6. Algorithm and threshold analysis for each definition of the target condition.
|
Target condition Testa |
Datasets (n) |
Lesions (cases) |
Pooled sensitivity (95% CI) |
Pooled specificity (95% CI) |
Datasets (n) |
Lesions (cases) |
Pooled sensitivity (95% CI) |
Pooled specificity (95% CI) |
| a. Invasive melanoma and atypical intraepidermal melanocytic variants | In‐person | Image‐based | ||||||
| No algorithm: any threshold | 8 | 4707 (849) | 0.88 (0.75 to 0.95) |
0.87 (0.80 to 0.92) |
24 | 4498 (941) | 0.76 (0.70 to 0.82) |
0.79 (0.71 to 0.85) |
| No algorithm: correct diagnosis | ‐ | ‐ | ‐ | ‐ | 18 | 4118 (795) | 0.77 (0.69 to 0.83) |
0.84 (0.76 to 0.89) |
| No algorithm: excise decision | ‐ | ‐ | ‐ | ‐ | 10 | 831 (263) | 0.79 (0.69 to 0.86) |
0.55 (0.50 to 0.61) |
| Pattern: any threshold or NR | 6 | 4307 (296) | 0.92 (0.87 to 0.95) |
0.92 (0.68 to 0.98) |
20 | 4621 (989) | 0.83 (0.76 to 0.88) |
0.87 (0.80 to 0.92) |
| Pattern: at ≥ 1 characteristics present | 1 | 220 (33) | 0.88 (0.72 to 0.97) |
0.79 (0.73 to 0.85) |
‐ | ‐ | ‐ | ‐ |
| Pattern: at ≥ 3 characteristics present | 1 | 68 (5) | 1.00 (0.48 to 1.00) |
0.56 (0.42 to 0.68) |
‐ | ‐ | ‐ | ‐ |
| Pattern: correct diagnosis | ‐ | ‐ | ‐ | ‐ | 19 | 4095 (896) | 0.81 (0.73 to 0.87) |
0.87 (0.80 to 0.92) |
| Pattern: excise decision | ‐ | ‐ | ‐ | ‐ | 3 | 933 (227) | 0.97 (0.68 to 1.00) |
0.72 (0.60 to 0.81) |
| ABCD at NR (likely > 5.45) | 1 | 235 (5) | 1.00 (0.48 to 1.00) |
0.90 (0.85 to 0.93) |
‐ | ‐ | ‐ | ‐ |
| ABCD at > 5.45 | 4 | 1203 (155) | 0.78 (0.58 to 0.90) |
0.93 (0.79 to 0.98) |
7 | 2471 (406) | 0.81 (0.60 to 0.92) |
0.81 (0.69 to 0.89) |
| ABCD at or likely > 5.45 (2 previous groups combined) | 5 | 1438 (160) | 0.81 (0.62 to 0.92) |
0.92 (0.82 to 0.97) |
‐ | ‐ | ‐ | ‐ |
| ABCD at > 4.75 | 1 | 309 (73) | 0.83 (0.69 to 0.92) |
0.45 (0.39 to 0.51) |
10 | 4242 (816) | 0.81 (0.67 to 0.90) |
0.72 (0.93 to 0.80) |
| Revised ABCD at ≥ 4 | ‐ | ‐ | ‐ | ‐ | 1 | 269 (84) | 0.87 (0.78 to 0.93) |
0.89 (0.83 to 0.93) |
| ABCD at 60% specificity | 1 | 356 (73) | 0.90 (0.81 to 0.96) |
0.60 (0.54 to 0.66) |
‐ | ‐ | ‐ | ‐ |
| ABCD at 70% specificity | 1 | 356 (73) | 0.85 (0.75 to 0.92) |
0.70 (0.64 to 0.75) |
‐ | ‐ | ‐ | ‐ |
| ABCD at 75% specificity | 1 | 356 (73) | 0.85 (0.75 to 0.92) |
0.75 (0.69 to 0.80) |
‐ | ‐ | ‐ | ‐ |
| ABCD at 80% specificity | 1 | 356 (73) | 0.77 (0.65 to 0.86) |
0.80 (0.75 to 0.84) |
‐ | ‐ | ‐ | ‐ |
| ABCD at 85% specificity | 1 | 356 (73) | 0.71 (0.59 to 0.81) |
0.85 (0.80 to 0.89) |
‐ | ‐ | ‐ | ‐ |
| ABCD at 90% specificity | 1 | 356 (73) | 0.64 (0.52 to 0.75) |
0.90 (0.86 to 0.93) |
‐ | ‐ | ‐ | ‐ |
| ABCDE at > 1.3 | 1 | 356 (73) | 1.00 (0.95 to 1.00) |
0.15 (0.11 to 0.20) |
‐ | ‐ | ‐ | ‐ |
| ABCDE at > 2.65 | 1 | 356 (73) | 0.97 (0.90 to 1.00) |
0.39 (0.33 to 0.45) |
‐ | ‐ | ‐ | ‐ |
| ABCDE at > 3.05 | 1 | 356 (73) | 0.95 (0.87 to 0.98) |
0.57 (0.51 to 0.62) |
‐ | ‐ | ‐ | ‐ |
| ABCDE at > 3.6 | 1 | 356 (73) | 0.90 (0.81 to 0.96) |
0.70 (0.64 to 0.75) |
‐ | ‐ | ‐ | ‐ |
| ABCDE at > 4.25 | 1 | 356 (73) | 0.82 (0.71 to 0.90) |
0.82 (0.77 to 0.86) |
‐ | ‐ | ‐ | ‐ |
| ABCDE at > 4.9 | 1 | 356 (73) | 0.74 (0.62 to 0.84) |
0.90 (0.86 to 0.93) |
‐ | ‐ | ‐ | ‐ |
| ABCDE at ≥ 4 | ‐ | ‐ | ‐ | ‐ | 1 | 269 (84) | 0.90 (0.82 to 0.96) |
0.87 (0.81 to 0.92) |
| 7FFM at ≥ 2 | 1 | 401 (60) | 0.80 (0.68 to 0.89) |
0.89 (0.85 to 0.92) |
4 | 2200 (340) | 0.89 (0.76 to 0.96) |
0.84 (0.78 to 0.89) |
| 7PCL at ≥ 2 | 1 | 638 (108) | 0.93 (0.86 to 0.97) |
0.98 (0.97 to 0.99) |
‐ | ‐ | ‐ | ‐ |
| 7PCL at ≥ 3 | 2 | 11137 (127) | 0.67 (0.46 to 0.83) |
0.96 (0.88 to 0.99) |
11 | 3408 (798) | 0.80 (0.63 to 0.91) |
0.67 (0.51 to 0.80) |
| 7PCL at ≥ 5 | ‐ | ‐ | ‐ | ‐ | 1 | 322 (70) | 0.67 (0.55 to 0.78) |
0.83 (0.78 to 0.87) |
| 7PCL at NR | ‐ | ‐ | ‐ | ‐ | 4 | 1936 (360) | 0.72 (0.56 to 0.84) |
0.79 (0.61 to 0.90) |
| Revised 7PCL at NR (likely ≥ 1) | ‐ | ‐ | ‐ | ‐ | 1 | 1678 (238) | 0.61 (0.54 to 0.67) |
0.88 (0.86 to 0.89) |
| Revised 7PCL at ≥ 1 | ‐ | ‐ | ‐ | ‐ | 1 | 300 (100) | 0.88 (0.80 to 0.94) |
0.51 (0.44 to 0.58) |
| Revised 7PCL for FU: major change | ‐ | ‐ | ‐ | ‐ | 1 | 70 (12) | 0.67 (0.35 to 0.90) |
0.60 (0.47 to 0.73) |
| Menzies at 2 negative and ≥ 1 positive | 1 | 206 (23) | 0.83 95) |
0.69 (0.62 to 0.75) |
4 | 1856 (317) | 0.78 (0.38 to 0.96) |
0.63 (0.39 to 0.81) |
| Menzies at NR | ‐ | ‐ | ‐ | ‐ | 2 | 60 (26) | 0.77 (0.57 to 0.89) |
0.82 (0.66 to 0.92) |
| 3PCL at ≥ 2 | ‐ | ‐ | ‐ | ‐ | 7 | 1505 (363) | 0.74 (0.61 to 0.85) |
0.60 (0.42 to 0.76) |
| 4‐point (scored 3PCL) at > 2 | ‐ | ‐ | ‐ | ‐ | 1 | 75 (32) | 0.84 (0.67 to 0.95) |
0.81 (0.67 to 0.92) |
| Hofman algorithm at NR | ‐ | ‐ | ‐ | ‐ | 1 | 254 (75) | 0.87 (0.77 to 0.93) |
0.88 (0.82 to 0.92) |
| CASH at ≥ 6 | ‐ | ‐ | ‐ | ‐ | 1 | 477 (119) | 0.78 (0.70 to 0.85) |
0.51 (0.46 to 0.56) |
| CASH at ≥ 8 | ‐ | ‐ | ‐ | ‐ | 2 | 190 (56) | 0.97 (0.79 to 1.00) |
0.69 (0.60 to 0.76) |
| Chaos/Clues at = 2 | ‐ | ‐ | ‐ | ‐ | 2 | 940 (148) | 0.82 (0.75 to 0.87) |
0.53 (0.36 to 0.70) |
| Acral 3‐step | ‐ | ‐ | ‐ | ‐ | 1 | 107 (25) | 0.96 (0.80 to 1.00) |
0.91 (0.83 to 0.96) |
| b. Invasive melanoma | In‐person | Image‐based | ||||||
| No algorithm: threshold NR | 3 | 190 (62) | 0.87 (0.76 to 0.93) |
0.96 (0.91 to 0.98) |
6 | 683 (202) | 0.77 (0.59 to 0.88) |
0.79 (0.63 to 0.90) |
| Pattern analysis: threshold NR | 1 | 45 (16) | 0.81 (0.54 to 0.96) | 0.97 (0.82 to 1.00) |
1 | 119 (24) | 1.00 (0.86 to 1.00) |
0.97 (0.91 to 0.99) |
| ABCD at > 4.2 | 1 | 495 (23) | 0.88 (0.69 to 0.97) |
0.64 (0.60 to 0.68) |
‐ | ‐ | ‐ | ‐ |
| ABCD at > 4.75 | ‐ | ‐ | ‐ | ‐ | 2 | 330 (85) | 0.76 (0.66 to 0.84) |
0.84 (0.73 to 0.91) |
| ABCD at > 5.45 | 2 | 832 (242) | 0.79 (0.74 to 0.84) |
0.90 (0.58 to 0.98) |
1 | 258 (64) | 0.45 (0.33 to 0.58) |
0.94 (0.89 to 0.97) |
| 7PCL at NR | ‐ | ‐ | ‐ | ‐ | 1 | 332 (217) | 0.90 (0.85 to 0.94) |
0.79 (0.71 to 0.86) |
| Menzies at 2 negative and ≥ 1 positive | ‐ | ‐ | ‐ | ‐ | 4 | 4184 (715) | 0.91 (0.83 to 0.96) |
0.71 (0.68 to 0.74) |
| 3PCL at > NR | ‐ | ‐ | ‐ | ‐ | 1 | 332 (217) | 0.82 (0.77 to 0.87) |
0.40 (0.31 to 0.50) |
| Kenet (modified) at MM likely | 1 | 54 (12) | 1.00 (0.74 to 1.00) |
0.95 (0.84 to 0.99) |
1 | 258 (64) | 0.75 (0.63 to 0.85) |
0.94 (0.89 to 0.97) |
| Kenet (modified) at MM possible | 1 | 54 (12) | 1.00 (0.74 to 1.00) |
0.45 (0.30 to 0.61) |
1 | 258 (64) | 0.89 (0.79 to 0.95) |
0.87 (0.82 to 0.91) |
| CASH at ≥ 8 | ‐ | ‐ | ‐ | ‐ | 1 | 332 (217) | 0.82 (0.76 to 0.86) |
0.72 (0.63 to 0.80) |
| Kreusch algorithm | ‐ | ‐ | ‐ | ‐ | 1 | 265 (96) | 0.98 (0.93 to 1.00) |
0.83 (0.77 to 0.89) |
| Menzies for amelanotic at 1 | ‐ | ‐ | ‐ | ‐ | 1 | 332 (217) | 0.91 (0.87 to 0.95) | 0.70 (0.61 to 0.79) |
| Menzies for amelanotic at 0 | ‐ | ‐ | ‐ | ‐ | 1 | 332 (217) | 1.00 (0.98 to 1.00) | 0.52 (0.43 to 0.62) |
| c. Any skin cancer or lesion with high risk of progression to melanoma | In‐person | Image‐based | ||||||
| No algorithm at NR | 1 | 231 (77) | 0.90 (0.81 to 0.95) |
0.94 (0.89 to 0.97) |
2 | 83 (32) | 0.78 (0.61 to 0.89) |
0.75 (0.61 to 0.85) |
| Pattern analysis: threshold NR | 1 | 3372 (98) | 0.90 (0.82 to 0.95) |
1.00 (0.99 to 1.00) |
1 | 119 (37) | 1.00 (0.91 to 1.00) |
0.96 (0.90 to 0.99) |
| ABCD at > 5.45 | 1 | 200 (46) | 0.98 (0.88 to 1.00) |
0.98 (0.94 to 1.00) |
‐ | ‐ | ‐ | ‐ |
| 3PCL at ≥ 2 | 1 | 77 (39) | 0.85 (0.69 to 0.94) |
0.26 (0.13 to 0.43) |
1 | 150 (44) | 0.91 (0.78 to 0.97) |
0.72 (0.62 to 0.80) |
| 3PCL: three‐point checklist; 7FFM: seven features for melanoma; 7PCL: seven‐point checklist; ABCD(E): asymmetry, border, colour, differential structures (enlargement);CASH: colour, architecture, symmetry and homogeneity; CI: confidence interval; FU: follow‐up; MM: malignant melanoma; NR: not reported | ||||||||
aAll analyses by algorithm were undertaken using the bivariate normal model (BVN).