Table 1.Non-contrast-enhanced ultrasound(NEUS).Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model.Confidence intervals were calculated using bootstrap methods. Non-significant differences are shaded gray.Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
NEUS / Sensitivity % (95% CI) / Comparison Difference% (95% CI) / Positive Predictive Value
% (95% CI) / Comparison
Difference
% (95% CI)
Lesions
< 20 mm / 37.3
(26, 48.9)
n=14 I2=96.4 (93.1, 98.2) / -35.4
(-51, -18) p=2e-05 / 71.1
(55.6, 77.6) n=3
I2=68.2
(0, 80) / -13.1
(-31.4, -3.11)
p=0.01
F1
Lesions
> 20mm / 72.7
(59, 83) n=13 I2=99.3 (98.4, 99.7) / 84.3
(75.9, 89.1) n=3
I2=96.5 (64.3, 99)
Studies performed in Asia / 68.4
(58.3, 76.3) n=13 I2=95.5 (88.5, 98.4) / 14.2
(-1.16, 28.2) p=0.04 / 91.4
(66, 91.4) n=2
I2=93.2
(0, 93.2) / 16.5
(-12.5, 25.9) p=0.2
Studies not performed in Asia / 54.1
(43.1, 66.7) n=16
I2=97
(90.6, 99) / 74.8
(68.8, 80.5) n=8
I2=86.9 (49.2, 94)
Explant-only studies / 48.9
(38.4, 58.4) n=14 I2=95.1 (86.8, 98.1) / -21.8
(-34.4, -8.4) p=0.0005 / 78.3
(66.5, 88.8) n=7
I2=96
(77.4, 98) / 1.46
(-9.74, 12.3) p=0.4
Non-explant-only studies / 70.7
(61, 77.8) n=15 I2=98.7 (94.7, 99.6) / 76.8
(76.2, 77.8) n=3
I2=0
Studies published
≥ 2006 / 68.3
(60.5, 76.2)
n=13 I2=95.7 (87.9, 98.4) / 15
(-0.514, 29.6) p=0.03 / 83.8
(76.2, 91.8) n=4
I2=95.9
(69, 97.3) / 14.9
(4.6, 26.3) p=0.0003
Studies published
< 2006 / 53.3
(41.1, 66.7) n=16 I2=96.9 (90.3, 99) / 68.9
(60.6, 75.1) n=6
I2=62
(0, 86.5)
Table 2. Contrast-enhanced ultrasound (CEUS).Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model.Confidence intervals were calculated using bootstrap methods. Non-significant differences are shaded gray. Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
CEUS / Sensitivity % (95% CI) / Comparison Difference% (95% CI) / Positive Predictive Value
% (95% CI) / Comparison
Difference
% (95% CI)
Lesions
< 20 mm / 67.0
(55.3, 75.8) n=15 I2=92.6 (83.9, 96.5) / -19
(-32.5, -5.33) p=0.003 / 89.6
(89.5, 89.6) n=2
I2=0 / -
Lesions
> 20mm / 86
(73.6, 92.9) n=9
I2=98.8 (78.3, 99.7) / -
Studies performed in Asia / 84.9
(80.9, 87.9) n=36 I2=89.1 (81.4, 94.3) / 3.33
(-3.87, 12.5) p=0.2 / 91
(87.6, 94) n=13 I2=87.9 (61.5, 96.1) / 8.32
(-0.308, 15) p=0.03
F1
Studies not performed in Asia / 81.6
(72.4, 87.7) n=22 I2=94.3 (86.9, 97.3) / 82.7
(76.1, 89.7) n=4
I2=69.4
(0, 88.6)
Explant-only studies / - / - / - / -
Non-explant-only studies / 83.8
(79.4, 86.7) n=59 I2=92.7 (86.9, 96.3) / 89.3
(85.7, 92.5) n=17 I2=89.6 (59.5, 97.7)
Studies published
≥ 2006 / 82.3
(76.9, 85.8) n=45 I2=94.2 (89.7, 97.2) / -6.95
(-13.2, -2.11) p=0.006 / 89.1
(85, 92.7) n=12 I2=92.2 (64.1, 97.5) / -1.39
(-7.99, 8.19) p=0.4
Studies published
< 2006 / 89.2
(85.5, 92.2) n=14 I2=93.6 (78.6, 97.8) / 90.5
(79.3, 95.1) n=5
I2=81.9 (16.7, 94.5)
Table 3.Contrast-enhanced multi-detector or helical computed tomography (CT) studies. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Non-significant differences are shaded gray. Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
CT / Sensitivity % (95% CI) / Comparison Difference% (95% CI) / Positive Predictive Value
% (95% CI) / Comparison
Difference
% (95% CI)
Lesions
< 20 mm / 59.1
(53.9, 63.9) n=54 I2=94.8 (91.6, 97.1) / -27.8
(-34.5, -20)
p=<0.0001 / 77.3
(67, 84.5) n=14 I2=92.2 (83.7, 96.7) / -18.9
(-29.8, -11.2) p=1e-05
Lesions
> 20mm / 86.9
(79.8, 91) n=38 I2=99.4 (97.7, 99.9) / 96.1
(93.5, 98.9) n=8
I2=96
(86.6, 98.6)
Studies performed in Asia / 75.8
(72.3, 79) n=61 I2=92.4 (89.8, 94.6) / 6.29
(0.182, 13.5) p=0.03 / 90
(86.4, 92.2) n=29 I2=77.5 (52.1, 89.9) / 9.86
(4.46, 15.7) p=0.0001
Studies not performed in Asia / 69.5
(62.8, 74.4) n=45 I2=95.8
(92, 97.9) / 80.1
(74.5, 84.3) n=21 I2=84.9 (69.7, 92.1)
Explant-only studies / 67.2
(57.5, 72.8) n=30 I2=95.7 (90.9, 98.3) / -8.33
(-17.5, -1.48) p=0.01 / 81.7
(75.2, 86.7) n=20
I2=92
(83.8, 96.3) / -6.72
(-13.5, -0.804) p=0.02
Non-explant-only studies / 75.5
(72.3, 78.5) n=76
I2=97
(94.5, 98.7) / 88.4
(85.2, 90.9) n=30 I2=93.9 (84.9, 97.6)
Studies published
≥ 2006 / 73.6
(70.3, 76.8) n=63 I2=93.8 (90.8, 96.5) / 1.24
(-5.06, 9.62) p=0.4 / 88.5
(85.0, 90.9) n=33 I2=80.7 (63.7, 91) / 8.18
(1.67, 14.9) p=0.006
Studies published
< 2006 / 72.3
(64.2, 77.6) n=43 I2=96.1 (93.3, 97.8) / 80.3
(73.6, 85.7) n=17
I2=87
(75.4, 92.2)
Table 4.Gadolinium-enhanced magnetic resonance imaging without inclusion of hepatobiliary phase(GdMR).Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Non-significant differences are shaded gray.Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
GdMR / Sensitivity % (95% CI) / Comparison Difference% (95% CI) / Positive Predictive Value
% (95% CI) / Comparison
Difference
% (95% CI)
Lesions
< 20 mm / 63.8
(57.9, 69.7) n=34
I2=95
(86.3, 98.3) / -25
(-32.1, -17.5)
p<0.0001 / 87.5
(77.0, 93.6) n=10 I2=87.2 (48.1, 96.8) / -7.05
(-17.6, -0.529)
p=0.03
F1
Lesions
> 20mm / 88.8
(84.1, 92.9) n=21 I2=97.1
(94, 98.4) / 94.6
(92.7, 98.5) n=6
I2=63.4
(0, 96.3)
Studies performed in Asia / 78.3
(73.5, 82.1) n=43 I2=90.3
(85, 94.1) / 4.71
(-1.84, 10.5) p=0.07 / 86.8
(82.3, 90.5) n=21 I2=80.6 (61.7, 90.8) / 8
(-0.803, 20.6) p=0.08
Studies not performed in Asia / 73.6
(69.4, 78.3) n=32 I2=93.7 (75.3, 98.6) / 78.8
(66, 86.5) n=15 I2=90.9 (79.6, 95.4)
Explant-only studies / 74.5
(70, 79) n=19 I2=68.8 (32.4, 84.9) / -3.09
(-8.71, 2.76) p=0.1 / 73.0
(56.9, 83.4) n=10 I2=91.4 (73.3, 97) / -14.5
(-30, -3.14) p=0.02
Non-explant-only studies / 77.6
(73.8, 81) n=55 I2=95.3 (90.5, 98.6) / 87.5
(83.5, 90.7) n=26 I2=97.2 (89.8, 99.2)
Studies published
≥ 2006 / 78.0
(74.4, 81.5) n=46 I2=93.6 (85.5, 97.7) / 4.7
(-1.83, 11.6) p=0.08 / 88.8
(85, 91.9) n=23 I2=75.3 (49.9, 88) / 14.3
(4.88, 26.4) p=0.002
F1
Studies published
< 2006 / 73.3
(67.1, 78.6) n=29 I2=88.6 (82.1, 92.3) / 74.5
(62.3, 83) n=13 I2=91.1 (79.8, 96)
Table 5.Gadolinium-enhanced magnetic resonance imaging utilizing the hepatobiliary agent gadoxetate.Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Non-significant differences are shaded gray.Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
GxMR / Sensitivity % (95% CI) / Comparison Difference% (95% CI) / Positive Predictive Value
% (95% CI) / Comparison
Difference
% (95% CI)
Lesions
< 20 mm / 83.6
(78.6, 88.5) n=21 I2=97.7 (80.8, 99.4) / -14.4
(-19.8, -9.1) p=0.001 / 94.1
(88.1, 98.1) n=7
I2=91
(19.9, 98.1) / -
Lesions
> 20mm / 98.0
(95.1, 99.3) n=11 I2=95.2 (75.4, 98.6) / -
Studies performed in Asia / 84.4
(80.5, 87.3) n=40 I2=88.4 (80.9, 93.7) / -1.72
(-10.9, 9.65) p=0.3 / 93.9
(90.2, 96.2) n=23 I2=87.2 (56.3, 96.2) / -2.68
(-7.05, 1.34) p=0.1
Studies not performed in Asia / 86.1
(73.8, 94) n=8
I2=99.1 (74.1, 99.8) / 96.5
(66.3, 98.4) n=4
I2=80
(0, 96.7)
Explant-only studies / 80.5
(69.6, 88.4) n=6
I2=94.9 (90.6, 97.2) / -4.97
(-15.6, 4.55) p=0.2 / 97.0
(94.2, 99.4)
n=5
I2=68.9
(0, 89.1) / 3.51
(-0.354, 8.04) p=0.05
Non-explant-only studies / 85.5
(81.5, 88.5) n=42 I2=97.6
(90, 99.7) / 93.5
(89.6, 96.0) n=22 I2=64.9 (19.5, 85.6)
Studies published
≥ 2006 / 84.9
(81.1, 87.8) n=46 I2=95.6 (86.8, 98.7) / -0.458
(-6.58, 18.4) p=0.7 / 94.4
(91.1, 96.5) n=26
I2=84
(52.2, 95.1) / -
Studies published
< 2006 / 85.3
(65.9, 85.3) n=2
I2=79.6
(0, 91.7) / 88.0
n=1
Table 6.Superparamagnetic iron oxide-enhanced magnetic resonance imaging(SpMR).Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model.Confidence intervals were calculated using bootstrap methods. Non-significant differences are shaded gray. Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
SpMR / Sensitivity % (95% CI) / Comparison Difference% (95% CI) / Positive Predictive Value
% (95% CI) / Comparison
Difference
% (95% CI)
Lesions
< 20 mm / 69.1
(63.2, 78.0) n=11 I2=86.9 (2.43, 93.9) / -26.9
(-33.9, -16.4) p=0.003 / 54.2
n=1 / -
Lesions
> 20mm / 96.1
(86.9, 98.7) n=8
I2=97.2 (29.6, 99.2) / 100
n=1
Studies performed in Asia / 79.6
(72.5, 83.8) n=25 I2=93.9 (84.6, 98.1) / -0.0952
(-12, 13.5) p=0.4 / 88.8
(82.0, 93.0) n=14 I2=87.8 (64.2, 96.1) / 18.9
(-4.53, 43.8) p=0.1
F1
Studies not performed in Asia / 79.7
(64.7, 89.1) n=9
I2=97.1
(89, 99) / 69.9
(47.9, 69.9)
n=2
I2=96.4
(0, 98.5)
Explant-only studies / 91.1
(86.2, 95.2) n=8
I2=82.2 (41.2, 92.8) / 14.6
(7.9, 22.8) p=2e-05 / 84.5
(59.4, 94.4) n=5
I2=92
(0, 98.6) / -2.56
(-22.9, 10.1) p=0.4
Non-explant-only studies / 76.4
(69.6, 81.5) n=26 I2=93.3 (85.7, 96.8) / 87
(79.3, 92.5) n=11 I2=94.5 (75.9, 99.5)
Studies published
≥ 2006 / 80.6
(70.6, 86.2) n=16
I2=95
(84.5, 98.6) / 2.02
(-8.54, 11.9) p=0.3 / 93.3
(88.7, 96.5) n=8
I2=68.5
(0, 89.1) / 13.6
(4.54, 26) p=0.003
F1
Studies published
< 2006 / 78.5
(69.7, 84.5) n=18 I2=94.6 (86.2, 98.3) / 79.7
(67.6, 87.8) n=8
I2=90.4
(60, 95.8)
Table 7.Overall inter-modality sensitivity overall comparison.Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods.Off-diagonal entries are estimations of differences between modalities.Non-significant differences are shaded gray.
Overall Comparison / GdMR / GxMR / SpMR / CT / CEUS / NEUSGdMR / 77.5
(73.1, 79.3) n=75 I2=93.1 (87.6, 97.2) / -8.1
(-12.9, -3.9) p=0.0009 / -3.0
(-8.8, 3.1) p=0.1 / 3.9
(-1.1, 7.7) p=0.08 / -6.9
(-11.9, -2.5) p=0.001 / 18.2
(8.4, 25.3) p=0.0003
GxMR / — / 85.6
(81.1, 87.7) n=48
I2=95.5
(86.6, 98.7) / 5.1
(-0.4, 11.6)
p=0.04 / 12.0
(7.2, 16.2) p<0.0001 / 1.2
(-3.5, 6.1) p=0.4 / 26.3
(16.8, 33.8) p<0.0001
SpMR / — / — / 80.5
(73.6, 83.9) n=34 I2=96.2 (91.3, 98.2) / 6.9
(0.3, 12.1) p=0.02 / -3.9
(-10.5, 1.9) p=0.08 / 21.2
(10.5, 29.1) p=2e-05
CT / — / — / — / 73.6
(69.7, 76) n=106 I2=94.2 (91.5, 96.4) / -10.8
(-15.2, -5.8) p=1e-05 / 14.3
(5.1, 21.9) p=0.001
CEUS / — / — / — / — / 84.4
(79.4, 86.7) n=59
I2=92.7 (86.9, 96.3) / -25.1
(-32.7, -15.6)
p<0.0001
NEUS / — / — / — / — / — / 59.3
(51.3, 67.1) n=29
I2=96.7
(92.9, 98.8)
Table 8. Inter-modality positive predictive value overall comparison. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray.Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
Overall Comparison / GdMR / GxMR / SpMR / CT / CEUS / NEUSGdMR / 83.6
(77.2, 87.5) n=36 I2=89.8 (80.2, 95.6) / -10.6
(-17.3, -5.8) p<0.0001
F1 / -2.78
(-11, 5.2) p=0.2 / -2.22
(-8.6, 2.9) p=0.2 / -5.74
(-12.8, -0.5) p=0.03
F1 / 6.22
(-3.7, 13.9) p=0.08
GxMR / — / 94.2
(90.9, 96.3) n=27
I2=84.4
(54.2, 95.1) / 7.82
(2.0, 15.9) p=0.008
F1 / 8.38
(4.6, 12.4) p=1e-05
F1 / 4.86
(0.4, 8.9) p=0.01 / 16.8
(8.6, 23.7) p=3e-05
F1
SpMR / — / — / 86.4
(77.9, 91.4) n=16 I2=84.2 (57.2, 92.7) / 0.566
(-7.1, 7.0) p=0.4 / -2.96
(-11.6, 3.2) p=0.2
F1 / 9
(-2.1, 17.3) p=0.03
F1
CT / — / — / — / 85.8
(82.4, 88.4) n=50 I2=87.8 (80.6, 92.8) / -3.52
(-8.2, 0.7) p=0.06
F1 / -8.44
(0.3, 15.6) p=0.01
F1
CEUS / — / — / — / — / 89.3
(85.7, 92.5) n=17
I2=89.6 (59.5, 97.7) / -12
(-19.2, -3.6) p=0.001
F1
NEUS / — / — / — / — / — / 77.4
(71.1, 85.5) n=10 I2=94.5 (41.1, 97.6)
Table 9. Inter-modality sensitivity comparison for lesions < 20 mm in size. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray.
Lesions< 20 mm / GdMR / GxMR / SpMR / CT / CEUS / NEUS
GdMR / 63.8
(57.9, 69.7) n=34
I2=95
(86.3, 98.3) / -19.8
(-27.3, -12) p<0.0001 / -5.3
(-14.8, 3.7) p=0.1 / 4.73
(-2.8, 12.5) p=0.1 / -3.18
(-13.9, 9.5) p=0.3 / 26.5
(14.0, 39.7) p=5e-05
GxMR / — / 83.6
(78.6, 88.5) n=21
I2=97.7
(80.8, 99.4) / 14.5
(5.1, 22.5) p=0.0008 / 24.5
(17.6, 31.7) p<0.0001 / 16.6
(6.3, 29) p=0.0009 / 46.3
(33.7, 58.7) p<0.0001
SpMR / — / — / 69.1
(63.2, 78.0) n=11 I2=86.9 (2.43, 93.9) / 10
(2.0, 19.5) p=0.01 / 2.12
(-8.8, 16.3) p=0.4 / 31.8
(18.7, 45.7) p<0.0001
CT / — / — / — / 59.1
(53.9, 63.9) n=54 I2=94.8 (91.6, 97.1) / -7.91
(-18.6, 3.8) p=0.08 / 21.8
(9.2, 34.2) p=0.0005
CEUS / — / — / — / — / 67
(55.3, 75.8) n=15
I2=92.6 (83.9, 96.5) / 29.7
(13.8, 44.3) p=0.0001
NEUS / — / — / — / — / — / 37.3
(26.0, 48.9)
n=14 I2=96.4 (93.1, 98.2)
Table 10. Inter-modality positive predictive value comparison for lesions < 20 mm in size. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray. Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
Lesions< 20 mm / GdMR / GxMR / SpMR / CT / CEUS / NEUS
GdMR / 87.5
(77, 93.6) n=10 I2=87.2 (48.1, 96.8) / -6.6
(-17.3, 1.08) p=0.07
F1 / — / 10.2
(-1.2, 21.8) p=0.04 / -2.08
(-12.4, 4.02) p=0.3 / 16.4
(3.0, 29.7) p=0.003
F1
GxMR / — / 94.1
(88.1, 98.1) n=7
I2=91
(19.9, 98.1) / — / 16.8
(8.03, 27.9) p=0.0003
F1 / 4.55
(-1.5, 8.5) p=0.04
F1 / 23
(13.2, 37.7) p=9e-05
F1
SpMR / — / — / 54.2
n=1 / — / — / —
CT / — / — / — / 77.3
(67, 84.5) n=14 I2=92.2 (83.7, 96.7) / -12.3
(-22.4, -5) p=0.0006 / 6.17
(-7.5, 20.4) p=0.2
CEUS / — / — / — / — / 89.6
(89.5, 89.6) n=2
I2=0 / 18.5
(12.0, 34.1) p<0.0001
F1
NEUS / — / — / — / — / — / 71.1
(55.6, 77.6) n=3
I2=68.2
(0, 80)
Table 11.Inter-modality sensitivity comparison for explant-only studies. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray.
Explant only / GdMR / GxMR / SpMR / CT / CEUS / NEUSGdMR / 74.5
(70, 79) n=19 I2=68.8 (32.4, 84.9) / -6
(-15.6, 5.61) p=0.1 / -16.6
(-22.8, -10) p<0.0001 / 7.3
(0.17, 17.8) p=0.04 / — / 25.6
(14.9, 36.8) p<0.0001
GxMR / — / 80.5
(69.6, 88.4) n=6
I2=94.9
(90.6, 97.2) / -10.6
(-22.2, -1.05) p=0.02 / 13.3
(2.3, 26.3) p=0.01 / — / 31.6
(17.5, 45.1) p=1e-05
SpMR / — / — / 91.1
(86.2, 95.2) n=8
I2=82.2
(41.2, 92.8) / 23.9
(16.7, 34.6) p<0.0001 / — / 42.1
(31.6, 53.6) p<0.0001
CT / — / — / — / 67.2
(57.5, 72.8) n=30 I2=95.7 (90.9, 98.3) / — / 18.2
(4.7, 29.6) p=0.002
CEUS / — / — / — / — / — / —
NEUS / — / — / — / — / — / 48.9
(38.4, 58.4) n=14 I2=95.1 (86.8, 98.1)
Table 12. Inter-modality positive predictive value comparison for explant-only studies. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayesestimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray.Entries which include “F1” indicate that the F1 score difference confidence interval met significance.
Explant only / GdMR / GxMR / SpMR / CT / CEUS / NEUSGdMR / 73.0
(56.9, 83.4) n=10 I2=91.4 (73.3, 97) / -24.0
(-40.5, -13.1) p=0.009
F1 / -11.5
(-29.8, 8.8) p=0.1 / -8.7
(-24.7, 3.43) p=0.1 / — / -5.3
(-24.8, 9.3) p=0.3
GxMR / — / 97.0
(94.2, 99.4) n=5
I2=68.9
(0, 89.1) / 12.5
(2.02, 37.8) p=0.03 / 15.3
(9.36, 22.4) p=0.01
F1 / — / 18.7
(5.11, 30.3) p=0.01
F1
SpMR / — / — / 84.5
(59.4, 94.4) n=5
I2=92
(0, 98.6) / 2.79
(-16.5, 15.4) p=0.3 / — / 6.21
(-18.1, 20.5) p=0.2
F1
CT / — / — / — / 81.7
(75.2, 86.7) n=20
I2=92
(83.8, 96.3) / — / 3.42
(-8.6, 16.2) p=0.3
F1
CEUS / — / — / — / — / — / —
NEUS / — / — / — / — / — / 78.3
(66.5, 88.8) n=7
I2=96
(77.4, 98)
Table 13. Inter-modality sensitivity comparison for studies published in or after 2006. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray.
Studies published≥ 2006 / GdMR / GxMR / SpMR / CT / CEUS / NEUS
GdMR / 78.0
(74.4, 81.5) n=46 I2=93.6 (85.5, 97.7) / -6.8
(-11.5, -1.86) p=0.003 / -2.53
(-9.6, 6.8) p=0.3 / 4.45
(-0.4, 9.2)
p=0.03 / -4.26
(-9.4, 1.9) p=0.07 / 9.73
(1.01, 18.3) p=0.01
GxMR / — / 84.9
(81.1, 87.8) n=46
I2=95.6
(86.8, 98.7) / 4.29
(-2.8, 13.3) p=0.2 / 11.3
(6.5, 15.7) p<0.0001 / 2.57
(-2.5, 8.5) p=0.2 / 16.6
(7.8, 24.8)
p=1e-05
SpMR / — / — / 80.6
(70.6, 86.2) n=16
I2=95
(84.5, 98.6) / 6.98
(-2.2, 13.9) p=0.04 / -1.72
(-10.6, 6.3) p=0.4 / 12.3
(0.1, 21.8)
p=0.01
CT / — / — / — / 73.6
(70.3, 76.8) n=63 I2=93.8 (90.8, 96.5) / -8.7
(-13.6, -2.8) p=0.001 / 5.28
(-3.3, 13.8)
p=0.1
CEUS / — / — / — / — / 82.3
(76.9, 85.8) n=45
I2=94.2 (89.7, 97.2) / 14
(4.2, 22.4) p=7e-04
NEUS / — / — / — / — / — / 68.3
(60.5, 76.2)
n=13 I2=95.7 (87.9, 98.4)
Table 14. Inter-modality positive predictive value comparison for studies published in or after 2006. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray.Entries which include “F1” indicate that the F1 score confidence interval met significance.
Studies published≥ 2006 / GdMR / GxMR / SpMR / CT / CEUS / NEUS
GdMR / 88.8
(85, 91.9) n=23 I2=75.3 (49.9, 88) / -5.6
(-9.9, -1.25) p=0.006
F1 / -4.5
(-9.62, 0.60) p=0.05 / 0.3
(-4.01, 4.92) p=0.4 / -0.3
(-5.6, 4.7) p=0.5
F1 / 5.1
(-5.04, 12.9) p=0.1
F1
GxMR / — / 94.4
(91.1, 96.5) n=26
I2=84
(52.2, 95.1) / 1.1
(-3.64, 5.56) p=0.3 / 5.9
(2.12, 9.85) p=0.002
F1 / 5.3
(0.385, 9.68) p=0.01 / 10.6
(0.638, 18.1) p=0.01
F1
SpMR / — / — / 93.3
(88.7, 96.5) n=8
I2=68.5
(0, 89.1) / 4.8
(0.279, 9.76) p=0.03
F1 / 4.2
(-1.25, 9.54) p=0.06 / 9.5
(-0.57, 17.5) p=0.02
F1
CT / — / — / — / 88.5
(85, 90.9) n=33 I2=80.7 (63.7, 91) / -0.6
(-5.72, 3.9) p=0.4
F1 / 4.7
(-5.65, 12.1) p=0.2
F1
CEUS / — / — / — / — / 89.1
(85, 92.7) n=12
I2=92.2 (64.1, 97.5) / 5.3
(5.6, 13.5) p=0.1
F1
NEUS / — / — / — / — / — / 83.8
(76.2, 91.8) n=4
I2=95.9
(69.0, 97.3)
Table 15. Inter-modality sensitivity comparison for studies notoriginating in Asia. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray.
Studies not originating in Asia / GdMR / GxMR / SpMR / CT / CEUS / NEUSGdMR / 73.6
(69.4, 78.3)
n=32 I2=93.7 (75.3, 98.6) / -12.5
(-21.6, 0.1) p=0.01 / -6.1
(-16.6, 8.8)
p=0.1 / 4.07
(-2.4, 12.1)
p=0.1 / -7.96
(-15.6, 1.9) p=0.03 / 19.5
(6.6, 31.8) p=0.002
GxMR / — / 86.1
(73.8, 94.0)
n=8
I2=99.1
(74.1, 99.8) / 6.44
(-7.7, 22.2)
p=0.2 / 16.6
(5.1, 27.3) p=0.003 / 4.53
(-7.2, 17.0) p=0.2 / 32.0
(15.0, 45.9) p=4e-05
SpMR / — / — / 79.7
(64.7, 89.1) n=9
I2=97.1
(89, 99) / 10.1
(-3.2, 22.3) p=0.05 / -1.91
(-15.6, 11.7)
p=0.4 / 25.5
(7.2, 40.6) p=0.002
CT / — / — / — / 69.5
(62.8, 74.4) n=45 I2=95.8
(92, 97.9) / -12
(-21.3, -2.7) p=0.005 / 15.4
(1.6, 27.9) p=0.01
CEUS / — / — / — / — / 81.6
(72.4, 87.7) n=22
I2=94.3 (86.9, 97.3) / 27.5
(12.1, 40.2)
p=0.0001
NEUS / — / — / — / — / — / 54.1
(43.1, 66.7) n=16
I2=97
(90.6, 99)
Table 16.Inter-modality positive predictive value comparison for studies not originating in Asia. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray. Entries which include “F1” indicate that the F1 score confidence interval met significance.
Studies not originating in Asia / GdMR / GxMR / SpMR / CT / CEUS / NEUSGdMR / 78.8
(66, 86.5) n=15 I2=90.9 (79.6, 95.4) / -17.8
(-31.1, -9.6) p=0.004
F1 / 8.88
(-18.8, 35.8) p=0.3 / -1.33
(-14.2, 8.0) p=0.5 / -3.9
(-18.7, 5.9) p=0.3 / 3.9
(-10.2, 13.3) p=0.2
GxMR / — / 96.5
(66.3, 98.4) n=4
I2=80
(0, 96.7) / 26.7
(4.83, 50.3) p=0.003 / 16.4
(11.1, 22.6) p=0.004
F1 / 13.8
(4.8, 20.3) p=0.004 / 21.7
(14.7, 28.2) p=0.004
F1
SpMR / — / — / 69.9
(47.9, 69.9)
n=2
I2=96.4
(0, 98.5) / -10.2
(-35.0, 12.8) p=0.3 / -12.8
(-39.4, 10.8) p=0.2
F1 / -4.9
(-30.6, 18.3) p=0.3
CT / — / — / — / 80.1
(74.5, 84.3) n=21 I2=84.9 (69.7, 92.1) / -2.6
(-12.3, 4.8) p=0.3
F1 / 5.3
(-2.8, 12.6) p=0.08
CEUS / — / — / — / — / 82.7
(76.1, 89.7) n=4
I2=69.4
(0, 88.6) / 7.9
(0.8, 17.3) p=0.03
F1
NEUS / — / — / — / — / — / 74.8
(68.8, 80.5) n=8
I2=86.9 (49.2, 94)
Table 17. Inter-modality sensitivity comparison for studies with internalcomparisons. Off-diagonal entries are estimations of differences between modalities. Pooled per-lesion values are empirical Bayes estimates using a beta-binomial model. Confidence intervals were calculated using bootstrap methods. Off-diagonal entries are estimations of differences between modalities. Non-significant differences are shaded gray
Within study comparison / GxMR (Gadoxetate) / CTGxMR (Gadoxetate) / 83.8
(79.3, 87.9) n=18
I2=83.8 (68.7, 91.5) / 13.9
(7.3, 20.5)p<0.0001
CT / — / 69.8
(64.5, 74.6) n=18
I2=89.5
(79.9, 94.6)
Within study comparison / GdMR / CT
GdMR / 71.2
(64.8, 77.1) n=24
I2=97.3
(87, 99.2) / 8.08
(-0.7, 19.5)
p=0.05
CT / — / 63.1
(52.6, 69.2) n=24
I2=95.1
(87.7, 98.3)
Within study comparison / GdMR / CEUS
GdMR / 73.9
(46.0, 85.6) n=4
I2=92.2
(0, 95.9) / 8.7
(-14.4, 32.2) p=0.3
CEUS / — / 65.3
(44.1, 77.2) n=4
I2=93.5
(0, 98.5)
Within study comparison / SpMR / CT
SpMR / 81
(65.5, 89.4) n=9
I2=97.7 (90.5, 99.2) / 8.7
(-8.0, 21.7) p=0.1
CT / — / 72.3
(62, 81.4)
n=9
I2=94.7
(83.6, 97.8)
Within study comparison / GxMR (Gadoxetate) / CEUS
GxMR (Gadoxetate) / 81.7
(73.5, 91.9) n=6
I2=14.3
(0, 67.8) / 15.9
(2.93, 29.4) p=0.007
CEUS / — / 65.8
(56.1, 75) n=6
I2=87.4
(51, 95.5)
Within study comparison / CT / CEUS
CT / 79.4
(71.2, 85.6) n=14
I2=91.3 (83.7, 95.6) / 0.393
(-10.3, 13.7) p=0.5
CEUS / — / 79
(66.4, 86.6) n=14
I2=94.7
(86.2, 98)