Utility of Lesion Diameter in the Clinical Diagnosis of Cutaneous Melanoma FREE

In 1985, several of us devised the ABCD criteria for melanoma screening¹ to facilitate the early diagnosis of cutaneous melanoma. The ABCD criteria are evidence-based guidelines established to remind physicians that Asymmetry, Border irregularity, Color variegation, and Diameter larger than 6 mm are features characteristic of melanoma. In recent publications, our group reviewed evidence supporting the concept of lesion change in the diagnosis of cutaneous melanoma and formally recommended expansion of the ABCD criteria to include an E criterion for Evolution.^{2 - 3} Rapidly growing melanomas, especially nodular melanomas, frequently lack ABCD features.^{4 - 5} Thus, the addition of an E criterion greatly aids in the diagnosis of banal-appearing melanomas whose sole but critical concerning clinical feature is that of growth or change over time.⁵

As the incidence of melanoma continues to increase worldwide,⁶ the ABCDE criteria remain important to public education in the early recognition of melanoma. An important challenge to the ABCDE criteria has been the recent recognition of small melanomas (≤ 6 mm in diameter). Several authors^{7 - 11} have described such tumors in US, Australian, Italian, and Israeli patient cohorts. Our 2004 publication, which reviewed all available small melanoma data from patient cohorts of more than 30 persons, concluded that small-diameter melanomas likely compose 3% to 14% of all cutaneous melanomas but rarely result in recurrence, metastasis, or death.^{2 ,7 - 12} Nevertheless, the existence of small-diameter melanomas complicates the evaluation of pigmented lesions suggestive of melanoma. Some authors^{7 ,9} have argued that a D criterion of larger than 6 mm may be misleading because strict adherence to this guideline will result in the failure to biopsy small melanomas. Yet others¹³ point out that more than 99% of atypical melanocytic lesions biopsied by dermatologists are benign. Clearly, additional data are needed to help physicians balance their desire to biopsy atypical pigmented lesions suggestive of melanoma with the knowledge that most pigmented skin lesions are benign.

To our knowledge, no large studies have described the relationship between in vivo lesion diameter and frequency of melanoma among lesions biopsied because of concern about melanoma. The main purpose of our study was to answer the question of whether the D criterion of the current ABCDE criteria should be adjusted downward (eg, changing the diameter guideline from 6 to 5 mm) in light of the known existence of small-diameter melanomas. Our study elucidates the relationship between lesion diameter and the diagnosis of melanoma among 1657 consecutive atypical pigmented skin lesions biopsied by dermatologists.

Our group, in collaboration with investigators at Electro-Optical Sciences Inc (EOS), collected and analyzed diameter measurements and histopathologic diagnoses of 1657 consecutive pigmented skin lesions in 1323 patients accrued as part of the second phase of the multicenter MelaFind Study.¹⁴ Patients of any age, race, ethnicity, or sex with pigmented lesions suggestive of melanoma and requiring skin biopsy were eligible for enrollment in the study. The specific reason for biopsy of each lesion was not recorded; however, lesions suggestive of melanoma were those that generally possessed 1 or more of the ABCDE features or were of particular concern to the patient or physician.

Physicians who participated in the study were composed of a group of dermatologists in both academic and community practices (see “Clinical Principal Investigators”) across the United States. They were chosen for their particular interest in the diagnosis of pigmented skin lesions and/or a substantial volume of patients potentially requiring skin biopsies. Of the 1657 lesions, 44.7% were examined in academic practices and 55.3% were examined in community practices. The proportion of melanomas examined in academic practices was 7.6%, vs 8.9% examined in community practices. The observed differences were small and not statistically significant (P = .32). Institutional review board approvals were obtained for all participating sites, and all patients signed written informed consent at enrollment.

MelaFind is an experimental computerized skin imaging system devised by EOS to diagnose melanoma.^{14 - 15} The probe, just slightly larger than a cordless telephone, acquires digital images of pigmented skin lesions illuminated by different spectral bands of light from visible to near infrared. It automatically extracts pertinent data for image analysis by the EOS central computer. In addition, EOS is developing an image analysis algorithm to serve as an aid for physicians in the diagnosis of melanoma. The images and histopathologic diagnoses collected in this study were part of that development effort; hence, no diagnostic information was provided to the physicians from the MelaFind system at any time in the study. For this article, the function of the system was to accurately measure the in vivo diameters of pigmented lesions deemed by dermatologists to be potentially malignant. Diameter measurement is not incorporated into the diagnostic computer algorithm being developed to identify lesions as benign or malignant. The maximum diameter is defined as the longest distance between 2 points located on the lesion border. Diameter measurements are reported to hundredths of millimeters. Image segmentation is based on lesion images obtained at wavelengths of 430 and 500 nm. A proprietary mathematical operation automatically defines which pixels are in the lesion and which are not. Additional methodologic details can be found in 2 previous publications.^{14 - 15} All lesions in the cohort were biopsied, and histopathologic diagnoses were independently rendered by 2 dermatopathologists (one diagnosis was rendered by 1 of us [M.C.M.] for all lesions and the second diagnosis was rendered by 1 of 3 of us [P.G., R.K., or V.P.]) without previous discussion of diagnostic features. To resolve occasional differences in diagnoses between the 2 dermatopathologists, consultation by a third dermatopathologist (1 of 3 of us [P.G., R.K., or V.P.]) was obtained. Such differences occurred in only 6% of cases, a proportion that compares favorably to the considerable disagreement reported across the literature for diagnosis of melanocytic lesions by dermatopathologists.^{16 - 18}

At initiation of this analysis, 1657 lesions had been accrued and appraised in the study. We sorted these lesions into categories based on diameter and diagnosis. In our analysis, we were particularly interested in quantifying the number and proportion of melanomas among this cohort of clinically suspicious pigmented lesions and among its smaller subset of melanocytic neoplasms (n = 1380). Sensitivity (true positive/[true positive + false negative]) and specificity (true negative/[false positive + true negative]) were calculated for various cut points for the diagnosis of melanoma (invasive and in situ). The following definitions were used: true positive was defined as melanomas with diameters greater than the cutoff value, false positive was defined as nonmelanomas with diameters greater than the cutoff value, true negative was defined as nonmelanomas with diameters less than the cutoff value, and false negative was defined as melanomas with diameters less than the cutoff value. A receiver operating characteristic (ROC) curve was estimated nonparametrically to visually summarize the accuracy of the predictions.¹⁹ The unit of analysis throughout the study was pigmented lesion(s), not individual patients. In patients with more than 1 clinically suspicious pigmented lesion, each was assessed separately with regard to its diameter and diagnosis. In this study the patients contributed 1.25 lesions per patient. When we accounted for multiple lesions per patient, the area under the curve for the ROC curve was exactly the same (0.68; 95% confidence interval, 0.63-0.73) as when we used the regular nonclustered methods.²⁰

A total of 852 women and 805 men participated in the study. The mean age of the women was 43.5 years (median age, 42 years); the mean age of the men was 50 years (median age, 51 years). Among the 1657 lesions under study, 138 (8.3%) were histologically diagnosed as melanoma, including invasive (n = 54) and in situ (n = 84) lesions. Other diagnoses included dysplastic nevi (n = 1016), other melanocytic nevi (eg, Spitz and congenital) (n = 226), and other nonmelanocytic lesions (eg, seborrheic keratosis) (n = 277) (Table 1).

Eight hundred four lesions (48.5%) were greater than 6 mm in diameter, and 853 (51.5%) were 6 mm in diameter or less. A representative MelaFind image used to calculate lesion diameter is shown in Figure 1. Within each 1-mm diameter range from 2.01 to 6.00 mm, the proportion of melanomas did not vary significantly, remaining stable at 3.6% to 4.5% (Table 2 and Figure 2). However, we observed a nearly 100% increase in the proportion of melanomas when comparing the 5.01- to 6.00-mm category (4.3%) to the 6.01- to 7.00-mm category (8.3%). Among lesions 6.01 mm in diameter or larger, the proportion of melanomas ranged from 8.1% to 21.9%, roughly increasing as lesion diameters increased. We analyzed the proportions of invasive and in situ melanomas separately and found that invasive melanomas were diagnosed in 13 of 853 lesions (1.5%) that were 6 mm in diameter or smaller and in 41 of 804 lesions (5.1%) that were larger than 6 mm in diameter, whereas in situ melanomas were diagnosed in 22 of 853 lesions (2.6%) that were 6 mm in diameter or smaller and in 62 of 804 lesions (7.7%) that were larger than 6 mm in diameter. Recognizing the importance of Breslow thickness to prognosis in melanoma, we also analyzed mean tumor thickness for invasive melanomas in each interval listed in Table 2 but could not identify trends or a reasonable cutoff based on tumor thickness (data not shown).

We were also interested in determining whether the striking increase in melanoma proportion among lesions greater than 6 mm would also be observed when the analysis was restricted only to melanocytic neoplasms (nevi and melanoma). Table 3 reports the frequency of melanomas vs nevi of various diameters. Melanomas comprise in situ and invasive lesions; the remaining melanocytic neoplasms comprise dysplastic nevi and other nevi, which include Spitz nevi, congenital nevi, and blue nevi. We again observed a 100% increase in the proportion of melanomas among the 6.01- to 7.00-mm category (9.9%) vs the 5.01- to 6.00-mm category (4.9%). There was a generally constant proportion of melanomas (4.0%-5.3%) among lesions 6 mm in diameter or smaller but a significantly larger and increasing proportion of melanomas (9.9%-28.6%) among lesions with diameters greater than 6 mm.

Although it was not the intent of this study to analyze the diameter criterion as a stand-alone screening test for melanoma, we performed an ROC analysis to assess the clinical utility of various diameter measurements in the diagnosis of melanoma. Specifically, we calculated the sensitivity and specificity of diameters of 4, 5, 6, 7, 8, 9, and 10 mm for the diagnosis of melanoma (Table 4). As diameter increased from 4 to 10 mm, sensitivity decreased for the diagnosis of melanoma and specificity increased. The 6-mm diameter criterion yielded a sensitivity of 75% and a specificity of 54%. In comparison, a 5-mm cutoff had a sensitivity of 83% and a specificity of 38% and a 7-mm cutoff had a sensitivity of 63% and a specificity of 65%.

Our study design is unique in several ways. First, we used computer technology to measure the diameters of pigmented lesions with great accuracy and precision, thus considerably reducing the possibility of interobserver variability introduced by human measurements of skin lesions. Second, we measured pigmented lesions in vivo and reported biopsy results of all lesions studied. Third, we studied a large cohort of suspicious but mostly benign lesions to better understand the frequency of melanomas among clinically suspicious atypical pigmented lesions. With the exception of 1 study,⁹ other authors^{7 - 11 ,21} have focused their attention on populations of ex vivo biopsy-confirmed small melanomas. Ex vivo specimen measurement may result in underestimation of lesion diameter by as much as 20%.^{22 - 23} The rigor of our study design allows much greater insight into the relationship between lesion diameter and diagnosis than any previous study by using accurate lesion measurements rather than the rough estimates not uncommon in dermatologic practice.

The goal of this study was to determine if a diameter measurement of greater than 6 mm was still a useful guideline in the diagnosis of melanoma. Ideally, we would want to determine the frequency of melanomas among all pigmented skin lesions; however, this would be exceedingly difficult. Melanoma is still a relatively rare tumor, and the average adult patient has between 11 and 50 nevi, with 5% of patients having at least 1 atypical nevus.²⁴ Thus, to conduct a clinical trial to calculate the frequency of melanomas among all pigmented skin lesions would require the removal and histopathologic diagnosis of all pigmented skin lesions on each patient, a nearly impossible and extraordinarily expensive task. The present study took advantage of a cohort of suspicious clinically atypical lesions that were biopsied primarily because of a concern for melanoma. This study design would likely enrich the cohort with melanomas compared with a study that biopsied all pigmented lesions, yet even within this more selective cohort, the rate of melanoma was low. Although we do not know how many melanomas were missed by this group of physicians, the physicians who enrolled patients into the MelaFind study were experienced dermatologists who are, arguably, less likely to miss many malignant lesions or to select benign lesions for further appraisal than their less specialized colleagues. Nevertheless, the fact that most lesions biopsied in the MelaFind Study (853 of 1657) were less than 6 mm in diameter suggests a substantial concern among physicians not to miss small-diameter melanomas. Despite the inherent selection bias of our study design, our data provide strong support for the idea that small melanomas are rare, even among a cohort of clinically atypical pigmented lesions that experienced physicians may excise.

We recommend that a diameter criterion of larger than 6 mm remain a part of the ABCDE criteria. Lesion diameter greater than 6 mm seems to be a significant indicator of increased suspicion for melanoma because our data reveal a significant increase in the proportion of melanomas among lesions larger than 6 mm when compared with lesions 6 mm or smaller (Figure 2 and Table 2). In addition, data from the ROC analysis suggest that downward revision of the diameter criterion to 5 mm would result in a significant loss of specificity. Because more than 50% of all suspicious lesions biopsied in this study were 6 mm or smaller yet only 1.5% of those were invasive melanoma and 2.6% were in situ melanoma, it seems that many biopsies are being performed on small lesions based on their A, B, C, and E characteristics, with a low specificity for detecting melanoma. It is likely that lowering the D criterion to 5 mm would only further decrease the specificity of such biopsies.

For pigmented lesions with diameters that measure 6 mm or less, we believe that the use of the ABCDE² criteria in combination can be an effective management strategy. Specifically, identifying lesions that possess more atypical features (eg, A, B, C, or E) is likely to be more specific for melanoma. Thomas et al²⁵ demonstrated that specificity for the diagnosis of melanoma increases markedly as the number of ABCDE criteria present in a pigmented lesion increase from 1 or more (36%) to 4 or more (94%); however, there is a concomitant decrease in sensitivity from 97% to 54%, respectively. Our ROC analysis indicates that a diameter criterion of larger than 6 mm is not sensitive or specific enough to serve as a stand-alone screening test for melanoma, consistent with the observations of Thomas et al. Unfortunately, in the present study we were unable to examine the utility of the A, B, C, or E criteria or combinations thereof; however, Thomas et al found that horizontal enlargement had the greatest combination of sensitivity and specificity for melanoma, followed by the D criterion, so it may be especially helpful to consider lesional evolution (E), such as horizontal enlargement, in lesions 6 mm or smaller.

The significant increase in the proportion of melanomas among lesions greater than 6 mm in diameter suggests that a lesion size of 6 mm may have biological significance. The current theory regarding cellular senescence in melanocytic nevi^{26 - 27} predicts that nevus cells undergo a certain number of cell divisions, after which cellular mechanisms of long-term growth arrest (eg, tumor suppressor genes) halt proliferation. Loss of these growth arrest mechanisms is common in melanoma²⁸ ; however, much additional work is needed to correlate the frequency of these biological alterations with lesion size.

In conclusion, the present study reaffirms that a diameter criterion of larger than 6 mm is a useful guideline for the early detection of cutaneous melanoma. We do not recommend downward revision of the D criterion at this time. In the United States, rates of melanoma and nonmelanoma skin cancers have markedly increased, and skin biopsy rates have more than doubled in 20 years.²⁹ In an era that demands greater data to support clinical decision making, the ABCDE criteria are valuable evidence-based guidelines to aid physicians in decisions regarding the biopsy of pigmented lesions of the skin.

Correspondence: David Polsky, MD, PhD, Ronald O. Perelman Department of Dermatology, New York University School of Medicine, 550 First Ave, Room H-100, New York, NY 10016 (David.Polsky@nyumc.org).

Accepted for Publication: June 20, 2007.

Author Contributions: Dr Polsky had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Abbasi, Prieto, Osman, Friedman, Rigel, Kopf, and Polsky. Acquisition of data: Abbasi, Gutkowicz-Krusin, Mihm, Googe, King, and Prieto. Analysis and interpretation of data: Abbasi, Yancovitz, Panageas, Prieto, Osman, and Polsky. Drafting of the manuscript: Abbasi, Prieto, and Polsky. Critical revision of the manuscript for important intellectual content: Abbasi, Yancovitz, Gutkowicz-Krusin, Panageas, Mihm, Googe, King, Prieto, Osman, Friedman, Rigel, Kopf, and Polsky. Statistical analysis: Panageas. Administrative and technical support: Gutkowicz-Krusin. Study supervision: Osman and Polsky.

Financial Disclosure: Dr Gutkowicz-Krusin is an employee of EOS; Drs Mihm, Googe, King, and Prieto are dermatopathologists for the MelaFind Study, sponsored by EOS; Dr Friedman is a consultant for and shareholder in EOS; Dr Rigel is a consultant for EOS; Drs Kopf and Polsky are investigators for the MelaFind Study sponsored by EOS; and Dr Polsky is a consultant for EOS.

Role of the Sponsor: The data were extracted from a clinical study sponsored by EOS; EOS provided data to the research team but had no role in study design, data analysis, or data interpretation.

Clinical Principal Investigators: Jeffrey Callen, MD, University of Louisville, Louisville, Kentucky; Armand Cognetta, MD, Dermatology Associates of Tallahassee, Tallahassee, Florida; Caron Grin, MD, University of Connecticut, Hartford; Kenneth Gross, MD, Skin Surgery Medical Group, San Diego, California; Allan Halpern, MD, Memorial Sloan-Kettering Cancer Center, New York, New York; Peter Lee, MD, University of Minnesota, Minneapolis; Seth Lerner, MD, Adult and Pediatric Dermatology Specialists, Trumbull, Connecticut; Norman Levine, MD, University of Arizona Health Sciences Center, Tucson; Gary Monheit, MD, Total Skin and Beauty Dermatology, Birmingham, Alabama; Gary Peck, MD, Washington Cancer Institute, Washington, DC; Harold Rabinovitz, MD, Skin & Cancer Associates, Plantation, Florida; Jennifer Schwartz, MD, University of Michigan, Ann Arbor; Nancy Thomas, MD, PhD, University of North Carolina, Chapel Hill; Yardy Tse, MD, Dermatology Associates of San Diego, Encinitas, California.