Background

Dermoscopy (dermatoscopy, epiluminescence microscopy, incident light microscopy, surface microscopy)
is widely used nowadays for the diagnosis of pigmented skin lesions. The method is getting more and more
popular, not only among dermatologists, but also among oncologists, surgeons, pediatricians,
and even general physicians. However, there is still a need for better standardization of the dermoscopic
terminology, which has not been reviewed since the last (and only) Consensus Meeting on Dermoscopy
was held in Hamburg more than ten years ago, in November, 1989. With a new Consensus Meeting,
we will attempt to refine the dermoscopic terminology and to simplify the diagnostic categories
of pigmented skin lesions. Because dermoscopic images today can be easily transmitted
via telecommunication devices, we plan to organize a virtual Consensus Meeting via the Internet,
known as the Consensus Net Meeting on Dermoscopy (CNMD).

Goals

Main issue

To prove the validity of a two-step procedure for the dermoscopic classification of pigmented skin lesions.

Additional issues

1. Consensus on terminology of dermoscopic criteria by refinement of established criteria and definition
   of new criteria

2. Diagnostic categories, as far as diagnosable by dermoscopy

3. Consensus on management decisions based on dermoscopic evaluation

4. Consensus on follow-up strategies, including digital documentation

5. Consensus on validity of teledermoscopy and elaboration of standard requirements for teledermoscopy (DDDI-standard)

Work plan

The Preliminary scientific design, including the Web design, was presented during an informal meeting in San Francisco at the AAD meeting on March 10, 2000. The Materials and Methods section has since been refined based on the comments and suggestions of the participants. The actual CNMD will start on July 1 and last until October 31, 2000.

First World Congress of Dermoscopy Rome, February 23-25, 2001

The results of the CNMD will be discussed with Consensus Board Members at the World Congress of Dermoscopy in Rome.

Materials and Methods

Materials

a) Cases

• Pigmented skin lesions included in this study are taken from the histopathology files of the Department of Dermatology, University Federico II of Naples (Italy); the Department of Dermatology, University of L’Aquila (Italy); the Sydney Melanoma Unit, Royal Prince Alfred Hospital, Camperdown (Australia); and the Skin and Cancer Associates, Plantation, Florida (USA). Each case includes complete information regarding relevant clinical* and histopathologic data as well as clinical and dermoscopic images and histopathologic specimens.

*Clinical data:

1. age, sex, skin phototype, total number of nevi (0-10, 10-50, >50), personal and/or familial history of melanoma;

2. location, diameter, and duration of the lesion, as well as the history concerning any morphologic changes within the last year before the excision of the lesion.

b) Standards used for Digital Documentation of Dermoscopic Images (DDDI-standard)

• Clinical and dermoscopic images of each case will be obtained using a Heine Dermaphot, with 10-fold magnification of the lesion. The color slides will be converted to digital format using a Kodak Photo CD system.

• Using Adobe Photoshop software all Photo CD images (RGB, 768 x 512 pixel, 72 DPI)
  will be converted to jpeg format with high compression rate.
  Each image will then be optimized for color, brightness,
  and contrast by using the Adobe Photoshop standards (Auto levels and Unsharp mask).

c) Criteria for Cases Selection

• Before inclusion in the study all digital images will be evaluated by a professional photographer
  for focus, brightness and contrast. Only images with excellent quality will be considered for inclusion.

• Only those lesions having a clinical diameter of less than 14 x 10 mm will be included. This is a basic prerequisite in order to evaluate the entire surface of the lesion.

• Each case will then be included in one of two groups of pigmented skin lesions:
  1) Stereotypical examples (low level of diagnostic difficulty);
  2) Dermoscopically equivocal examples (high level of diagnostic difficulty).

• All cases will be evaluated by 3 participants for their level of diagnostic difficulty.
  Each case will be assigned to group 1 or 2 by the agreement of at least 2 out of 3 evaluators.

• All cases will be reviewed by a panel of histopathologists and only histopathologically
  unequivocal lesions will be included in the study.

Methods

• 128 pigmented skin lesions will be randomly subdivided in a training set of 20 cases and a test
  set of 108 cases.

• On the Internet a tutorial will be given concerning the unifying concept of dermoscopy
  (with complete definitions of criteria and example images).

• By entering a password each participant will be able to evaluate 20 cases from the training set.
  For each case, the participants will complete an electronic data sheet.
  Immediately following submission of the completed electronic data sheet, the participants
  will receive the set of dermoscopic criteria of that case as judged by the instructors
  and the histopathologic diagnosis provided by the histopathology panelists.
  This "real-time feedback" should assist the participant in mastering the dermoscopic criteria.
  Each participant will then be able to evaluate on-line 108 cases of the test set.
  A total of 4 months (July 1 to October 31, 2000) will be given for evaluating the cases. 

• At the end of the evaluation period, participants will re-examine 20 randomly selected cases
  from the test set of 108 cases, to test for intra-observer agreement.

• The histopathologic diagnoses for all cases will be posted on the web site after October 31, 2000,
  following evaluation of all cases of the test set and evaluation of the additional 20 cases
  for determining intra-observer agreement.

• The minimum required resolution for viewing cases is 800 x 600 pixels and 24-bit color graphic capability.

• At the beginning of the study all participants will be asked to complete a questionnaire concerning their personal experience in dermoscopy and their preferred diagnostic methods, as well as their computer system specifications. This information will be useful in understanding whether the variability of the dermoscopic methods used in clinical practice, the different learning curves and training experiences,
  and the various computer systems used are affecting the intra- and inter-observer agreement
  on dermoscopic criteria and dermoscopic diagnoses.

Electronic Data Sheet for evaluation of cases

This will include:

1. Clinical information concerning lesion and patient (see Materials section)

2. Melanocytic algorithm for differentiating melanocytic and non-melanocytic pigmented skin lesions

3. Criteria for diagnosing melanocytic lesions by pattern analysis

4. Criteria for the ABCD rule with automatic calculation of the score

5. Criteria for the 7-point checklist with automatic calculation of the score

6. Criteria for the Menzies' method with automatic calculation of the score

7. Diagnostic categories for the final diagnosis together with management recommendations and final remarks

Statistics

Rates of inter- and intra-observer agreement will be analyzed by calculating the percentage of agreement and determining the k index for agreement compared with that expected by chance alone for each descriptor considered [Fleiss LJ, 1981]. Subgroups of descriptors will be compared for distribution of the associated k values by means of two-tailed Wilcoxon rank-sum test. A k value of 1.0 indicates full agreement beyond chance. Values greater than 0.75 are generally considered excellent, values between 0.40 and 0.75 fair to good, and values less then 0.40 poor [Fleiss LJ, 1981]. A p value <= 0.05 (two-sided) will be considered statistically significant.

Donner's method will be used to calculate the sample size for testing two coefficients of inter-observer agreement [Donner A, 1998; Walter et al, 1998]. The sample size requirement for comparing two coefficients of inter-observer agreement will be calculated for dichotomous and continuous outcome variables.

In order to measure intra-class agreement between parameters, for each group we calculated two different numbers of cases necessary to detect both a moderate and a substantial strength of agreement. Choosing the level of significance as alpha=0.05 and power (1-beta=0.80) for testing H0: K1=K2 versus H1: K1 <> K2, the required sample size for group testing would be 36 cases for moderate (k index of agreement = 0.41-0.60) and 97 cases for substantial strength of agreement (k= 0.61-0.80).

Additional statistical analysis will be conducted to test the validity of procedures to detect melanoma.
For this reason, cases will be divided in two groups (melanoma and non-melanoma), based on results of the histopathologic assessment. Using a univariate analysis, the relationship of each criterion with the diagnosis of melanoma will be evaluated by chi-square analysis. Logistic regression analysis will be used to test the diagnostic impact of the whole set of dermoscopic criteria in a multivariate approach, and the probability of each case belonging to the group of melanomas will be calculated. Sensitivity, specificity, efficacy and diagnostic accuracy will be assessed for clinical diagnosis, dermoscopic diagnosis, and the diagnosis provided by the logistic regression approach.

References

1. Donner A. Sample size requirement for the comparison of two or more coefficients of inter-observer agreement.
   Statist Med 1998;17:1157-1168.

2. Walter SD, Eliasziw M, Donner A. Sample size and optimal designs for reliability studies. Statist Med 1998;17:101-110.

3. Fleiss LJ. Statistical methods for rate and proportions. 2nd ed. New York, John Wiley & Sons, 1981: 212-236.