< CNMD \ Results
Dermoscopy of pigmented skin lesions: Results of a consensus meeting via the Internet|
Giuseppe Argenziano, MD,a H. Peter Soyer, MD,b Sergio Chimenti, MD,c Renato Talamini, ScD,d Rosamaria Corona, MD, DSc,e Francesco Sera, DStat,e Michael Binder, MD,f Lorenzo Cerroni, MD,b Gaetano De Rosa, MD,g Gerardo Ferrara, MD,h Rainer Hofmann-Wellenhof, MD,b Michael Landthaler, MD,i Scott W. Menzies, MBBS, PhD,j Hubert Pehamberger, MD,f Domenico Piccolo, MD,k Harold S. Rabinovitz, MD,l Roman Schiffner, MD,i Stefania Staibano, MD,g Wilhelm Stolz, MD,i Igor Bartenjev, MD,m Andreas Blum, MD,n Ralph Braun, MD,o Horacio Cabo, MD,p Paolo Carli, MD,q Vincenzo De Giorgi, MD,q Matthew G. Fleming, MD,r James M. Grichnik, MD, PhD,s Caron M. Grin, MD,t Allan C. Halpern, MD,u Robert Johr, MD,v Brian Katz, MD,e Robert O. Kenet, MD, PhD,x Harald Kittler, MD,f Jürgen Kreusch, MD,y Josep Malvehy, MD,z Giampiero Mazzocchetti, MD,aa Margaret Oliviero, ARNP,l Fezal Özdemir, MD,bb Ketty Peris, MD,k Roberto Perotti, MD,cc Ana Perusquia, MD,dd Maria Antonietta Pizzichetta, MD,d Susana Puig, MD,z Babar Rao, MD,ee Pietro Rubegni, MD,cc Toshiaki Saida, MD,ff Massimiliano Scalvenzi, MD,gg Stefania Seidenari, MD,hh Ignazio Stanganelli, MD,ii Masaru Tanaka, MD,jj Karin Westerhoff, MD,kk Ingrid H. Wolf, MD,b Otto Braun-Falco, MD,ll Helmut Kerl, MD,b Takeji Nishikawa, MD,jj Klaus Wolff, MD,f and Alfred W. Kopf, MDmm
Naples, Rome, Aviano, Benevento, L’Aquila, Florence, Atessa, Siena, Modena, Milan, and Ravenna, Italy; Graz and Vienna, Austria; Regensburg, Tübingen, Lübeck, and Munich, Germany; Camperdown, Australia; Plantation and Miami, Florida; Ljubljana, Slovenia; Geneva, Switzerland; Buenos Aires, Argentina; Milwaukee, Wisconsin; Durham, North Carolina; Farmington, Connecticut; Barcelona, Spain; Izmir, Turkey; Mexico City, Mexico; New Brunswick, New Jersey; Matsumoto and Tokyo, Japan; Skövde, Sweden; and New York, New York
From the Departments of Dermatology at Second University of Naples,a University of Graz,b University Tor Vergata of Rome,c University of Vienna,f University of Regensburg,i University of L’Aquila,k University Clinical Center Ljubljana,m University of Tübingen,n University Hospital Geneva,o Hospital de Clínicas, University of Buenos Aires,p University of Florence,q Medical College of Wisconsin,r University of Connecticut School of Medicine,t University at Luebeck,y University of Barcelona,z Medical Faculty, Ege University,bb University of Siena,cc Hospital Español,dd Shinshu University School of Medicine,ff University Federico II,gg University of Modena,hh Keio University School of Medicine,jj Skövde Hospital,kk University of Munich,ll and New York University School of Medicine, New York University Medical Centermm; National Cancer Instituted; Istituto Dermopatico dell’Immacolatae; Institute of Pathology, University Federico IIg; Pathologic Anatomy Service, Gaetano Rummo General Hospitalh; Sydney Melanoma Unit, Sydney Cancer Centre, Melanoma and Skin Cancer Research Institute, University of Sydney at Royal Prince Alfred Hospitalj; Skin and Cancer Associates, Department of Dermatology, University of Miami School of Medicinel; Department of Medicine, Division of Dermatology, Duke University Medical Centers; Dermatology Service, Department of Medicine, Memorial Sloan Kettering Cancer Centeru; Pigmented Lesion Clinic, Department of Dermatology and Cutaneous Surgery, University of Miami, School of Medicinev; Department of Medicine, New York-Presbyterian Hospital and Weill Medical College of Cornell Universityx; Dermatology Service, Vittorio Emanuele Hospitalaa; Division of Dermatology, Robert Wood Johnson Medical School–University of Medicine and Dentistry of New Jerseyee; and Cutaneous Oncology Unit, Niguarda Ca’ Granda Hospital, and Center for Cancer Prevention, Ravenna Hospital.ii
All technical issues related to this Internet study supported by Edra Medical Publishing and New Media , Milan, Italy.
Conflict of interest: None identified.
A case-by-case summary of data from all colleagues taking part of this virtual study is presented in an atlas called Dermoscopy of Pigmented Skin Lesions: An Atlas Based on the Consensus Net Meeting on Dermoscopy 2000. Milan, Italy: Edra Medical Publishing and New Media; 2001.
Reprint requests: H. Peter Soyer, MD, Department of Dermatology, University of Graz, Auenbruggerplatz 8 - A8036 Graz, Austria. E-mail: email@example.com.
J Am Acad Dermatol 2003;48:679-93.
Copyright© 2003 by the American Academy of Dermatology, Inc. 0190-9622/2003/$30.00 + 0
Background: There is a need for better standardization of the dermoscopic terminology in assessing pigmented skin lesions.
Objective: The virtual Consensus Net Meeting on Dermoscopy was organized to investigate reproducibility and validity of the various features and diagnostic algorithms.
Methods: Dermoscopic images of 108 lesions were evaluated via the Internet by 40 experienced dermoscopists using a 2-step diagnostic procedure. The first-step algorithm distinguished melanocytic versus nonmelanocytic lesions. The second step in the diagnostic procedure used 4 algorithms (pattern analysis, ABCD rule, Menzies method, and 7-point checklist) to distinguish melanoma versus benign melanocytic
lesions. Values, log odds ratios, sensitivity, specificity, and positive likelihood ratios were estimated for all diagnostic algorithms and dermoscopic features.
Results: Interobserver agreement was fair to good for all diagnostic methods, but it was poor for the majority of dermoscopic criteria. Intraobserver agreement was good to excellent for all algorithms and features considered. Pattern analysis allowed the best diagnostic performance (positive likelihood ratio: 5.1), whereas alternative algorithms revealed comparable sensitivity but less specificity. Interobserver agreement on management decisions made by dermoscopy was fairly good (mean value: 0.53).
Conclusion: The virtual Consensus Net Meeting on Dermoscopy represents a valid tool for better standardization of the dermoscopic terminology and, moreover, opens up a new territory for diagnosing and managing pigmented skin lesions. (J Am Acad Dermatol 2003;48:679-93.)
In all, 51 experienced clinicians in the field of dermoscopy were invited to participate in the CNMD, and 40 actively participated in this study. Geographic distribution of the participating board members is as follows: 24 participants from Europe (9 countries); 11 from the United States; and 5 from the rest of the world (2 participants from Japan, and 1 each from Argentina, Australia, and Mexico). In response to the initial questionnaire, 21 of the 40 participants selected pattern analysis as their preferred method for dermoscopic examination of PSL, whereas ABCD rule, Menzies method, and 7-point checklist were preferred by 9, 4, and 3 participants, respectively. Three board members indicated that they were not using specific diagnostic systems. Tables I and II show the results concerning the interobserver and intraobserver reproducibility on diagnostic algorithms and dermoscopic criteria, respectively, as expressed by κ values. Concerning the reproducibility of the first-step diagnosis, pattern analysis, the ABCD rule, Menzies method, and the 7-point checklist, the 40 observers were able to classify PSL with fair to good interobserver agreement and nearly excellent to perfect intraobserver agreement (Table I). Fair interobserver agreement was found in the assessment of global dermoscopic patterns, pigment network, regression and vascular structures, ABCD asymmetry, and Menzies symmetry of pattern. However, 10 dermoscopic features (dots/globules, streaks, blue-whitish veil, blotches, and hypopigmentation from pattern analysis; border, color, and dermoscopic structures from the ABCD rule; and color and positive features from the Menzies method) did not exhibit sufficient interobserver reproducibility, with κ values less than 0.40. Because the 7-point checklist is on the basis of the assessment of the same criteria considered in pattern analysis, similar reproducibility results were obtained for both diagnostic methods. Remarkably, the intraobserver agreement was shown to be good to excellent for all dermoscopic criteria considered (Table II). To assess the validity of dermoscopic criteria for the diagnosis of melanoma, odds ratios were calculated; the results are shown in Tables III and IV. Among the various global features of a given PSL, the feature most predictive for the diagnosis of melanoma was the multicomponent pattern, whereas the globular, cobblestone, homogeneous, and starburst patterns were most predictive for the diagnosis of benign melanocytic lesions (see Appendix II for detailed definitions of criteria). Atypical pigment network, irregular streaks, and regression structures were the local features (included in both pattern analysis and the 7-point checklist) that showed the highest association with melanoma, followed by irregular dots/globules, irregular blotches, and bluewhitish veil (Table III). Vascular structures were not found to be significantly associated with melanoma because they were rarely detectable in this series of cases. Typical pigment network, regular dots/globules, regular streaks, and regular blotches were mostly associated with benign melanocytic lesions. Among the features assessed in the ABCD rule, the asymmetry on both axes exhibited the highest association with melanoma, followed by the presence of more than 4 colors and more than 3 different dermoscopic features. Within the Menzies method the highest association with melanoma was scored by the presence of an asymmetrical distribution of dermoscopic patterns, followed by the presence of more than 1 color and 1 or more positive dermoscopic features. By contrast, the presence of a single color and a symmetrical distribution of pattern were associated with benign melanocytic lesions (Table IV) (see Appendix IV for definitions of criteria). Table V shows the results in terms of sensitivity and specificity that were obtained by 40 observers evaluating 108 PSL. The 40 colleagues were able to correctly classify more than 95% of melanocytic lesions and more than 90% of nonmelanocytic lesions (first step of the unifying concept of dermoscopy) with a positive likelihood ratio of 10. Concerning the differentiation between benign melanocytic lesions and melanoma (second step), the classic dermoscopic approach for diagnosing melanoma, ie, pattern analysis, allowed the best diagnostic performance (sensitivity, 83.7%; specificity, 83.4%; and positive likelihood ratio, 5.1), whereas the alternative algorithms (ABCD rule, Menzies method, and 7-point checklist) revealed similar sensitivity compared with pattern analysis but lower specificity (11.9%-13.4% less specificity) and lower positive likelihood ratio (from 2.8-3.0). Remarkably, when sensitivity and specificity were calculated as a "consensus diagnosis," meaning the specific diagnosis made by the majority of observers, all diagnostic methods allowed better results in terms of sensitivity (100% by pattern analysis; 96.3% by alternative algorithms). Table VI shows the comparison of the 4 diagnostic algorithms for sensitivity and specificity. The sensitivity of the Menzies method was higher than those of the ABCD rule (P=.010) and the 7-point checklist (P=.039), whereas it was not different from that of pattern analysis (P=.442). Remarkably, pattern analysis showed specificity significantly higher than all other systems (P=.000). The lesions included in this study were all considered equivocal from a clinical point of view and had been excised for histopathologic examination. The results assessing the role of dermoscopy for the management decision of PSL show that, on average, 99.2% of melanomas and 98.7% of basal cell carcinomas were judged to require excision or at least follow up examination using digital documentation systems. The mean proportion of melanomas and basal cell carcinomas judged to require excision were 94.8% and 96.6%, respectively, whereas the mean proportion of melanomas and basal cell carcinomas judged to require follow up examination were 4.4% and 2.2%, respectively. Remarkably, 46.4% (mean value) of benign PSL were judged by dermoscopy not to require excision. The interobserver agreement on management decisions made by dermoscopy was fairly good, with a mean κ value of 0.53 (+/- 0.08 SD). Definitions of dermoscopic criteria and diagnostic methods (Appendices I-V) were refined on the basis of the comments and suggestions of the participants.
|Table I. Interobserver and intraobserver agreement on the dermoscopic diagnosis of pigmented skin lesions using various diagnostic algorithms |
| ||Interobserver agreement*||Intraobserver agreement|
|CI, Confidence interval.*Distribution of 4320 observations (40 observations for each of 108 lesions) according to the ratings made by the observers †Melanocytic versus nonmelanocytic lesion.|
|Table II. Interobserver and intraobserver agreement on individual diagnostic criteria used in pattern analysis, ABCD rule, Menzies method, and seven-point checklist|
| ||Interobserver agreement*||Intraobserver agreement †|
|Pattern analysis|| || || || |
|ABCD rule|| || || || |
|Color of the lesion||0.17||0.16-0.18||1.00||0.07-1.00|
|Symmetry of pattern||0.57||0.56-0.58||0.85||0.04-1.00|
|Atypical pigment network||0.45||0.44-0.46||0.63||0.09-1.00|
|Atypical vascular pattern||0.27||0.26-0.28||1.00||0.06-1.00|
|Regression structures|| 0.46||0.45-0.47||0.66||0.18-1.00|
|CI, Confidence interval. |
*Interobserver agreement is on the basis of 85 melanocytic lesions classified by at least 30 of 40 observers; distribution of 3264 observations (38.4 observations for each of 85 lesions) according to the ratings made by the observers.
†Intraobserver agreement is on the basis of 20 lesions of the test set that were re-evaluated by 40 observers.
|Table III. Association between individual diagnostic criteria and melanoma diagnosis obtained by pattern analysis in 85 melanocytic lesions classified by at least 30 of 40 observers: odds ratios and 95% confidence intervals obtained by generalized estimating equation regression models|
|Distribution of assignments*|| |
|Variable||Melanoma||Nonmelanoma||Odds ratio||95% CI|
|Global pattern|| || || || |
|Pigment network || |
|Streaks|| || || || |
|Blue-whitish veil|| || || || |
|Blotches|| || || || |
|Hypopigmentation|| || || || |
|Regression structures|| || || || |
|Vascular structures|| || || || |
|CI, Confidence interval|
*Distribution of 3264 observations (38 4 observations for each of 85 lesions) according to the ratings made by the observers
†Odds ratio is statistically significant for differentiation between melanoma and nonmelanoma since 95% CI does not include the unit
|Table IV. Association between individual diagnostic criteria and melanoma diagnosis obtained by ABCD rule, Menzies method, and 7-point checklist on 85 melanocytic lesions classified at least by 30 of 40 observers*|
|Distribution of assignments*|| |
|Variable||Melanoma||Nonmelanoma||Odds ratio||95% CI|
|ABCD rule|| || || || |
|Asymmetry|| || || || |
|Border|| || || || |
|Color|| || || || |
|Dermoscopic structures|| || || || |
|Menzies method|| || || || |
|Color of the lesion|| || || || |
|More than 1 color||0.99||0.95||18.5||4.1-83.7‡|
|Symmetry of pattern|| || || |
|> 1 positive features||0.94||0.68||7.5||3.0-18.9‡|
|7-point checklist|| || || || |
|Atypical pigment network||0.77||0.27||9.0||4.9-16.3‡|
|Atypical vascular pattern||0.09||0.06||1.5||0.6-3.7|
|CI, Confidence interval|
*Odds ratios and 95% confidence intervals obtained by generalized estimating equation regression models.
†Distribution of 3264 observations (38.4 observations for each of 85 lesions) according to the ratings made by the observers
‡Odds ratio is statistically significant for differentiation between melanoma and nonmelanoma since 95% CI does not include the unit.
|DS Medica S.r.l. | CF/P.IVA 12676030153 | © DS Medica S.r.l. | Privacy | Contatti