TABLE 2.
Diagnostic test results, per‐protocol and intention‐to‐diagnose analyses
| Training cohort | Advanced fibrosis ≥ F3 | Cirrhosis = F4 | ||||
|---|---|---|---|---|---|---|
| MFAP4 | TE | ELF | MFAP4 | TE | ELF | |
| Prevalence, n (%) | 42/153 (27) | 37/138 (27) | 42/153 (27) | 28/153 (18) | 23/138 (17) | 28/153 (18) |
| AUROC (95% CI) | 0.88 (0.81‐0.94) | 0.95 (0.91‐0.98) | 0.91 (0.85‐0.96) | 0.91 (0.83‐0.98) | 0.95 (0.92‐0.99) | 0.95(0.91‐0.98) |
| AUROC vs AUROC‐MFAP4 | — | P = .027 | P = .348 | — | P = .234 | P = .257 |
| Brier test | 0.109 | 0.095 | 0.098 | 0.082 | 0.071 | 0.072 |
| Hosmer‐Lemeshow test |
4.27 (P = .832) |
26.01 (P = .001) |
6.85 (P = .553) |
6.72 (P = .567) |
8.22 (P = .412) |
2.98 (P = .935) |
| Optimal cut‐off | 88.7 Y | 15.5 L | 10.5 L | 88.7 Y | 19.7 L | 11.1 Y |
| Correctly classifies n (%) | 132 (86) | 124 (90) | 132 (86) | 132 (86) | 123 (89) | 139 (91) |
| TP/FP/FN/TN | 32/11/10/100 | 32/9/5/92 | 34/13/8/98 | 25/18/3/107 | 22/14/1/101 | 25/11/3/114 |
| Sensitivity (%) | 76 (61‐88) | 87 (71‐96) | 81 (66‐91) | 89 (72‐98) | 96 (78‐100) | 89 (72‐98) |
| Specificity (%) | 90 (83‐95) | 91 (84‐96) | 88 (81‐94) | 86 (78‐91) | 88 (80‐93) | 91 (85‐96) |
| PPV (%) | 74 (59‐87) | 78 (62‐89) | 72 (57‐84.4) | 58 (42‐73) | 61 (44‐77) | 69 (52‐84) |
| NPV (%) | 91 (84‐96) | 95 (88‐98) | 93 (86‐97) | 97 (92‐99) | 99 (95‐100) | 97 (93‐100) |
| Pretest odds | 0.37 | 0.37 | 0.37 | 0.22 | 0.20 | 0.22 |
| LR (+) | 7.69 (4.28‐13.8) | 9.71 (5.14‐18.3) | 6.91 (4.06‐11.8) | 6.2 (3.97‐9.69) | 7.86 (4.77‐12.9) | 10.1 (5.69‐18.1) |
| LR (−) | 0.26 (0.15‐0.46) | 0.15 (0.07‐0.34) | 0.22 (0.12‐0.40) | 0.13 (0.04‐0.37) | 0.05 (0.01‐0.24) | 0.12 (0.04‐0.34) |
| Validation cohort | MFAP4 | TE | ELF | MFAP4 | TE | ELF |
| Prevalence n (%) | 20/113 (18) | 18/110 (16) | 20/113 (18) | 17/113 (15) | 15/110 (14) | 17/113 (15) |
| AUROC (95% CI) | 0.92 (0.83‐1.00) | 0.98 (0.96‐1.00) | 0.94 (0.88‐0.99) | 0.91 (0.79‐1.00) | 0.97 (0.95‐1.00) | 0.92 (0.85‐0.98) |
| AUROC vs AUROC‐MFAP4 | — | P = .247 | P =.84 | — | P = .255 | P = .865 |
| Brier test | 0.061 | 0.047 | 0.069 | 0.062 | 0.058 | 0.073 |
| Hosmer‐Lemeshow test |
14.86 (P = .062) |
16.01 (P = .042) |
9.17 (P = .328) |
14.76 (P = .064) |
15.13 (P = .057) |
5.30 (P = .725) |
| Cut‐off | 88.7 Y | 15.5 L | 10.5 L | 88.7 Y | 19.7 L | 11.1 Y |
| Correctly classifies n (%) | 104 (92) | 108 (98) | 100 (88) | 105 (93) | 102 (93) | 101 (89) |
| TP/FP/FN/TN | 11/0/9/93 | 18/2/0/90 | 14/7/6/86 | 10/1/7/95 | 12/5/3/90 | 10/5/7/91 |
| Sensitivity (%) | 55 (32‐77) | 100 (82‐100) | 70 (46‐88) | 59 (33‐82) | 80 (52‐96) | 59 (33‐82) |
| Specificity (%) | 100 (96‐100) | 98 (92‐100) | 93 (85‐97) | 99 (94‐100) | 95 (88‐98) | 95 (88‐98) |
| PPV (%) | 100 (72‐100) | 90 (68‐99) | 67 (43‐85) | 91 (59‐100) | 71 (44‐90) | 67 (38‐88) |
| NPV (%) | 91 (84‐96) | 100 (96‐100) | 94 (86‐98) | 93 (86‐97) | 97 (91‐99) | 93 (86‐97) |
| Pretest odds | 0.22 | 0.20 | 0.22 | 0.18 | 0.16 | 0.18 |
| LR (+) | (High) | 46 (11.7‐181) | 9.3 (4.31‐20) | 56.5 (7.72‐413) | 15.2 (6.24‐37) | 11.3 (4.41‐29) |
| LR (−) | 0.45 (0.28‐0.73) | 0 | 0.32 (0.17‐0.63) | 0.42 (0.24‐0.74) | 0.21 (0.08‐0.58) | 0.43 (0.25‐0.77) |
| Total cohort | MFAP4 | TE | ELF | MFAP4 | TE | ELF |
| Prevalence n (%) | 62/266 (23) | 55/248 (22) | 62/266 (23) | 45/266 (17) | 38/248 (15) | 45/266 (17) |
| AUROC (95% CI) | 0.90 (0.85‐0.95) | 0.96 (0.94‐0.99) | 0.92 (0.88‐0.96) | 0.90 (0.84‐0.96) | 0.96 (0.94‐0.98) | 0.93 (0.90‐0.97) |
| AUROC vs AUROC‐MFAP4 | — | P = .011 | P = .440 | — | P = .061 | P = .314 |
| Brier test | 0.0922 | 0.076 | 0.086 | 0.079 | 0.102 | 0.072 |
| Hosmer‐Lemeshow test |
7.28 (P = .506) |
50.13 (P = .000) |
10.85 (P = .210) |
14.09 (P = .080) |
23.18 (P = .003) |
3.32 (P = .912) |
| Cut‐off | 62.0 Y | 15.5 L | 10.5 L | 60.3 Y | 19.7 L | 10.1 Y |
| Correctly classifies n (%) | 224 (84) | 232 (94) | 232 (87) | 214 (80) | 225 (91) | 218 (82) |
| TP/FP/FN/TN | 53/33/9/171 | 50/11/5/182 | 48/20/14/184 | 42/49/3/172 | 34/19/4/191 | 42/45/3/176 |
| Sensitivity (%) | 86 (74‐93) | 91 (80‐97) | 77 (64‐87) | 93 (82‐99) | 90 (75‐97) | 93 (82‐99) |
| Specificity (%) | 84 (78‐89) | 84 (90‐97) | 90 (85‐94) | 77.8 (72‐99) | 91 (86‐95) | 80 (74‐85) |
| PPV (%) | 62 (51‐72) | 82 (70‐91) | 71 (58‐81) | 46 (36‐57) | 64 (50‐77) | 48 (37‐59) |
| NPV (%) | 95 (91‐98) | 97 (94‐99) | 93 (88‐96) | 98 (95‐100) | 98 (95‐99) | 98 (95‐100 |
| Pretest odds | 0.30 | 0.28 | 0.30 | 0.20 | 0.18 | 0.20 |
| LR (+) | 5.28 (3.8‐7.34) | 16 (8.93‐28.5) | 7.9 (5.1‐12.2) | 4.21 (3.25‐5.45) | 9.89 (6.35‐15.4) | 4.58 (3.49‐6.02) |
| LR (−) | 0.17 (0.09‐0.32) | 0.10 (0.04‐0.22) | 0.25 (0.16‐0.40) | 0.09 (0.03‐0.26) | 0.12 (0.05‐0.29) | 0.08 (0.03‐0.25) |
| Intention‐to‐diagnose analysis | MFAP4 | TE | ELF | MFAP4 | TE | ELF |
| Prevalence n (%) | 20/113 (18) | 20/113 (18) | 20/113 (18) | 17/113 (15) | 17/113 (15) | 17/113 (15) |
| AUROC (95% CI) | 0.92 (0.83‐1.00) | 0.93 (0.83‐1.00) | 0.94 (0.88‐0.99) | 0.91 (0.79‐1.00) | 0.91 (0.80‐1.00) | 0.92 (0.85‐0.98) |
| AUROC vs AUROC‐MFAP4 | — | P = .918 | P = .84 | — | P = .924 | P = .865 |
| Brier test | 0.061 | 0.063 | 0.069 | 0.062 | 0.067 | 0.073 |
| Hosmer‐Lemeshow test |
14.86 (P = .062) |
13.05 (P = .110) |
9.17 (P = .328) |
14.76 (P = .064) |
12.43 (P = .133) |
5.30 (P = .725) |
| Cut‐off | 88.7 Y | 15.5 L | 10.5 L | 88.7 Y | 19.7 L | 11.1 Y |
| Correctly classifies n (%) | 104 (92) | 109 (96) | 100 (88) | 105 (93) | 103 (91) | 101 (89) |
| TP/FP/FN/TN | 11/0/9/93 | 19/3/1/90 | 14/7/6/86 | 10/1/7/95 | 13/6/4/90 | 10/5/7/91 |
| Sensitivity (%) | 55 (32‐77) | 95 (75‐100) | 70 (46‐88) | 59 (33‐82) | 77 (50‐93) | 59 (33‐82) |
| Specificity (%) | 100 (96‐100) | 97 (91‐99) | 93 (85‐97) | 99 (94‐100) | 94 (87‐98) | 95 (88‐98) |
| PPV (%) | 100 (72‐100) | 86 (65‐97) | 67 (43‐85) | 91 (59‐100) | 68 (43‐87) | 67 (38‐88) |
| NPV (%) | 91 (84‐96) | 99 (94‐100) | 94 (86‐98) | 93 (86‐97) | 96 (90‐99) | 93 (86‐97) |
| Pretest odds | 0.22 | 0.22 | 0.22 | 0.18 | 0.19 | 0.18 |
| LR (+) | (High) | 29.4 (9.63‐90.1) | 9.3 (4.31‐20) | 56.5 (7.72‐413) | 12.2 (5.4‐28) | 11.3 (4.41‐29) |
| LR (−) | 0.45 (0.28‐0.73) | 0.05 (0.01‐0.35) | 0.32 (0.17‐0.63) | 0.42 (0.24‐0.74) | 0.25 (0.11‐0.59) | 0.43 (0.25‐0.77) |
Y in the cut‐off indicates that the value was identified by optimizing the Youden index and an L indicates that the value was identified based on published literature.
Abbreviations: AUROC, area under the receiver operating characteristics curve; TP, true positive; FP, false positive; FN, false negative; TN, true negative; PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio.