|
Home | Issues | eMJA shop | My account | Classifieds | Contact | More... | Topics | Search |
Justine Mayer
Abstract - Introduction - Methods - Data sources - Study selection and data extraction - Results - Design and validity of selected studies - Results of selected studies - Discussion - Acknowledgements - References - Authors' details
Make a comment - Register to be notified of new articles by e-mail - Current contents list - ©MJA1997
Abstract |
Objective: To assess the evidence that dermatoscopy
improves the accuracy of diagnosis of melanomas in clinical
practice. Data sources: MEDLINE 1983-January 1997, EMBASE 1980-1996, and bibliographies of retrieved articles. Study selection and data extraction: Studies selected were original studies with formal methods and results sections comparing diagnostic accuracy of dermatoscopy for malignant melanoma with another clinical method; the criterion standard was excision biopsy with histopathological examination; and accuracy of dermatoscopic diagnosis was determined over a spectrum of stages of melanoma and skin lesions commonly confused with melanoma. Data were extracted by a single observer. Data synthesis: 579 articles were identified; six studies met the inclusion criteria. Positive likelihood ratios for dermatoscopy for diagnosis of melanoma ranged from 2.9 to 10.3. Dermatoscopy had 10%-27% higher sensitivity than clinical diagnosis in the two studies with the most clinically equivocal lesions. However, when sensitivity of clinical diagnosis was more than 84%, sensitivity of dermatoscopy was only slightly higher. One study of dermatologists with no training in dermatoscopy showed a significant decrease in sensitivity. Conclusions: Variability between studies in methods, observers and types of pigmented skin lesions and lack of studies in primary care make generalisation of results difficult. Dermatoscopy appeared not to improve the accuracy of diagnosis enough to alter the clinical management of most pigmented skin lesions. Further research with more explicit methods is needed. |
Introduction |
Malignant melanoma is an important cause of morbidity and mortality
in Australia. Public health strategies emphasise, in addition to
reducing sun exposure, early diagnosis and treatment. It has been
claimed that dermatoscopy improves the accuracy of diagnosis of
malignant melanomas,1,2 but
calls have also been made for a systematic review of the evidence3 (see Box 1 for description of
dermatoscopy).
An ideal diagnostic test achieves maximum levels of sensitivity and specificity. However, as sensitivity reflects a test's ability to identify individuals with the disease, most clinicians considering the diagnosis of melanoma would prefer a test with high sensitivity. Furthermore, diagnostic tests should be performed only if they have the potential to alter clinical management. For pigmented skin lesions, this means helping decide whether to excise the lesion. This review aimed to identify primary studies that researched the diagnostic accuracy of dermatoscopy compared with clinical diagnosis and to assess their implications for clinical practice. |
Methods | The review followed current recommendations on systematic reviews of the accuracy of diagnostic tests.7,8 |
Data sources | MEDLINE Express (Silverplatter Winspirs) on CD-ROM databases January 1983-January 1997 and Elsevier Science EMBASE Database 1980-1996 were searched. The Medline search was conducted with exploded MeSH headings "melanoma" or "skin neoplasms", combined with "sensitivity and specificity", and with text words "dermoscop*", "dermatoscop*", "skin surface microscop*", "epiluminesc*", or "incident light microscop*". Similar strategies were used to search EMBASE. |
Study selection and data extraction |
Abstracts of articles were reviewed by the author for studies that met
the selection criteria:
Articles describing studies that appeared to meet the criteria were retrieved for more detailed review (after translation if not in English). Bibliographies of articles retrieved were also checked for articles not listed or incorrectly indexed in MEDLINE or EMBASE. Articles that did not meet the selection criteria included those on the terminology of dermatoscopic techniques, development of criteria for diagnosing pigmented skin lesions, observer error of diagnosis, accuracy of diagnosis of a single type of pigmented skin lesion, and accuracy of digital imaging computer programs. Studies providing data on the sensitivity and specificity of dermatoscopy but with no comparison group were also excluded.9,10 Articles that met the selection criteria were reviewed for study validity and applicability to clinical practice.11 The methods section of each article was checked for explicit mention of the criterion standard of histopathological examination, blinding of observers, spectrum of pigmented skin lesions, study setting, patient demographics, prevalence of melanoma, sample size, intraobserver and interobserver error, and the proportion of pigmented skin lesions in which no dermatoscopic diagnosis could be made. |
Results | Three hundred and thirty-six articles were identified from MEDLINE and 243 from EMBASE. Six articles met the inclusion criteria.4,12-16 All were indexed in MEDLINE and one in EMBASE and all were in English. No further articles meeting the inclusion criteria were found on review of the bibliographies of retrieved articles. |
Design and validity of selected studies |
The designs of the six studies are summarised in Box 2. Assessment of
the studies' internal and external validity was hampered by lack of
information. Only Binder et al.14
provided sufficient details of methods to enable the study to
be repeated.
All studies were set in specialist dermatology clinics, but none provided further details of the setting (i.e., whether a secondary or tertiary referral centre). They investigated selected groups of pigmented skin lesions, with melanomas comprising 15%-41%, but only one study explicitly stated how lesions were chosen for entry into the study.12 All appeared to include only lesions to be excised because of presumed risk of malignancy or, possibly, patient request. None commented on the suitability of the sample to detect clinically important differences in the diagnostic methods being evaluated, and none provided data on all presenting lesions or followed them up to calculate the true false negative rate. Although all studies used the criterion standard of histopathological examination, none commented on how the histological diagnosis was reached, although pathologists may vary in classifying pigmented skin lesions.24,.25 Only four studies provided objective data on melanoma diameter or Breslow depth.4,12,14,15 Five studies assessed dermatoscopy results in vivo and one, from photographic slides.14 Observers in the in vivo studies were assumed to be blinded to the histopathological results, but it was not apparent that clinical and dermatoscopic diagnoses were independent, despite the importance of history in the diagnosis of malignant lesions. Five studies did not provide the criteria used for clinical diagnosis, and only Binder et al. provided complete details of the method of non-dermatoscopic examination. The diagnostic criteria and method of surface microscopy used differed between the studies. Three studies used handheld monocular dermatoscopes with x 10 magnification,4,13,16 and two studies used binocular stereomicroscopes with magnification up to x 40.12,14 One study used both types of instrument, but did not clearly distinguish their results.15 No study remarked that a skin lesion was unable to be diagnosed by dermatoscopy. |
Results of selected studies |
Results of the six studies are shown in Box 2. Five of the studies
compared dermatoscopy with clinical diagnosis or diagnosis by
microscopy without oil. Likelihood ratios for a positive diagnosis
of melanoma by dermatoscopy ranged from 2.9 to 10.3 (meaning that
pigmented lesions diagnosed as melanoma by dermatoscopy were 2.9 to
10.3 times morelikely to be melanoma than not). Two studies found that
dermatoscopy had higher sensitivity than non-dermatoscopic
diagnosis,12,13 one found
no difference,15 and one
found that dermatoscopy increased sensitivity when performed by
experts, but decreased it when performed by dermatologists without
training in the technique.14
Pazzini et al. provided insufficient data to calculate
sensitivity.16
Nachbar et al.4 compared two forms of dermatoscopy -- with explicit structured diagnostic criteria and without these criteria, but combined with clinical diagnosis. They found that dermatoscopy with structured ABCD criteria19,21 (specific for dermatoscopy and differing from the ABCDE criteria for clinical diagnosis) had higher sensitivity and specificity than dermatoscopy without such criteria. Clinical diagnosis varied widely in sensitivity and specificity between the studies, most likely because the lesions varied in their ease of diagnosis. Observers may also have varied in their ability to make the diagnosis. This variability prevented formal meta-analysis and estimation of a single summary statistic for the diagnostic benefit of dermatoscopy. The studies that provided results for handheld monocular dermatoscopes with x 10 magnification are most relevant to clinical practice.4,13 Cristofolini et al.13 found that dermatoscopy with pattern-analysis criteria had slightly higher sensitivity (88%) and specificity (79%) than clinical diagnosis with ABCDE criteria.18 However, the latter was already highly accurate (sensitivity, 85%; specificity, 75%). Nachbar et al. found that dermatoscopy with structured criteria had higher sensitivity (93% versus 84%) and specificity (91% versus 84%) than without such criteria.4 Interestingly, the studies that showed the greatest improvement in sensitivity for dermatoscopy were those that had the lowest baseline sensitivity for clinical diagnosis. Steiner et al.12 found that dermatoscopy increased sensitivity from 59% to 86% (an increase of 27%) compared to microscopy without oil. Binder et al.14 found dermatoscopy performed by experts increased sensitivity from 58% to 68% (increase of 10%) compared to clinical diagnosis. Both these studies appeared to include small difficult-to-diagnose lesions, suggesting that dermatoscopy may be most useful in these circumstances. In contrast, when clinical diagnosis was very accurate (as found by Soyer et al.15) dermatoscopy was of no benefit. This study also provided data on the accuracy of individual dermatoscopy criteria (not shown). The study by Binder et al. also examined two further important issues. They found that, although dermatoscopy had higher sensitivity than clinical diagnosis when performed by experts, it had lower sensitivity when performed by dermatologists with no formal training in the technique; these observers misdiagnosed more melanomas with dermatoscopy. Further, while experts showed moderate interobserver and intraobserver agreement (average k = 0.47 and 0.56, respectively), non-experts showed only fair agreement (average k = 0.29 and 0.36, respectively). This study used photographic slides and it may be problematic generalising these results to clinical practice. |
Discussion |
Despite the comprehensive search strategy used for this review,
publication bias is still possible, and small studies of
dermatoscopy with negative results may not have been submitted or
accepted for publication.
For dermatologists working within specialist clinics the evidence on dermatoscopy is:
However, as none of the studies were conducted in a primary care setting, the place of dermatoscopy in general practice is unknown. In addition, none of the studies commented that dermatoscopy improved the sensitivity and specificity of diagnosis enough to alter the clinical management of the pigmented skin lesion. In fact, dermatoscopy only increased the accuracy of diagnosis of equivocal lesions that were to undergo biopsy anyway. Given the low threshold clinicians already have for excision biopsy, the modest impact of dermatoscopy on likelihood ratios suggests it does not improve diagnostic accuracy enough to alter clinical management of most pigmented skin lesions. Further research is currently being undertaken using dermatoscopy to describe the sensitivities and specificities of individual features of pigmented lesions.6 This may prove to be more clinically relevant in the long term. In one study, the use of dermatoscopy by untrained dermatologists resulted in more melanomas being missed than with magnification alone. Non-experts also showed more interobserver and intraobserver variation. This suggests that the benefits of dermatoscopy depend on training and experience with its use. However, none of the studies commented on the type of training needed to develop competence. Two of the studies investigated the effect of using defined diagnostic criteria in diagnosing melanoma. Their results were consistent with the view that simple structured explicit criteria, such as ABCD criteria, either with or without dermatoscopy, result in a high sensitivity and specificity for diagnosis of melanoma. To provide better evidence on the value of dermatoscopy in Australian clinical practice, future research should be more explicit in the methods used and should select lesions representative of those seen in primary and secondary care in Australia. An investigation in clinical practice, where the prevalence of melanoma is low, is not practicable. However, it would be possible to compare the accuracy of diagnosis from dermatoscopic and standard magnified images of skin lesions by dermatologists and general practitioners. This type of study would also allow further assessment of intraobserver and interobserver variability and could be combined with assessment of the impact of training. This approach would allow the potential clinical benefits and limitations of dermatoscopy to be more clearly understood. |
Acknowledgements | I gratefully thank Dr P S Morris (Menzies School of Health Research, Darwin, NT), Associate Professor L Piterman (Department of Community Medicine, Monash University, Melbourne, VIC), and three anonymous reviewers for helpful comments on an earlier draft of the manuscript; Mariella Stroschio and Antje Haase for translation of Italian and German articles; the RACGP Resource Centre librarians, Jane Ryan and Diane Horrigan, for their skills in searching Embase and hard work in locating the articles; the Royal Darwin Hospital library for use of Medline ; and the RACGP Research and Development Fund. |
References |
(Received 14 Nov 1996, accepted 16 May 1997) |
|
Home | Issues | eMJA shop | My account | Classifieds | More... | Contact | Topics | Search |
©MJA 1997
<URL: http://www.mja.com.au/>
© 1997 Medical Journal of Australia.
We appreciate
your comments.