Dermatoscopy of basal cell carcinoma: Morphologic variability of global and local features and accuracy of diagnosis

Article Outline

• Abstract
• Methods
  • Lesion selection and dermatoscopic equipment
  • Dermatoscopic criteria of BCC
  • Test set of lesions
  • Statistical analysis
• Results
  • Global dermatoscopic features
  • Local dermatoscopic features
  • Distribution of dermatoscopic patterns
  • Accuracy and interrater reliability of dermatoscopic diagnosis of BCC
• Discussion
• References
• Copyright

Background

Early detection of basal cell carcinoma (BCC) is crucial to reduce the morbidity of this tumor.

Objective

We sought to investigate the variability and diagnostic significance of dermatoscopic features of BCCs.

Methods

We conducted retrospective dermatoscopic analysis of 609 BCCs and 200 melanocytic and nonmelanocytic lesions, and assessment of interrater reliability of dermatoscopic BCC criteria.

Results

Lesions included nonpigmented (15.1%), lightly pigmented (33.2%), pigmented (42.7%), and heavily pigmented (9%) BCCs. Classic BCC patterns including arborizing telangiectasia (57.1%), blue/gray ovoid nests (47.5%), ulceration (39.2%), multiple blue/gray globules (26.1%), leaflike areas (15.9%), and spoke-wheel areas (9%) were significantly increased in pigmented BCCs compared with nonpigmented and heavily pigmented BCCs (P = .0001). Among nonclassic BCC patterns, we detected short fine superficial telangiectasia (10%) and multiple small erosions (8.5%), and described two new patterns named “concentric structures” (7.6%) and “multiple in-focus blue/gray dots” (5.1%). Dermatoscopic features suggestive of melanocytic lesions (eg, multiple brown to black dots/globules, blue/white veillike structures, and nonarborizing vessels) were observed in 40.6% BCCs and significantly increased in heavily pigmented BCCs (P < .0001). Expert observers provided an accurate (sensitivity: 97%) and reliable (K: 87%) dermatoscopic diagnosis of BCC, although a significant difference in terms of specificity (P = .0002) and positive predictive value (P = .0004) was found. Arborizing telangiectasia, leaflike areas, and large blue/gray ovoid nests represented reliable and robust diagnostic parameters.

Limitation

The study was retrospective.

Conclusion

BCCs show a large spectrum of global and local dermatoscopic features; heavily pigmented BCCs show the most challenging combinations of dermatoscopic features.

Key words: basal cell carcinoma, dermatoscopy, early diagnosis of skin tumors

Basal cell carcinoma (BCC) is a slowly growing malignant epithelial skin tumor predominantly affecting middle-aged and fair-skinned individuals.¹ Clinicopathologic appearances of BCC are various and include nodular, superficial, morpheic, and pigmented variants.¹, ² Epidemiologic data indicate that the global incidence is increasing worldwide particularly in younger age groups and BCC emerges as a growing public health problem.³ The differential diagnosis between BCC and other skin lesions is of critical importance, and serious morbidity may result from an undiagnosed tumor.

Noninvasive procedures such as dermatoscopy, high-frequency ultrasound, optical coherence tomography, and reflectance confocal microscopy have been developed for the diagnosis of skin cancers.⁴, ⁵, ⁶, ⁷ Among these, dermatoscopy represents so far the diagnostic tool with the highest clinical impact in dermatologic practice to better differentiate benign from malignant skin lesions and to detect tumors in an early stage.⁶, ⁷

The dermatoscopic model for the diagnosis of the pigmented variant of BCC is based on the absence of a pigmented network and the presence of at least one positive feature including: (1) ulceration (not associated with a recent history of trauma); (2) multiple blue/gray globules; (3) leaflike areas; (4) large blue/gray ovoid nests; (5) spoke-wheel areas; and (6) arborizing (treelike) telangiectasia.⁸ However, BCC may exhibit a large variety of clinical and dermatoscopic characteristics that are the result of a wide range of combinations of histopathological features.¹, ⁸ Different case reports described BCCs with unusual clinical and dermatoscopic aspects and recent studies emphasized the possible role of additional features such as short fine telangiectasia and multiple small ulcerations as further criteria for the dermatoscopic diagnosis of superficial BCC.⁹, ¹⁰, ¹¹, ¹², ¹³, ¹⁴, ¹⁵

The aims of this study were to: (1) describe the morphologic variability of the dermatoscopic features in a large set of BCCs; (2) define the distribution of different dermatoscopic features in specific types of BCC; and (3) evaluate the accuracy and reliability of dermatoscopic diagnosis and the diagnostic significance of BCC-specific patterns.

Back to Article Outline

Methods 

Lesion selection and dermatoscopic equipment 

Dermatoscopic images of the skin lesions included in this study were from the digital databases of the outpatient clinics of the Departments of Dermatology of the University of L'Aquila (L'Aquila, Italy) and of the tertiary referral center of the Sydney Melanoma Diagnostic Center (Sydney, Australia). Lesions were collected between January 1991 and May 2007 using 3 pieces of dermatoscopic equipment: (1) SolarScan system (Polartechnics Ltd, Sydney, Australia); (2) Nevuscreen system (Arkè s.a.s., Avezzano, Italy); and (3) Dermaphot (Heine Ltd, Herrshing, Germany). Large lesions present on the databases but not completely comprised within the field of view were not included in the study. Demographic data such as age, sex, and anatomic site of the lesions were also collected for each patient.

Dermatoscopic criteria of BCC 

In the first phase of the study, dermatoscopic images of 609 histopathologically proven BCCs were scored for global and local dermatoscopic features as previously described.⁸, ¹⁶ Individual lesions were categorized according to the degree of dermatoscopic pigmentation as: (1) nonpigmented BCCs, characterized by the absence of dermatoscopic brown, black, blue, or gray pigmentation; (2) lightly pigmented BCCs, showing pigmented areas involving less than 30% of the lesion's surface; (3) pigmented BCCs, displaying pigmented areas involving 30% to 70% of the lesion's surface; and (4) heavily pigmented BCCs, characterized by the presence of dermatoscopic pigmentation in more than 70% of the lesion's surface. Scored colors included blue, black, gray, light brown, dark brown, and red. Red and/or white structureless areas were evaluated to define the surface of nonpigmented regions.⁸

According to morphologic definition, local dermatoscopic aspects that displayed specific criteria to diagnose pigmented BCC were classified as classic BCC patterns (ie, ulceration, multiple blue/gray globules, leaflike areas, large blue/gray ovoid nests, spoke-wheel areas, and arborizing telangiectasia).⁸ Local features not classifiable in this category but representing a possible variation on the theme of the above patterns were categorized as nonclassic BCC patterns (Table I). Finally, dermatoscopic patterns more frequently observed in melanocytic lesions compared with pigmented BCCs were included in the category of melanocytic patterns.⁸

Table I. Morphologic definition of classic basal cell carcinoma patterns and nonclassic basal cell carcinoma patterns

BCC, Basal cell carcinoma.

Test set of lesions 

In the second phase of the study, the interrater reliability of dermatoscopic features and diagnosis of BCC was assessed among 3 observers experienced in dermatoscopic evaluation (G. A., I. Z., and H. P. S.). The test set comprised 300 skin lesions including 150 BCCs (38 pigmented, 38 heavily pigmented, 37 nonpigmented, and 37 lightly pigmented), 50 melanomas (median Breslow thickness 0.4 mm; range 0-2.7 mm), 50 melanocytic nevi, and 50 nonmelanocytic skin lesions, with a similar degree and distribution of pigmentation, randomly selected from our digital databases (Table II). The test set was presented in a mixed order to the observers, who scored the images blinded to the histopathologic diagnosis and without knowledge of any clinical data of the patients and lesions. Observers were first asked to make the diagnosis (BCC vs not BCC) of each lesion; subsequently, if a BCC was diagnosed, observers reported the BCC dermatoscopic features categorized as classic and nonclassic BCC patterns.

Table II. Diagnosis and frequency of lesions included in test set

BCC, Basal cell carcinoma.

Statistical analysis 

Differences between proportions of contingency tables were assessed using the Chi-square test. Chi-square statistic for trend was used to test the null hypothesis of no association between proportion of variation and categorical variables. All P values cited are two-sided, and values of P less than .05 were considered statistically significant. Sensitivity, specificity, positive predictive value, and negative predictive value for the dermatoscopic diagnosis of BCC were calculated for each observer as compared with histopathologic diagnosis. Classic and nonclassic BCC patterns were considered as positive dermatoscopic features to diagnose BCC. The agreement between ratings made by 3 observers on dermatoscopic diagnosis and specific patterns of BCC (interrater reliability) was estimated using Cohen kappa statistic with 95% confidence intervals.¹⁷, ¹⁸ Populations with trait prevalence closer to 50% were considered well-balanced indices for more reliable K values.¹

Back to Article Outline

Results 

Dermatoscopic images of BCCs were from 609 individuals, including 357 male (58.6%) and 252 female (41.4%) patients with a median age of 63 years (range: 17-98 years). Individual lesions were located on the face (173/609; 28.4%), back (168/609; 27.6%), front of the trunk (154/609; 25.3%), scalp (59/609; 9.7%), lower (21/609; 3.4%) and upper (17/609; 2.8%) extremities, and neck (17/609; 2.8%). The populations included predominantly non-Asian or non-black-skinned individuals.

Global dermatoscopic features 

Lesions included 260 of 609 (42.7%) pigmented, 202 of 609 (33.2%) lightly pigmented, 92 of 609 (15.1%) nonpigmented, and 55 of 609 (9%) heavily pigmented BCCs. The colors observed by dermatoscopy were blue (495/609; 81.3%), red (430/609; 70.6%), light brown (321/609; 52.7%), gray (299/609; 49.1%), dark brown (167/609; 27.4%), and black (65/609; 10.7%). Multiple colors (≥3) were identified in 276 of 609 (45.3%) lesions. Nonpigmented and lightly pigmented BCCs exhibited a shiny white to red background in 245 of 294 (83.3%) lesions, and a white scarlike depigmentation appearance in 49 of 294 (16.6%) lesions. Heavily pigmented BCCs showed a multicolored background in 37 of 55 (67.3%) lesions, and a single color background in 5 of 55 (9.1%) lesions.

Local dermatoscopic features 

Dermatoscopic evaluation showed the presence of classic BCC patterns in 583 of 609 (95.7%) lesions. In all, 499 of 517 (96.5%) pigmented BCCs had the overall pattern of an absent network and one or more positive feature as previously described.⁸ The most common pattern was the arborizing (treelike) telangiectasia, identified in 348 of 609 (57.1%) BCCs. Large blue/gray ovoid nests were detected in 289 of 609 (47.5%) lesions, ulceration was seen in 239 of 609 (39.2%) lesions, and multiple blue/gray globules were observed in 159 of 609 (26.1%) lesions. In all, 97 (15.9%) and 55 (9%) of the 609 BCCs showed leaflike areas and spoke-wheel areas, respectively (Fig 1). In all, 499 of 517 (96.5%) pigmented BCCs had the overall pattern of an absent pigmented network and one or more positive feature as previously described.⁸

• 
  • View Large Image
  • Download to PowerPoint

• Fig 1. 

  Classic basal cell carcinoma criteria included: arborizing (treelike) telangiectasia (→), large blue/gray ovoid nests (⫫), ulceration (▴), multiple blue/gray globules (■), maple-leaflike areas (●), and spoke-wheel areas (⊥).

Nonclassic BCC patterns were shown in 159 of 609 (26.1%) BCCs and included: (1) short fine superficial telangiectasia (61/609; 10%); (2) multiple small erosions (52/609; 8.5%); (3) concentric structures (46/609; 7.6%); and (4) multiple in-focus blue/gray dots (not pepper-like) (31/609; 5.1%) (Fig 2). Among pigmented BCCs, the overall dermatoscopic appearance of 9 of 517 (1.8%) lesions showed the presence of one nonclassic BCC pattern in the absence of pigmented network and classic positive criteria.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 2. 

  Nonclassic basal cell carcinoma dermatoscopic criteria included: short fine superficial telangiectasia (▴), multiple small erosions (→), concentric structures (⫫), and multiple in-focus blue/gray dots (⊥).

Dermatoscopic patterns indicative of melanocytic lesions were found in 247 of 609 (40.6%) BCCs. In such cases the following features were detected: (1) multiple brown to black dots/globules (132/609; 21.6%); (2) blue/white veillike structures (87/609; 14.3%); (3) pigmented network (11/609; 1.8%); (4) multiple blue/gray dots (pepper-like) (7/609; 1.2%); and (5) radial streaming or pseudopods (4/609; 0.6%). In addition, nonarborizing vascular features were observed in 70 of 609 BCCs (11.5%) and included dotted vessels (45/609; 7.4%), linear irregular vessels (30/609; 4.9%), hairpin vessels (16/609; 2.6%), and comma vessels (6/609; 1%).

Distribution of dermatoscopic patterns 

Classic BCC patterns were significantly more frequent in lightly pigmented and pigmented types of BCC (P = .0001) than in nonpigmented and heavily pigmented BCCs. In contrast, melanocytic patterns were significantly more frequent in heavily pigmented BCCs as compared with the other variants (P < .0001) (Fig 3). No significant difference in terms of distribution of nonclassic BCC patterns in different types of BCC was found. In addition, the frequency of melanocytic patterns significantly increased with the increasing rate of the pigmentation of BCCs (test for trend P value < .0001). Table III summarizes the overall distribution of dermatoscopic patterns according to the different types of BCC.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 3. 

  Two cases of heavily pigmented basal cell carcinomas (BCC) dermatoscopically characterized by prevalent distribution of brown to black dots and globules (→) and diffuse blue-white veillike structure (▴). Interestingly, first lesion lacks pigmented network and has multiple blue/gray globules that would satisfy classic pattern observed in pigmented BCC (⊥).

Table III. Distribution of dermatoscopic patterns according to different types of basal cell carcinoma

BCC, Basal cell carcinoma.

Accuracy and interrater reliability of dermatoscopic diagnosis of BCC 

Sensitivity, specificity, positive predictive value, and negative predictive value for the dermatoscopic diagnosis of BCC obtained by observers 1, 2, and 3 are reported in Table IV. There was no significant difference between the sensitivity and negative predictive value between observers 1, 2, and 3. A significant difference was detected between specificity and positive predictive value (P = .0002 and P = .0004, respectively). The interrater reliability of dermatoscopic diagnosis of BCC obtained a Cohen kappa statistic value of .87 (95% confidence interval: .82-.91).¹⁷, ¹⁸ Table V summarizes the overall prevalence of dermatoscopic BCC criteria obtained by the observers and the interrater reliability of ratings.

Table IV. Sensitivity, specificity, positive predictive value, and negative predictive value for dermatoscopic diagnosis obtained by observers 1, 2, and 3 in test set

NPV, Negative predictive value; PPV, positive predictive value.

Table V. Overall observers' prevalence index of dermatoscopic basal cell carcinoma patterns and interrater reliability of ratings (K coefficient)

CI, Confidence interval; BCC, basal cell carcinoma.

Back to Article Outline

Discussion 

In this study we investigated the dermatoscopic variability of global and local features of BCCs and analyzed the interobserver agreement (interrater reliability) on patterns and diagnosis of this neoplasm. We validated the dermatoscopic method previously described to diagnose pigmented BCC,⁸ and showed the presence of the 6 classic BCC dermatoscopic criteria in 96.5% of our set of pigmented BCCs. The frequency of multiple blue/gray globules, leaflike areas, and spoke-wheel areas was almost identical to that reported in the previous study, whereas we detected a higher frequency of arborizing (treelike) telangiectasia and ulceration and a lower rate of large blue/gray ovoid nests.⁸ Lesion selection could account for these differences, as we included in our set both pigmented and nonpigmented types of BCC.

BCC may also display dermatoscopic features commonly found in melanocytic lesions such as brown to black dots/globules, blue/white veil, pigmented network, pseudopods, radial streaming, or a polymorphous vascular pattern.⁹, ¹⁰, ¹¹, ¹², ¹⁶ In the current study, dermatoscopic features suggestive of melanocytic lesions were observed in 40.6% of BCCs and mainly included multiple brown to black dots/globules, blue/white veillike structures, and vascular features such as dotted, linear irregular, hairpin, and comma vessels. Remarkably, we showed that the frequency of melanocytic patterns linearly increased with the pigmentation of the lesions, having a prevalent distribution in the heavily pigmented BCCs. These findings suggest that, among pigmented BCCs, the heavily pigmented variant represents the most difficult type to be differentiated from both melanocytic nevi and melanomas, and confirm that the global aspect should always be evaluated in the dermatoscopic interpretation of the lesions.

Local dermatoscopic features classified in this study as nonclassic BCC patterns included short fine telangiectasia, which were observed in 10% of BCCs and have been interpreted as an early variant of the arborizing vessels. In addition, multiple small erosions were detected in 8.5% of BCCs and thought to represent the initial stage of clear-cut ulceration. Consistent with our observation, short fine telangiectasia and multiple small ulcerations randomly located in a shiny white to red structureless background have been reported as additional dermatoscopic criteria for the diagnosis of superficial BCC.^14,15 Other nonclassic BCC patterns to our knowledge described for the first time in our study included concentric structures and multiple in-focus blue/gray dots, which were seen in 7.6% and 5.1% of the lesions, respectively. Concentric structures appeared as round, concentrically overlapped structures that resembled the early stage of a spoke-wheel area lacking the characteristic peripheral radial projection. Furthermore, multiple in-focus blue/gray dots represented foci of sharply focused dermatoscopic features that were differentiated from the pepper-like dermatoscopic appearance of melanophages and from black dots of melanocytic lesions. Based on our results, we hypothesized that the nonclassic BCC criteria described herein may further support the diagnosis of BCC, particularly in early lesions that may lack BCC classic patterns.

The dermatoscopic model proposed by Menzies et al⁸ to diagnose pigmented BCCs is indeed not an effective discriminator for lesions lacking pigmentation. In the current study we also examined the dermatoscopic features of 92 nonpigmented BCCs showing that the arborizing telangiectasia was the most common criterion, followed by ulceration, short fine superficial telangiectasia, nonarborizing vessels, and multiple small erosions. We believe that features such as ulceration or small erosions and arborizing or short fine superficial telangiectasia might be useful to diagnose nonpigmented variant of BCC, although further studies are needed to validate these observations.

Our group previously demonstrated the substantial reproducibility of the dermatoscopic diagnostic criteria¹⁹ proposed by Menzies et al.⁸ Interestingly, the K statistic coefficient is considered an appropriate measure of the reliability of the clinicians' ratings in areas such as diagnosis and the interpretation findings.¹⁷, ¹⁸ In the current study, all the observers provided high levels of diagnostic sensitivity for BCC, and the dermatoscopic diagnosis was considered substantially reliable because of the 87% of diagnostic K agreement. However, the prevalence index of ratings may influence the magnitude of K coefficient, and a more appropriate estimation of the reliability of findings can be obtained with trait prevalence closer to 50%.²⁰ In fact, among dermatoscopic patterns, arborizing telangiectasia, leaflike areas, and large blue/gray ovoid nests seem to represent reliable and robust parameters to diagnose BCC, whereas short fine superficial telangiectasia appears as a less reproducible diagnostic criterion.

In conclusion, we described a large spectrum of global and local dermatoscopic features of BCC, and we reported the 3 most robust and reliable BCC parameters (ie, arborizing telangiectasia, leaflike areas, and large blue/gray ovoid nests). In addition, we demonstrated that heavily pigmented BCCs show the most challenging combinations of dermatoscopic features.

Back to Article Outline

References 

1. Roewert-Huber J, Lange-Asschenfeldt B, Stockfleth E, Kerl H. Epidemiology and etiology of basal cell carcinoma. Br J Dermatol. 2007;157(Suppl):47–51
   • View In Article
   • CrossRef
2. Leiter U, Garbe C. Epidemiology of melanoma and non-melanoma skin cancer–the role of sunlight. Adv Exp Med Biol. 2008;624:89–103
   • View In Article
   • CrossRef
3. Telfer NR, Colver GB, Morton CA. Guidelines for management of basal cell carcinoma. Br J Dermatol. 2008;159:35–48
   • View In Article
   • CrossRef
4. Kittler H, Pehamberger H, Wolff K, Binder M. Diagnostic accuracy of dermoscopy. Lancet Oncol. 2002;3:159–165
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (337 KB)
   • CrossRef
5. Marghoob AA, Swindle LD, Moricz CZ, Sanchez Negron FA, Slue B, Halpern AC, et al. Instruments and new technologies for the in vivo diagnosis of melanoma. J Am Acad Dermatol. 2003;49:777–797
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (1259 KB)
   • CrossRef
6. Mogensen M, Jemec JB. Diagnosis of nonmelanoma skin cancer/keratinocyte carcinoma: a review of diagnostic accuracy of nonmelanoma skin cancer diagnostic tests and technologies. Dermatol Surg. 2007;33:1158–1174
   • View In Article
   • CrossRef
7. Ulrich M, Stockfleth E, Roewert-Huber J, Astner S. Noninvasive diagnostic tools for nonmelanoma skin cancer. Br J Dermatol. 2007;157(Suppl):56–58
   • View In Article
   • CrossRef
8. Menzies SW, Ingvar C, Crotty KA, McCarthy WH. Surface microscopy of pigmented basal cell carcinoma. Arch Dermatol. 2000;136:1012–1016
   • View In Article
   • CrossRef
9. Ferrari A, De Angelis L, Peris K. Unusual clinical and dermoscopic features in two cases of pigmented basal cell carcinoma. J Am Acad Dermatol. 2005;53:1087–1089
   • View In Article
   • Full Text
   • Full-Text PDF (233 KB)
   • CrossRef
10. Felder S, Rabinovitz H, Oliviero M, Kopf A. Dermoscopic differentiation of a superficial basal cell carcinoma and squamous cell carcinoma in situ. Dermatol Surg. 2006;32:423–425
    • View In Article
    • MEDLINE
    • CrossRef
11. Felder S, Rabinovitz H, Oliviero M, Kopf A. Dermoscopic pattern of pigmented basal cell carcinoma, blue-white variant. Dermatol Surg. 2006;32:569–570
    • View In Article
    • MEDLINE
    • CrossRef
12. Rossiello L, Zalaudek I, Cabo H, Ferrara G, Gabriel C, Argenziano G. Dermoscopic-pathologic correlation in an unusual case of pigmented basal cell carcinoma. Dermatol Surg. 2006;32:1509–1512
    • View In Article
    • MEDLINE
    • CrossRef
13. Zalaudek I, Ferrara G, Broganelli P, Moscarella E, Mordente I, Giacomel J, et al. Dermoscopy patterns of fibroepithelioma of pinkus. Arch Dermatol. 2006;142:1318–1321
    • View In Article
    • CrossRef
14. Giacomel J, Zalaudek I. Dermoscopy of superficial basal cell carcinoma. Dermatol Surg. 2005;31:1710–1713
    • View In Article
    • MEDLINE
15. Scalvenzi M, Lembo S, Francia MG, Balato A. Dermoscopic patterns of superficial basal cell carcinoma. Int J Dermatol. 2008;47:1015–1018
    • View In Article
    • CrossRef
16. Menzies SW, Ingvar C, Crotty K, McCarthy W. Frequency and morphologic characteristics of invasive melanomas lacking specific surface microscopic features. Arch Dermatol. 1996;132:1178–1182
    • View In Article
17. Fleiss JL. Statistical methods for rates and proportions. 2nd ed.. New York: John Wiley and Sons; 1981;p. 212-36
    • View In Article
18. Sim J, Wright CC. The Kappa statistic in reliability studies: use, interpretation and sample size requirements. Phys Ther. 2005;85:257–268
    • View In Article
    • MEDLINE
19. Peris K, Altobelli E, Ferrari A, Fargnoli MC, Piccolo D, Esposito M, et al. Interobserver agreement on dermoscopic features of pigmented basal cell carcinoma. Dermatol Surg. 2002;28:643–645
    • View In Article
    • MEDLINE
    • CrossRef
20. Hoehler FK. Bias and prevalence effects on kappa viewed in terms of sensitivity and specificity. J Clin Epidemiol. 2000;53:499–503
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (117 KB)
    • CrossRef