Summary of findings'. 'What is the diagnostic accuracy of visual inspection for the detection of cutaneous invasive melanoma and atypical intraepidermal melanocytic variants in adults?
| Question | What is the diagnostic accuracy of visual inspection for the detection of cutaneous invasive melanoma and atypical intraepidermalmelanocytic variants in adults? | |||||
| Population | Adults with lesions suspicious for melanoma, including:
|
|||||
| Index test | Visual inspection with or without the use of any established algorithms or checklist to aid diagnosis, including:
|
|||||
| Target condition | Cutaneous invasive melanoma and atypical intraepidermal melanocytic variants | |||||
| Reference standard | Histology with or without long‐term follow‐up | |||||
| Action | If accurate, positive results ensure melanoma lesions are not missed but are appropriately referred and excised and those with negative results can be safely reassured and discharged. | |||||
| Number of studies | Total lesions | Total cases | ||||
| Quantity of evidence | 49a | 34,351 | 2499 | |||
| Limitations | ||||||
| Risk of bias | Potential risk for participant selection from case‐control design (6), inappropriate exclusion criteria (7) or lack of detail (27/49) All index test interpretation was blinded to reference standard diagnosis. Index test thresholds not clearly pre‐specified (22/33 in‐person evaluations; 13/16 image‐based) Low risk for reference standard (42/49); high concern from use of expert diagnosis (6). Blinding of reference standard to visual inspection diagnosis not reported in any study. High risk for participant flow due to differential verification (11), and exclusions following recruitment (15); 37 studies did not mention timing of tests |
|||||
| Applicability of evidence to question | Participant selection restricted to those with melanocytic lesions only (10), or to those with histopathology results (37) and included multiple lesions per participant (14) No description of index test diagnostic thresholds (24 in‐person; 13 image‐based) or reporting of average or consensus diagnoses (7 in‐person; 13 image‐based). Clinical images interpreted blinded to clinical information (11/16). Little information given concerning the expertise of the histopathologist (40/49). |
|||||
| Findings | ||||||
| 37 studies (providing 39 datasets) reported accuracy data for the primary target condition. We separated them a priori into in‐person (n = 28) and image‐based (n = 11) evaluations. Subsequent analysis confirmed differences in accuracy according to the different approaches to diagnosis (P < 0.001). Attempts to analyse studies by degree of prior testing were hampered by a lack of relevant information provided in the study publications and by the inclusion of lesions selected for biopsy or excision. Of the 28 in‐person evaluations, we could only clearly place 17 on the clinical pathway, and considered 11 to have provided insufficient information to allow us to identify the pathway (coded ‘unclear’ on pathway). The findings presented are based on results for in‐person evaluations that could be clearly placed on the clinical pathway. | ||||||
| Test: | In‐person visual inspection using any or no algorithm at any threshold | |||||
| Data: | Number of datasets | Total lesions | Total melanomas | |||
| All in‐person evaluations | 28 | 25,604 | 1748 | |||
| Studies clearly placed on the clinical pathway | 17 | 14,700 | 622 | |||
| Place on pathway: participants with limited prior testing (all lesions) | ||||||
| Datasets (n) | Lesions (n) | Melanomas (n) | Sensitivity (95% CI) | Specificity (95% CI) | ||
| 3 | 1339 | 55 | 92% (26 to 100) | 80% (74 to 85) | ||
| Numbers in a cohort of 1000 lesionsb | TP | FP | FN | TN | PPV | NPV |
| At a prevalence of 4% | 37 (10 to 40) |
195 (252 to 147) |
3 (30 to 0) |
765 (708 to 813) |
16% (4 to 21) | 100% (96 to 100) |
| At a prevalence of 9% | 83 (24 to 90) |
185 (239 to 139) |
7 (66 to 0) |
725 (671 to 771) |
31% (9 to 39) |
99% (91 to 100) |
| At a prevalence of 16% | 148 (42 to 160) |
171 (221 to 129) |
12 (118 to 0) |
669 (619 to 711) |
46% (16 to 55) |
98% (84 to 100) |
| Place on pathway: participants with limited prior testing (only lesions selected for excision) | ||||||
| Datasets (n) | Lesions (n) | Melanomas (n) | Sensitivity (95% CI) | Specificity (95% CI) | ||
| 2 | 4228 | 160 | 90% (70 to 97) | 81% (67 to 90) | ||
| Numbers in a cohort of 1000 lesionsb | TP | FP | FN | TN | PPV | NPV |
| At a prevalence of 4% | 36 (28 to 39) |
180 (312 to 96) |
4 (12 to 1) |
780 (648 to 864) |
17% (8 to 29) |
99% (98 to 100) |
| At a prevalence of 9% | 81 (63 to 88) |
170 (296 to 91) |
9 (27 to 2) |
740 (614 to 819) |
32% (18 to 49) |
99% (96 to 100) |
| At a prevalence of 16% | 144 (112 to 156) |
157 (273 to 84) |
16 (48 to 4) |
683 (567 to 756) |
48% (29 to 65) |
98% (92 to 99) |
| Place on pathway: referred participants (all lesions) | ||||||
| Datasets (n) | Lesions (n) | Melanomas (n) | Sensitivity (95% CI) | Specificity (95% CI) | ||
| 2 | 3494 | 61 | 75% (49 to 90) | 99% (95 to 100) | ||
| Numbers in a cohort of 1000 lesionsb | TP | FP | FN | TN | PPV | NPV |
| At a prevalence of 4% | 30 (20 to 36) |
13 (51 to 4) |
10 (20 to 4) |
947 (909 to 956) |
69% (28 to 90) |
99% (98 to 100) |
| At a prevalence of 9% | 67 (44 to 81) |
13 (48 to 4) |
23 (46 to 9) |
897 (862 to 906) |
84% (48 to 96) |
98% (95 to 99) |
| At a prevalence of 16% | 119 (78 to 144) |
12 (45 to 3) |
41 (82 to 16) |
828 (795 to 837) |
91% (64 to 98) |
95% (91 to 98) |
| Referred participants (only lesions selected for excision) | ||||||
| Datasets (n) | Lesions (n) | Melanomas (n) | Sensitivity (95% CI) | Specificity (95% CI) | ||
| 8 | 5331 | 258 | 77% (62 to 87) | 96% (90 to 98) | ||
| Numbers in a cohort of 1000 lesionsb | TP | FP | FN | TN | PPV | NPV |
| At a prevalence of 4% | 31 (25 to 35) |
41 (99 to 16) |
9 (15 to 5) |
919 (861 to 944) |
43% (20 to 68) |
99% (98 to 99) |
| At a prevalence of 9% | 69 (56 to 78) |
39 (94 to 15) |
21 (34 to 12) |
871 (816 to 895) |
64% (37 to 84) |
98% (96 to 99) |
| At a prevalence of 16% | 123 (99 to 139) |
36 (87 to 14) |
37 (61 to 21) |
804 (753 to 826) |
77% (53 to 91) |
96% (92 to 98) |
| Referred participants with equivocal lesions (only lesions selected for excision) | ||||||
| Datasets (n) | Lesions (n) | Melanomas (n) | Sensitivity (95% CI) | Specificity (95% CI) | ||
| 2 | 930 | 88 | 85% (56 to 96) | 89% (79 to 95) | ||
| Numbers in a cohort of 1000 lesionsb | TP | FP | FN | TN | PPV | NPV |
| At a prevalence of 4% | 34 (22 to 38) |
101 (197 to 48) |
6 (18 to 2) |
859 (763 to 912) |
25% (10 to 44) |
99% (98 to 100) |
| At a prevalence of 9% | 76 (50 to 86) |
96 (187 to 46) |
14 (40 to 4) |
814 (723 to 865) |
44% (21 to 66) |
98% (95 to 100) |
| At a prevalence of 16% | 136 (89 to 154) |
88 (172 to 42) |
24 (71 to 6) |
752 (668 to 798) |
61% (34 to 79) |
97% (90 to 99) |
| CI: confidence interval; FN: false‐negative; FP: false‐positive; NPV: negative predictive value; PPV: positive predictive value; TN: true negative; TP: true positive | ||||||
a37 of the 49 included studies (reporting on 39 cohorts of lesions) provide data for the primary target condition (defined as detection of cutaneous invasive melanoma and atypical intraepidermal melanocytic variants) and are the main focus of this 'Summary of findings' table; the summary of methodological quality is based on the full sample of 49 studies. bWe estimated number of true positives (TP), false‐positives (FP), false‐negatives (FN) and true negatives (TN) for a hypothetical cohort of 1000 lesions at the median and interquartile ranges of prevalence (25th and 75th percentiles), at average sensitivity and specificity and using the lower and upper limits of the 95% confidence intervals, denoted in brackets (lower limit to upper limit).