Dermoscopy of Squamous Cell Carcinoma and Keratoacanthoma

Objectives : To characterize dermoscopic criteria of squamous cell carcinoma (SCC) and keratoacanthoma and to compare them with other lesions. Design : Observer-masked study of consecutive lesions performed from March 1 through December 31, 2011. Setting : Primary care skin cancer practice in Brisbane, Australia. Participants : A total of 186 patients with 206 lesions. Main Outcome Measures : Sensitivity, specificity, predictive values, and odds ratios.


D
2][3][4] It is a valuable tool for pigmented and nonpigmented skin lesions.Dermoscopic criteria have been described for pigmented and nonpigmented Bowen disease (intraepithelial carcinoma) [5][6][7][8][9] and actinic keratosis, 10,11 two superficial forms of squamous cell carcinoma (SCC).However, few studies [12][13][14][15][16][17][18] have reported dermoscopic features of invasive cutaneous SCC or keratoacanthoma.A recent study 19 focused on a progression model of actinic keratosis and intraepidermal carcinomas to invasive carcinoma but was limited to facial cases and did not contrast dermoscopic fea-tures of SCC with those of other malignant or benign nonpigmented neoplasms.The study aims to better characterize the dermoscopic features of SCC and keratoacanthoma and to explore dermoscopic criteria that differentiate them from each other and from other nonpigmented skin lesions.

METHODS
A set of 103 cases for retrospective analysis was collected in a single private primary care skin cancer practice in Queensland, Australia, from January 19, 2007,  cluded because the image was not available or was of poor quality.An additional 53 cases were selected from April 8, 2010, through February 21, 2011.They were selected on the basis of image quality and represented 37% of the SCCs and keratoacanthomas treated during that interval.
Another set of cases for prospective analysis comprised 206 unselected, consecutive, nonpigmented raised lesions treated from March 1 through December 31, 2011, in the same private practice.All cases were excised or biopsied, and histopathologic features were regarded as the diagnostic criterion standard.
Dermoscopic images were taken either with a DermLite Fluid (nonpolarizing) dermoscope (3Gen, LLC) coupled to a Canon EOS 50D camera (Canon USA, Inc) or a DermLite DL3 dermoscope (polarizing or nonpolarizing mode) coupled to an Olympus E-450 camera (Olympus Corporation).Most images were acquired using nonpolarizing dermoscopy.All images were obtained using contact dermoscopy with alcohol gel for immersion or, in the case of significantly raised lesions, ultrasound gel.Ethics approval was obtained for this study from the Medical Research Ethics Committee of The University of Queensland, Brisbane, Australia, and all participants gave their informed consent.

EVALUATION OF DERMOSCOPIC IMAGES
Dermoscopic images were presented on a computer screen.Three observers (C.R., A.C., and H.K.) evaluated the presence or absence of dermoscopic criteria in consensus.In the first part of the study that included only keratoacanthomas and SCCs, the diagnosis was revealed to the observers.We evaluated vascular patterns according to the method of Kittler et al. 20 In short, we first describe vessel morphologic features using objectively defined geometric terms and then describe any specific arrangement of vessels.Morphologically, vessels may appear as dots, clods, straight lines, curved lines, looped lines, serpentine lines, helical lines, or coiled lines.If one vessel type dominates by far, the vascular pattern is called monomorphic; if multiple types of vessels are present, it is termed polymorphic.Most commonly, vessels are distributed randomly throughout a lesion (ie, the vessels are not arranged specifically).Specific arrangements of vessels with diagnostic significance include radial, branched, clustered, centered, serpiginous, reticular, and linear.Only one specific arrangement of vessels is possible per lesion except for the combination of radial and branched vessels.
In addition to vascular patterns, we evaluated the presence of keratin, scale, ulceration, blood spots, and white structures.White structures were further differentiated into white lines, white dots, white clods, white circles, and white structureless zones according to the basic principles of pattern analysis.
In the second part of the study, the evaluation of dermoscopic images was performed without knowledge of the histopathologic diagnosis in unselected consecutive cases.The evaluations of 4 observers (H.K., G.A., I.Z., and a dermatology resident) were used to calculate the interobserver agreement for white circles.

STATISTICAL ANALYSIS
Continuous data are given as mean(SD) unless otherwise specified.The sensitivity, specificity, and positive and negative predictive values and their 95% CIs were calculated according to standard formulas.We used the 2 test or the Fisher exact test for the comparison of proportions and logistic regression for multivariate analysis.Variables were included in the multivariate logistic regression analysis if the P value of the univariate analysis was Ͻ.10.Because of multicollinearity, we dropped "vessel arrangement" from the logistic regression model in favor of the simpler criterion "vessel morphologic features."Fleiss was used as a measure of interobserver agreement.All given P values are 2-tailed, and P Ͻ .05 was considered statistically significant.

RETROSPECTIVE ANALYSIS
The retrospective analysis included 43 keratoacanthomas and 60 SCCs from 80 patients (mean [SD] age, 68 [13]  a Data are presented as number (percentage) of cases unless otherwise indicated.
b More than 1 type of vessel could be present per lesion.c More than 1 type of white structure could be present per lesion.
presence of central keratin was more frequent in keratoacanthomas than in SCCs (51.2% vs 30.0%,P = .03).White circles were observed in both types of lesions but more commonly in SCCs than in keratoacanthomas (60% vs 25.6%, P = .001).With regard to vascular patterns, we found that radial arrangement of branched vessels was more common in keratoacanthomas than in SCCs (20.9% vs 1.7%, P = .03).The frequencies of all other dermoscopic criteria did not differ significantly between the two diagnoses (Table 1).

ANALYSIS OF CONSECUTIVE CASES
The prospective analysis consisted of 206 consecutive cases of raised, nonpigmented lesions, including 29 keratoacanthomas, 32 SCCs, and 145 cases with other diagnoses from 186 patients (mean [SD] age, 65 [13] years; 32.8% women).The range of diagnoses is given in Table 2.In accordance with the results of the retrospective study, central keratin was more common in keratoacanthomas than in SCCs (72.4% vs 40.6%, P = .01).
In contrast to the retrospective part of the study, white circles were equally common in both types of lesions (41.4% in SCCs vs 46.9% in keratoacanthomas, P = .67).
Because of the similar appearance of both lesions on dermoscopy, they were lumped together and contrasted with the other lesions in the set (Table 3).
As indicated in Table 4, the specificities of dermoscopic criteria depended on the differential diagnosis and thus on the clinical context.When keratoacanthoma and SCC were contrasted with basal cell carcinoma, the positive predictive value of keratin was 92% and that of white circles was 89%.When SCC and keratoacanthoma were contrasted with actinic keratosis and Bowen disease, the positive predictive value of keratin was 50% and that of white circles was 92%.In a multivariate logistic regression model (adjusted for age and site) that contrasted keratoacanthoma and SCC with all other diagnoses, only white circles (odds ratio [OR], 6.1; 95% CI, 2.4-13.3;P Ͻ .001),keratin (OR, 3.6; 95% CI, 1.5-8.7;P = .005),and blood spots (OR, 2.6; 95% CI, 1.1-6.3;P = .03)were independent positive predictors of keratoacanthoma and SCC.In this model, the presence of serpentine vessels decreased the odds for keratoacanthoma and SCC (OR, 0.3; 95% CI, 0.1-0.8;P = .02)significantly and was more in favor of another diagnosis.
The interobserver agreement for the presence or absence of white circles on dermoscopy was good (0.55; 95% CI, 0.44-0.65).On the basis of dermoscopic-pathologic correlation, white circles correspond to acanthosis and hypergranulosis of infundibular epidermis.

COMMENT
We identified several dermoscopic criteria that help differentiate keratoacanthoma and SCC from other equivocal, nonpigmented lesions.The most important dermoscopic clues to keratoacanthoma and SCC are white circles (Figure 1), keratin, blood spots, and white structureless zones.These criteria were significantly and independently associated with keratoacanthoma and SCC.
Although patterns of blood vessels were not statistically significant in a multivariate analysis, this does not mean that they are completely insignificant for clinical diagnosis.We think it is a major finding of our study that the significance of dermoscopic criteria depends on the context (ie, the differential diagnosis).This is certainly true of vascular patterns.When the differential diagnosis is keratoacanthoma and SCC vs basal cell carcinoma, the presence of coiled vessels is a strong clue to keratoacanthoma and SCC.However, the presence of coiled vessels is not helpful when the differential diagnosis is keratoacanthoma and SCC vs actinic keratosis and Bowen disease because coiled vessels are commonly found in all these lesions.For this differential diagnosis, the dermoscopic clues of white circles, blood spots, and white structureless zones are more robust than vascular patterns.A general principle in the dermoscopy of pigmented lesions is that pigment is more specific than vessels.In addition, for raised, nonpigmented lesions, we can also say that white structures and keratin are more specific than vessels.It is not surprising that keratin turned out to be a clue to keratoacanthoma and SCC.Keratin located centrally was the only criterion that differentiated keratoacanthoma from SCC, at least to a certain extent (Figure 2).The presence of central keratin was more frequent in keratoacanthomas than in SCCs in the retrospective (51.2% vs 30.0%) and prospective parts (72.4% vs 40.6%) of the study.This finding is not surprising because a central keratin plug is an architectural criterion for the diagnosis of keratoacanthoma on histopathologic analysis.
White circles and branched and radial vessels differentiated between SCC and keratoacanthoma only in the retrospective but not in the prospective series, possibly reflecting the lower number of keratoacanthomas in the prospective study and a greater selection bias in the retrospective study.We must conclude from these data that keratoacanthoma and SCC cannot confidently be differentiated by dermoscopy.This is not surprising given that dermatopathologists still debate whether keratoacanthoma is a highly differentiated type of SCC or a benign neoplasm that is completely different from SCC. 21 Even if it is true that keratoacanthomas are a distinct benign entity, most keratoacanthomas should still be excised because only dermatopathologists can tell the difference from SCC with certainty.Because keratoacanthoma and SCC are treated alike in a practical clinical sense and their appearance at dermoscopy overlaps, we aggregated SCC and keratoacanthoma data and performed an analysis of the dermoscopic features that differentiate them from other raised, nonpigmented lesions.Keratin again proved to be a useful dermoscopic clue, having the highest sensitivity (79%) for keratoacanthoma and SCC among all criteria.However, the specificity and the positive predictive value of keratin depended on the context, in synchrony with our findings regarding the significance of vascular patterns.When keratoacanthoma and SCC were contrasted with basal cell carcinoma, the positive predictive value of keratin was 92%, but when SCC and keratoacanthoma were contrasted with actinic keratosis and Bowen disease, the positive predictive value of keratin was only 50%.This is no surprise because keratin is usually absent in basal cell carcinoma but may be prominent in actinic keratoses and Bowen disease.This finding underlines how selection of control participants can One of the major results of our study is the identification of white circles as a new dermoscopic criterion (Figure 1) and a valuable clue to keratoacanthoma and SCC.At 87%, white circles had the highest specificity of all criteria examined for distinguishing SCC and keratoacanthoma from all other diagnoses.It was a robust criterion that was valid not only for the differentiation of keratoacanthoma and SCC from basal cell carcinoma but also for the clinically important differentiation of keratoacanthoma and SCC from actinic keratoses and Bowen disease.When SCC and keratoacanthoma were contrasted with actinic keratosis and Bowen disease, the positive predictive value of white circles was high (92%).We, therefore, believe that the presence of white circles in a raised, nonpigmented lesion should lead to consideration for biopsy.
As shown by dermoscopic-pathologic correlation, white circles correspond to acanthosis and hypergranulosis of infundibular epidermis (Figure 3).This finding explains why the circle is centered around a dilated infundibulum filled with a keratin plug that is visible as a yellow or an orange clod on dermoscopy (Figure 1B-D).White circles are different from keratin pearls, which his-topathologically correspond to intraepidermal keratin whorls or horn pearls. 22The keratin pearls as described by Jaimes et al 22 probably correspond to what we have termed white clods in our study.
The major weakness of our study is the influence of confirmation bias because we included only lesions that were excised or biopsied.Confirmation bias is evidenced by the incidence of 16.9% of benign lesions, mostly seborrheic keratoses, with the dermoscopic clue of white circles.Of course, by the time the prospective study commenced, all lesions with white circles were being biopsied, regardless of overall pattern.Another potential limitation of our study is that most dermoscopic images were taken by nonpolarized, dermoscopy and our results should be generalized to polarized dermoscopy with caution.However, the main dermoscopic clues to SCC and keratoacanthoma (white circles, keratin, and blood spots) will be visible by polarized and nonpolarized dermoscopy.
We tested the interobserver variability for white circles because, to our knowledge, they have not been described as a dermoscopic criterion.As indicated by a of 0.55, the interobserver variability was good, and white circles can be viewed as a reliable and robust criterion when evaluated by experts.However, this high level of interobserver agreement may not be sustainable among observers who are not experts.We also identified blood spots as a further significant criterion for the diagnosis of keratoacanthoma and SCC by dermoscopy.As shown in Figure 2, blood spots in keratoacanthoma and SCC are preferentially found within keratin.However, their significance does not reach that of white circles and keratin.As indicated by the univariate analysis, white structureless zones may also be a useful clue to SCC and keratoacanthoma, especially when other criteria, such as keratin or white circles, are missing.Although dermoscopy is increasingly used for nonpigmented lesions, the description of criteria is still not as advanced as for pigmented lesions.We have provided new evidence that dermoscopy is helpful for nonpigmented lesions beyond the diagnosis of amelanotic melanoma.The diagnostic utility of white circles is a new dermoscopic finding.It is relevant and reliable and shows high specificity and a high positive predictive value in differentiating keratoacanthoma and SCC from other raised, nonpigmented lesions.

Figure 1 .
Figure 1.White circles in squamous cell carcinoma.Clinical (A and C) and corresponding dermoscopy (B and D) images of 2 invasive squamous cell carcinomas with prominent white circles on dermoscopy.The white circles are centered around a dilated infundibulum filled with a keratin plug that is visible as a yellow or an orange clod on dermoscopy.The clinical differential diagnoses include basal cell carcinoma and amelanotic melanoma and other benign and malignant nonpigmented skin lesions.

Figure 2 .
Figure 2. Central keratin and blood spots as dermoscopic clues to squamous cell carcinoma and keratoacanthoma.Clinical (A and C) and corresponding dermoscopy (B and D) images of 2 keratoacanthomas with central keratin on dermoscopy.The dermoscopy of the keratoacanthoma shown in B also shows white circles.Blood spots are found within keratin (B and D) and help to differentiate keratin from scale.

Figure 3 .
Figure 3. Dermoscopic-pathologic correlation of white circles.This micrograph shows the dermatopathology of the squamous cell carcinoma shown in Figure 1D.White circles correspond to acanthosis and hypergranulosis of infundibular epidermis, which explains the geometric expression as a circle and why the circle is centered around a dilated infundibulum filled with a keratin plug that is visible as a yellow or an orange clod on dermoscopy.
through February 21, 2011.Selection criteria were the histopathologic diagnosis of invasive SCC or keratoacanthoma.Fifty cases were selected from January 19 through December 31, 2007; another 19 cases treated during that same interval were ex-

Table 3 . General and Dermoscopic Characteristics According to Diagnosis (Prospective Analysis) a
a Data are presented as number (percentage) of cases unless otherwise indicated.bUnivariate P value derived from the 2 test.

Table 4 . Diagnostic Significance of Dermoscopic Criteria in Favor of SCC and KA According to Differential Diagnosis affect
the reported specificity of various dermoscopic criteria.