (0.23–0.72) 0.40 (0.21–0.63) 0.002
Fibrotest 0.44 (0.25–0.68) 0.33 (0.18–0.53) <0.001
ELF - 9.26 (8.55–10.03) -
BMI: body mass index; T2DM: type 2 diabetes mellitus; NAS: NAFLD activity score; AST: aspartate aminotransferase; ALT: alanine aminotransferase; GGT: gamma glutamyl transferase; VCTE: vibration controlled transient elastography; FMV2G: FibroMeterV2G
Supplementary Table 7: Accuracy of non-invasive fibrosis tests, individually or in agreement-based combination (ELF group)
Method Fibrosis test Threshold DA Se Spe NPV PPV -LR +LR OR Biopsy a
Individual FIB4 1.30 65.7 81.1 57.9 86.0 49.1 0.33 1.93 5.9 -
fibrosis test VCTE 8.0 74.2 90.9 65.9 93.5 57.1 0.14 2.67 19.3 -
FMV2G 0.45 74.7 76.5 73.9 86.3 59.4 0.32 2.93 9.2 -
Fibrotest 0.48 72.0 54.5 80.7 78.0 58.5 0.56 2.82 5.0 -
ELF 9.8 80.6 66.7 87.5 84.0 72.7 0.38 5.33 14.0 -
Agreement-based FIB4 and VCTE 1.30 / 8.0 88.6 98.5 83.7 99.1 75.1 0.02 6.05 334.1 37.4
combination FIB4 and FMV2G 1.30 / 0.45 80.6 87.9 76.9 92.7 65.5 0.16 3.80 24.1 20.7
FIB4 and Fibrotest 1.30 / 0.48 85.8 85.6 84.5 92.1 73.4 0.17 5.52 32.3 32.1
FIB4 and ELF 1.30 / 9.8 89.4 85.6 91.3 92.7 83.1 0.16 9.83 62.3 32.6
VCTE and FMV2G 8.0 / 0.45 91.9 97.0 89.4 98.3 82.0 0.03 9.14 269.7 34.8
VCTE and Fibrotest 8.0 / 0.48 92.2 94.7 90.9 97.2 83.9 0.06 10.42 178.6 38.1
VCTE and ELF 8.0 / 9.8 92.9 93.9 92.4 96.8 86.1 0.07 12.4 189.1 31.1
FMV2G and Fibrotest 0.45 / 0.48 82.6 79.5 84.1 89.2 71.4 0.24 5.00 20.6 18.4
FMV2G and ELF 0.45 / 9.8 91.2 85.6 93.3 92.9 87.6 0.15 14.12 92.2 27.0
Fibrotest and ELF 0.48 / 9.8 90.1 78.8 95.8 90.0 90.4 0.22 18.91 85.4 27.8
a Rate (%) of patients requiring liver biopsy because of disagreement between non-invasive tests
DA: diagnostic accuracy (%); Se: sensitivity (%); Spe: specificity (%); NPV: negative predictive value (%); PPV: positive predictive value (%); -LR:
negative likelihood ratio; +LR: positive likelihood ratio; OR: odds ratio; VCTE: vibration controlled transient elastography (Fibroscan); FMV2G:
FibroMeterV2G
Supplementary Table 8: Diagnostic accuracy (rate of well-classified patients) of the EASL algorithm in the biopsy cohort as a function of age, sex, diabetes and BMI.
All biopsy cohort ELF group
Algorithm: FIB4/VCTE/FMV2G FIB4/VCTE/Fibrotest FIB4/VCTE/FMV2G FIB4/VCTE/Fibrotest FIB4/VCTE/ELF Age
(years)
< 45 90.8 92.0 89.6 91.0 91.0
45 - 55 82.3 84.0 87.9 90.1 86.8
55 - 65 78.6 79.4 81.9 83.3 87.0
≥ 65 78.6 80.4 80.0 82.0 86.0
Sex Male 79.0 80.8 82.0 83.6 85.6
Female 85.0 85.7 87.7 89.7 88.4
T2DM No 87.1 88.3 87.9 89.4 92.3
Yes 75.7 77.2 79.9 82.0 82.0
BMI (kg/m2)
< 30 84.9 86.0 87.1 87.9 92.9
30 – 35 81.1 81.4 83.1 85.6 86.0
> 35 76.8 79.5 81.7 85.0 82.5
VCTE: vibration controlled transient elastography (Fibroscan); FMV2G: FibroMeterV2G; T2DM: type 2 diabetes mellitus; BMI: body mass index
Supplementary Table 9: Diagnostic accuracy of “modified” EASL algorithm where patented serum tests are used in second-line testing
Group Algorithm DA Se Spe NPV PPV -LR +LR OR 2nd test a 3rd test b Biopsy
Whole biopsy FIB4 / FMV2G / VCTE 77.3 69.2 88.0 81.4 79.1 0.35 5.79 16.5
57.3 41.9 10.4
cohort FIB4 / Fibrotest / VCTE 76.4 55.2 90.2 78.7 75.5 0.50 5.66 11.4 33.8 8.2
ELF group FIB4 / FMV2G / VCTE 82.8 69.7 89.4 85.6 76.7 0.34 6.57 19.4
55.1
38.6 10.3
FIB4 / Fibrotest / VCTE 78.0 50.0 92.0 78.6 75.9 0.54 6.29 11.6 27.0 6.1
FIB4 / ELF / VCTE 83.6 62.1 94.3 83.3 84.5 0.40 10.93 27.2 26.5 3.0
a 2nd test: rate of patients requiring the second-line fibrosis test (%)
b 3rd test: rate of patients requiring the third-line fibrosis test (%)
DA: diagnostic accuracy (%); Se: sensitivity (%); Spe: specificity (%); NPV: negative predictive value (%); PPV: positive predictive value (%); -LR:
negative likelihood ratio; +LR: positive likelihood ratio; OR: odds ratio; Biopsy: rate of patients requiring liver biopsy (%); VCTE: vibration controlled transient elastography (Fibroscan); FMV2G: FibroMeterV2G
Supplementary Table 10: Patient characteristics of the diabetes clinics/primary care cohort (23).
All (n=230)
Diabetes Clinic (n=90)
Primary Care (n=140)
Age (years) 58.0 (49.0-66.0) 59.5 (52.0-66.0) 58.0 (45.5 – 68.0)
Male sex (%) 54.8 66.7 47.1
BMI (kg/m2) 32.5 (29.2-38.0) 33.4 (30.1-39.6) 32.2 (28.5-37.3)
BMI ≥30 kg/m2 (%) 69.6 76.6 65.0
Type 2 diabetes (%) 81.3 100.0 69.3
AST (IU/L) 24 (16-36) 21 (14-31) 26 (17-41)
ALT (IU/L) 33 (22-54) 29 (21-39) 34 (22-63)
GGT (IU/L) 34 (21-61) 28 (20-50) 37 (21-68)
Platelets (G/L) 241 (204-290) 240 (208-286) 242 (201-296)
Albumin (g/L) 41 (39-43) 41 (39-43) 42 (40-44)
FIB4 0.99 (0.68-1.38) 0.95 (0.69-1.34) 1.00 (0.67-1.50)
VCTE (kPa) 6.1 (4.8-9.2) 6.3 (4.8-9.7) 6.1 (4.6-9.1)
ELF 9.2 (8.6-9.8) 9.3 (8.6-9.7) 9.2 (8.5-10.0)
BMI: body mass index; ALT: alanine aminotransferase; AST: aspartate aminotransferase; GGT: gamma glutamyltransferase; VCTE: vibration controlled transient elastography.
