Summary of findings'. 'Summary of findings table.

Question:	What is the diagnostic accuracy of dermoscopy, in comparison to visual inspection, for the detection of keratinocyte skin cancer in adults?
Population:	Adults with skin lesions: suspicious for keratinocyte skin cancers, basal cell carcinoma (BCC) or cutaneous squamous cell carcinoma (cSCC) (e.g. non‐pigmented lesions); suspicious for any skin cancer, including melanoma (e.g. those with pigmented lesions only or mixed populations of pigmented and non‐pigmented lesions); or those at high risk of developing keratinocyte skin cancer
Index test:	Dermoscopy with or without the use of any established algorithms or checklist to aid diagnosis, including: in‐person evaluations (face‐to‐face diagnosis), and image‐based evaluations (diagnosis based on assessment of a dermoscopic image)
Comparator test	Visual inspection including: in‐person evaluations, and image‐based evaluations (diagnosis based on assessment of a clinical image)
Primary Target condition:	BCC or cSCC
Reference standard:	Histology with or without long‐term follow‐up
Action:	If accurate, negative results will stop patients having unnecessary excision or biopsy of skin lesions; positive results could inform the use of nonsurgical management options
	Number of studies		Total lesions			Total malignancies
Quantity of evidence	24		Visual Inspection: 8805 Dermoscopy: 6855			Visual Inspection: 2579 Dermoscopy: 1444
Limitations
Risk of bias: (in‐person (14); image‐based (12))	Potential risk of bias for participant selection from use of case‐control type design (3 image‐based), inappropriate exclusion criteria (3; 2) or lack of detail (8; 4). All visual inspection and dermoscopy interpretation considered blinded to reference standard diagnosis. Visual Inspection risk of bias not clear due to thresholds not clearly prespecified (8; 4). Threshold prespecification better reported for dermoscopy (6; 6). Low risk for reference standard (13; 11); high risk from use of expert diagnosis or > 20% of benign lesions with no histology (1; 1). High risk for participant flow due to differential verification (1; 1), and exclusions following recruitment (5; 6); timing of tests was not mentioned in (7; 7)
Applicability of evidence to question: (in‐person (14); image‐based (12))	High concern for participants (14; 12) due to restriction to those with histopathology results (13; 11) and including multiple lesions per participant (9; 2). High concern for Visual Inspection (7; 4) from lack of description of diagnostic thresholds. High concern for dermoscopy (3; 9) from no description of diagnostic thresholds (2; 4) or reporting of average or consensus diagnoses (2; 7). Dermoscopic image interpretation blinded to clinical images (10 image‐based). Unclear applicability of reference standard due to insufficient information concerning the expertise of the histopathologist (13; 11)
FINDINGS:
We included 24 studies. 14 studies reported data for in‐person visual inspection (n = 11) or in‐person dermoscopy (n = 8); 12 studies reported data for image‐based visual inspection (n = 4) or image‐based dermoscopy (n = 10). Two studies report both in‐person and image‐based data. The findings presented are based on results for the 21 studies reporting data for BCC alone or for cSCC alone. Due to the observed heterogeneity between studies, the results presented are points estimated from summary ROC curves rather than average sensitivity and specificity operating points. These are presented for illustrative purposes and should not be quoted as the actual performance of visual inspection or dermoscopy. We did not undertake analyses of studies by degree of prior testing due to a lack of relevant information provided in the study publications, most studies apparently being conducted in referred populations, and small study subgroups. There was not enough evidence to assess the use of algorithms or structured checklists for dermoscopy (or visual inspection)
Test (for BCC):	In‐person visual inspection alone versus visual inspection plus dermoscopy for the detection of BCC – any algorithm or threshold
Data analysed	Visual inspection				8 datasets ‐ 7017 lesions; 1586 cases
	Dermoscopy				7 datasets ‐ 4683 lesions; 363 cases
Resultsa	Sensitivity		Fixed specificity		Fixed sensitivity		Specificity
Visual inspection	79%		80%		80%		77%
Dermoscopy	93%						99%
Numbers applied to a hypothetical cohort of 1000 lesionsb
	TP	FN	FP	TN	TP	FN	FP	TN
At a prevalence of 10%	VI: 79 D: 93 ↑ 14	VI: 21 D: 7 ↓ 14	180	720	80	20	VI: 207 D: 9 ↓198	VI: 693 D: 891 ↑198
At a prevalence of 17%	VI: 134 D: 158 ↑24	VI: 36 D: 12 ↓ 24	166	664	136	34	VI: 191 D: 8 ↓183	VI: 639 D: 822 ↑183
At a prevalence of 53%	VI: 419 D: 493 ↑ 74	VI: 111 D: 37 ↓ 74	94	376	424	106	VI: 108 D: 5 ↓103	VI: 362 D: 465 ↑103
Consistency:	Wide range in prevalence of BCC; includes pigmented and non‐pigmented lesion populations and participants suspected of BCC or suspected of any malignancy, including melanoma. Sensitivities highly heterogeneous, particularly for visual‐inspection evaluations. Specificity for BCC lower in studies of non‐pigmented lesions
Test (for BCC):	Image‐based visual inspection alone versus visual inspection plus dermoscopy for the detection of BCC – any algorithm or threshold
Data analysed	Visual inspection				4 datasets ‐ 853 lesions; 156 cases
	Dermoscopy				9 datasets ‐ 2271 lesions; 737 cases
Results	Sensitivity		Fixed specificity		Fixed sensitivity		Specificity
Visual inspection	85%		80%		80%		87%
Dermoscopy	93%						96%
Numbers applied to a hypothetical cohort of 1000 lesionsc
	TP	FN	FP	TN	TP	FN	FP	TN
At a prevalence of 11%	VI: 94 D: 102 ↑ 8	VI: 16 D: 8 ↓ 8	178	712	88	22	VI: 116 D: 36 ↓80	VI: 774 D: 854 ↑80
At a prevalence of 16%	VI: 136 D: 149 ↑13	VI: 24 D: 11 ↓ 13	168	672	128	32	VI: 109 D: 34 ↓75	VI: 731 D: 806 ↑75
At a prevalence of 47%	VI: 400 D: 437 ↑ 37	VI: 70 D: 33 ↓ 37	106	424	376	94	VI: 69 D: 21 ↓48	VI: 461 D: 509 ↑48
Consistency:	Wide range in prevalence of BCC; includes mixed populations, as for in‐person evaluations. Sensitivities highly heterogeneous for visual inspection evaluations.
Test (for cSCC):	Visual inspection or dermoscopy for the detection of cSCC
	Datasets		Lesions	Cases	Sensitivity	(95%CIs)	Specificity	(95%CI)
Visual inspection (in‐person)	2		2684	538	57%	(53%, 61%)	79%	(77%, 81%)
Dermoscopy (image‐based)	2		717	119	55%	(29%, 79%)	84%	(32%, 98%)

^aNumbers for a hypothetical cohort of 1000 lesions are presented for two illustrative examples of points on the SROC curves: firstly for the sensitivities of tests at fixed specificities of 80%; and secondly for the specificities of tests at fixed sensitivities of 80%.
 ^bNumbers estimated at 25th, 50th (median) and 75% percentiles of BCC prevalence observed across 11 studies reporting in‐person evaluations of visual inspection (reported in eight studies) or visual inspection plus dermoscopy (reported in seven studies).
 ^cNumbers estimated at 25th, 50th (median) and 75% percentiles of BCC prevalence observed across 11 studies reporting image‐based diagnosis using clinical photographs (reported in four studies) or dermoscopic images (reported in nine studies)