JEADV
DOI: 10.1111/jdv.12483
ORIGINAL ARTICLE
Dermatoscopy of flat pigmented facial lesions P. Tschandl,1 C. Rosendahl,2,* H. Kittler1 1
Department of Dermatology, Division of General Dermatology, Medical University of Vienna, Vienna, Austria School of Medicine, The University of Queensland, Australia *Correspondence: C. Rosendahl. E-mail: [email protected]
2
Abstract Background The diagnosis of flat pigmented lesions on the face is challenging because of the morphologic overlap of biologically different lesions and the unknown significance of dermatoscopic patterns. Objective To better characterize dermatoscopic patterns of various types of flat pigmented facial lesions and to analyse their significance by calculating their relative risks and diagnostic values. Methods We prospectively analysed dermatoscopic images of 240 flat pigmented facial skin lesions collected consecutively from 195 patients (41.5% females, mean age: 61 14 years) between 2007 and March 2012 in a primary skin cancer practice situated in Queensland, Australia. Results Histopathologically 114 (47.5%) lesions were malignant (24 lentigo maligna, 21 basal cell carcinomas and 69 pigmented actinic keratoses). Compared with all other patterns the positive predictive value for lentigo maligna was highest for a pattern of circles (31.3%, 95% CI: 11.1–58.6%). A pattern of clods was associated with basal cell carcinoma. If grey structures were present the relative risk for malignancy was 2.2 (95%CI: 1.4–3.4). The best clues to differentiate pigmented actinic keratosis from other lesions were the presence of scale (positive predictive value: 72.2%, specificity: 94.2%), white circles (positive predictive value: 68.8%, specificity: 94.2%) and a sharply demarcated border (positive predictive value: 44.2%, specificity: 86.0%). Conclusions In flat lesions a pattern of circles bears the highest risk for facial melanoma but other patterns do not exclude it. Scale, white circles and a sharply demarcated border are clues to pigmented actinic keratoses. The presence of grey colour is a clue to malignancy regardless of pattern. Received: 26 November 2013; Accepted: 20 February 2014
Conflicts of interests None declared.
Funding sources None declared.
Introduction Dermatoscopy is a widely accepted method for the early diagnosis of skin cancer and has been proven to be accurate for melanoma and epithelial neoplasms.1,2 Dermatoscopy of melanoma may vary depending on anatomic site and whether the lesion is flat or raised. Dermatoscopy characteristics of melanoma in situ on the face have been described before by Schiffner,3,4 Tanaka5 and Pralong.6 However, the knowledge about the significance of dermatoscopic patterns with regard to the differentiation of lentigo maligna (LM) from other flat pigmented facial lesions is limited. In particular, the differentiation of LM from facial epithelial neoplasms is mostly based on case reports.7 Akay et al., who published the largest series of consecutive cases, found that the value of dermatoscopy to differentiate between melanocytic lesions and pigmented actinic keratosis (pAK) is limited.8 Caucasian skin chronically exposed to sun is susceptible to
JEADV 2014
benign and malignant pigmented facial lesions. Pigmented facial lesions frequently present to physicians and are challenging diagnostically because the differential diagnosis includes a variety of benign and malignant conditions with similar clinical appearances. In many cases, diagnostic uncertainty is not resolved by clinical inspection or by dermatoscopy. This usually leads to biopsy or excision to rule out melanoma in situ. On the other hand, flat, pigmented, facial lesions are often treated by cryotherapy9 or other local treatment modalities,10 sometimes without a definite histopathologic diagnosis. This procedure bears the risk of inappropriate treatment of melanoma and delay in diagnosis. The significance of dermatoscopic patterns has not been analysed prospectively and no clear guidelines exist to help the physician to select facial lesions for additional in vivo examinations like confocal microscopy, biopsy, local treatment or monitoring. Some dermatoscopic clues like asymmetric
© 2014 European Academy of Dermatology and Venereology
Tschandl et al.
2
pigmented follicular openings, rhomboids and circles within a circle that are used to differentiate LM from other flat pigmented lesions have not been tested prospectively and their specificity is unknown. In this study, we sought to identify dermatoscopic structures aiding in diagnosis and management of pigmented skin lesions on the face by blinded evaluation of a consecutive series of dermatoscopic images.
Methods We included dermatoscopically documented pigmented skin lesions on the face excised between January 2007 and March 2012 in a primary skin cancer clinic situated in Queensland, Australia. We excluded lesions without unequivocal histopathologic reports as well as raised lesions. The gold standard for diagnosis was the histopathologic report. All dermatoscopic images before October 2010 were taken with a Dermlite Fluid (non-polarized) dermatoscope and after that date with the same device and also in most cases with a Dermlite DL3 (3Gen, San Juan Capistrano, CA, USA) dermatoscope in both polarized and non-polarized modes. Always, some kind of immersion fluid, either 70% ethanol handwash gel, or ultrasound gel, was used when taking the photographs. Images were assessed digitally by author PT with no time limit for predefined patterns (Figs 1,2) and structures (Table S1) based on modified pattern analysis.11 A pattern consists of multiple basic elements (lines, dots, clods, circles or pseudopods) and has to take up a substantial part (20%) of the lesion, whereas ‘clues’ can be present just once, e.g. a single grey circle. If more than one pattern was present, we classified the lesion according to the most specific pattern. Patterns from most
(a)
(b)
(c)
(d)
specific to least specific were as follows: lines (reticular/ branched/radial/curved), circles, clods, dots and structureless. We added the following additional structures not previously described in modified pattern analysis (Fig. 3): ‘Four-dot clods’ are defined as four white dots, arranged in a square (Fig. 3a), this structure has been mentioned in the literature before as ‘rosettes’.12 A ‘double circle’6 was defined as a grey circle within another circle (Fig. 3b). Incomplete circles, also known as asymmetric pigmented follicular openings (APFO),3 were defined as brown or grey semicircles. A sharply demarcated scalloped border (‘moth-eaten border’) was defined as convex outline with abrupt cessation of pigmentation, white circles (Fig. 3c) were assessed as described before.13. ‘Rhomboids’ (Fig. 3c,d) are interfollicular lines that form a polygon (most commonly a rhomboid). Interfollicular reticular lines are reticular lines like that seen on non-facial skin. ‘Scale’ (Fig. 3d) was evaluated from the dermatoscopic, not from the clinical, image after application of fluid or gel. All features were treated as binary values (present or absent). Ethics statement
Ethics approval for this study was granted by the Ethics Committee of The University of Queensland, Australia. Statistical analysis
Differences in the presence of dermatoscopic features between groups were evaluated with the chi-squared test or Fisher’s exact test as appropriate. A logistic regression model with forward conditional entering of variables was used for the multivariate analysis. Sensitivity, specificity, positive and negative predictive values and their 95% confidence intervals were calculated
Figure 1 Sketches of basic dermatoscopic patterns as assessed within this study, namely, lines reticular, branched or curved (a, curved), circles (b), dots (c) and structureless (d).
JEADV 2014
© 2014 European Academy of Dermatology and Venereology
Dermatoscopy of flat pigmented facial lesions
3
(a)
(b)
(c)
(d)
(e)
(f)
Figure 2 Examples of dermatoscopic patterns as assessed in this study (representative examples of structures referred to are indicated by arrows): Interfollicular reticular lines (a; solar lentigo), grey circles (b; lentigo maligna), brown circles (c; lentigo maligna), grey dots (d; solar lentigo), clods (e; dermal naevus) and structureless (f; seborrhoeic keratosis).
according to standard formula. All statistical calculations were performed with IBMâ SPSSâ Statistics 20 software (IBM Corp., New York, NY, USA). A P-value < 0.05 was regarded statistically significant. ROC curves were calculated and fitted with a semiparametric estimation approach with proper binormal curve modelling using the ‘ROC-KIT’ software provided by the Department of Radiology of the University of Chicago (URL: http://metz-roc.uchicago.edu/MetzROC/software).
lesions, 50 (20.8%) were small in diameter (2 cm). All melanomas except one (Breslow 0.3 mm) were in situ. About a third of the lesions occurred on the cheek (32.3%), second most on the forehead (29.7%), followed by the nose (14.4%), malar region (11.4%), ear (9.6%) and chin and eyelid (1.3% respectively). Frequency and significance of patterns
Results General data
A total of 240 consecutive lesions (195 patients; 41.5% females, mean age: 61 14 years) were collected including 114 (47.5%) malignant (24 lentigo maligna (LM), 21 basal cell carcinomas (BCC) and 69 pigmented actinic keratosis or pigmented Bowens disease (pAK/pBD)) and 126 (52.5%) benign lesions (33 melanocytic nevi, 71 solar lentigines/seborrhoeic keratoses (SL/SebK), 19 lichen planus like keratoses (LPLK), three other skin lesions (vascular, inflammatory)). Of the 240
JEADV 2014
Frequencies and positive predictive values of patterns for different diagnoses are given in Table 1. A pattern of lines was the most common pattern (33%, n = 78) and SL/flat SebK was the most common diagnosis when this pattern was present (n = 35, 44.9%). A pattern of circles was found in 7% (n = 16) of lesions. More than two thirds of lesions with a pattern dominated by circles were malignant on histopathology (68.8%, n = 11), either pAK (n = 6) or LM (n = 5). Among benign lesions, the highest positive predictive value of circles was found for SL/SebK (PPV: 25.0%, 95% CI: 7.4–52.4%).
© 2014 European Academy of Dermatology and Venereology
Tschandl et al.
4
(a)
(c)
(b)
(d)
A pattern of clods was found in 8% (n = 20) of lesions and had the highest positive predictive value and specificity for BCCs (PPV 55.0%, 95% CI: 31.6–76.9%; specificity 95.9%, 95% CI: 92.3–98.1%). A pattern of dots occurred in 30% (n = 73) of all lesions with a similar frequency among all diagnoses (Table 1). The highest PPV and the highest specificity were found for pAK (PPV 37.0%, 95% CI: 26.0–49.1%,specificity73.1%,95%CI:65.8–79.6%). The structureless pattern was present in 22% (n = 53) of lesions, of which 43% (n = 23) were malignant. The highest PPV and the highest specificity were found for pAK (PPV 37.7%, 95% CI: 24.8–52.1%, specificity 80.7%, 95% CI: 74.0– 86.3%). Clues to malignant lesions (Table 2) Melanoma/Lentigo maligna (Fig. 4) Grey structures (grey dots, clods, circles or lines) were present in 95.8% (n = 23) of LM. The relative risk for LM when a grey structure was present was 8.9 (95% CI: 1.2–64.7). In 18.4% of lesions with incomplete circles (asymmetric pigmented follicular openings) the diagnosis was melanoma. The sensitivity and specificity for diagnosing LM were 58.3% and 71.3% respectively. Grey circles were present in 54.2% of LM, their PPV was 26.5% (95%CI: 15.0–41.1%) with a specificity of 83.3% respectively. The clue of a ‘circle within a circle’ was only found in one LM and therefore had a low sensitivity (4.2%). Rhomboids had a high specificity (91.7%) but their PPV was low (18.2%). Basal cell carcinoma Grey structures were found in 95.2% (n = 20) of BCCs. Dermatoscopic clues almost exclusively found
JEADV 2014
Figure 3 Clues for facial pigmented lesions not previously described in modified pattern analysis (representative examples of structures referred to are indicated by arrows): (a) Dermatoscopic image of a pigmented actinic keratosis with ‘four-dot clods’; (b) The dermatoscopic clue of ‘double circles’, histopathology revealed a seborrhoeic keratosis; (c) Dermatoscopic example of ‘rhomboids’ and ‘white circles’ in a pigmented actinic keratosis; (d) ‘Scale’(right arrow) as seen by dermatoscopy after appliance of immersion fluid on a pigmented actinic keratosis and ‘rhomboids’(left arrow).
in BCC were branched vessels [38.1%, n = 8, PPV 100.0% (95% CI: 62.9–100.0%), specificity 100% (95% CI: 98.3–100.0%)], serpentine vessels [(23.8%, n = 5, PPV 83.3% (95% CI: 36.1– 97.2%), specificity 99.5% (95% CI: 97.5–99.9%)] and ulceration [38.1%, n = 8, PPV 66.7% (95% CI: 35.0–89.9%), specificity 98.2 (95% CI: 95.4–99.5%)]. Pigmented actinic keratosis/pigmented Bowen’s disease (pAK/ pBD) Grey structures were also common in pAK/pBD (78.3%).
Incomplete circles were found in 43.5% (n = 30) of cases. The most specific clues were scale [PPV 72.2% (95% CI: 54.8– 85.8%) and specificity 94.2% (95% CI: 89.5–97.2%)], white circles [PPV 68.8% (95% CI: 50.0–83.9%) and specificity 94.2% (95% CI: 89.5–97.2%)] and four-dot clods [PPV 60.0% (95% CI: 15.4–93.5%) and specificity 98.8% (95% CI: 95.8–99.8%)]. Rhomboids had a high specificity [94.7% (95% CI: 90.2– 97.6%)] but a low PPV [59.1% (95% CI: 36.4–79.3%)]. Clues to benign lesions
Melanocytic nevus. The best clue for nevi was the presence of interfollicular reticular lines [PPV 18.1% (95% CI: 10.0–28.9%) and specificity 71.5 (95% CI: 64.8–77.5%)] which were present in 39.4% (n = 13) of all nevi. Grey structures were less common than in malignant lesions (66.7%, n = 22). Solar lentigo/seborrhoeic keratosis (SL/SebK) Interfollicular reticular lines were also common in SL/SebK. They occurred in 39.4% of SL/SebK. Their PPV was 38.9% (95% CI: 27.6–51.1%) and the specificity was 74.0% (95% CI: 66.7–80.4%). Incomplete circles showed a similar PPV (34.2%) and specificity (70.4%). A sharply demarcated border was found in 23.9% of lesions (PPV
© 2014 European Academy of Dermatology and Venereology
5
39.5% (95% CI: 25.0–55.6%), specificity 84.6 (95% CI: 78.3– 89.7%)). White dots/clods (‘milia’) and orange clods (‘comedo like openings’) were present in 15.5% (n = 11) of cases. Compared with other lesions on the face, grey structures were less common in SL/SebK (49.3%, n = 35).
LPLK, Lichen planus-like keratosis; PAK/pBD, pigmented actinic keratosis/pigmented Bowen’s disease; SL/SebK, Solar lentigo/seborrhoeic keratosis.
n.a.
13.7% (6.8–23.8%)
39.6% (26.5–54.0%) 37.7% (24.8–52.1%) 1.9% (0.3–10.1%) Structureless (n = 53)
3.8% (0.6–13.0%)
13.2% (5.5–25.4%)
n.a.
12.3% (5.8–22.1%) 15.1% (7.8–25.4%) 37.0% (26.0–.49.1%) 9.6% (4.0–18.8%) Dots (n = 73)
11.0% (4.9–20.5%)
10.0% (1.5–31.7%)) 30.0% (12.0–54.3%)
6.3% (1.0–30.3%) 25.0% (7.4–52.4%) n.a.
5.0% (0.8–25.0%)
37.5% (15.3–64.5%)) n.a.
55.0% (31.6–76.9%) n.a.
31.3%(11.1–58.6%) Circles (n = 16)
Clods (n = 20)
PPV for LPLK PPV for SL/SebK
44.9% (33.6–56.6%)
PPV for Naevus
11.5% (5.4–20.8%)
PPV for pAK/pBD
19.2% (11.2–29.7%)
PPV for basal cell carcinoma
14.1% (7.3–23.8%) Lines (n = 78)
n.a.
PPV for melanoma Pattern
Table 1 Positive predictive values and their 95% confidence intervals of dermatoscopic patterns according to different diagnoses JEADV 2014
10.3% (4.6–19.2%)
Dermatoscopy of flat pigmented facial lesions
Lichen planus-like keratosis (LPLK) Lichen planus-like keratoses on the face are dermatoscopically characterized by a pattern of dots [52.6%, n = 10, PPV 13.7% (95% CI: 6.8–23.8%), specificity 71.5% (95% CI: 65.1–77.4%)] or lines [42.1%, n = 8, PPV 10.3% (95% CI: 4.6–19.2%), specificity 68.3% (95% CI: 61.8– 74.4%)]. A sharply demarcated border was found in 26.3% (n = 5). Clues that are associated with malignant lesions were also common in LPLK: incomplete circles were found in 21.1% (n = 4) and grey structures in 94.7% (n = 18). Diagnostic models Lentigo maligna vs. pAK/pBD Since LM and facial pAK/pBD may have similar patterns on dermatoscopy, we built a multivariate diagnostic model including the most specific clues for each diagnosis. In the multivariate logistic model grey circles were associated with melanoma (odds ratio: 5.99, 95% CI: 1.27–28.18, P = 0.02). White circles (odds ratio for melanoma = 0.07, 95% CI: 0.01–0.70, P = 0.02), scale (odds ratio for melanoma = 0.09, 95% CI: 0.01–0.73, P = 0.02) and sharply demarcated border (odds ratio for melanoma = 0.09, 95% CI: 0.01–0.78, P = 0.03) were associated with pAK/pBD. Based on this model we created a simple score that adds one point if grey circles are present and subtracts one point if scale, white circles or a sharply demarcated border is present. The higher the score the higher is the risk for melanoma the lower the score the higher is the risk for pAK/ pBD. The diagnostic accuracy of this score expressed as area under the receiver operating characteristics curve was 0.87 (95% CI: 0.79–0.94). Lentigo maligna vs. solar lentigo/SebK/LPLK In a multivariate model including the most specific clues for LM and SL/SebK/ LPLK, the presence of any grey structure was significantly associated with melanoma (odds ratio: 16.13, 95% CI: 2.97–125.92, P = 0.008). A sharply demarcated scalloped border was in favour of SL/SebK/LPLK (odds ratio: 0.13, 95% CI: 0.02–1.08), but did not reach statistical significance (P = 0.06). If the presence of any grey structure was used as the sole criterion to differentiate melanoma from SL/SebK/LPLK the sensitivity was 95.8% and the specificity 41.1%. If LPLKs, which accounted for most false-positive cases, were excluded, the specificity of ‘any grey structure’ rose to 50.7%. Differentiation of benign from malignant lesions If the presence of any grey structure was used as the sole criterion to differentiate malignant lesions (melanoma, pAK/pBD and BCC) from
© 2014 European Academy of Dermatology and Venereology
Tschandl et al.
6
Table 2 Diagnostic indices of dermatoscopic clues to flat malignant facial lesions. Numbers in braces depict 95% confidence intervals RR
PPV
NPV
Sensitivity
Specificity
Melanoma Any grey structure
8.9, 95% CI: 1.2–64.7
13.3% (8.6–19.3%)
98.5% (91.9–99.8%)
95.8% (78.8–99.3%)
30.6% (24.5–37.2%)
Vessels as dots
3.5, 95% CI: 1.1–11.8
33.3% (5.3–77.3%)
90.6% (86.1–94.0%)
8.3% (1.3–27.0%)
98.1% (95.3–99.5%)
Incomplete circles
3.0, 95% CI: 1.4–6.5
18.4% (10.5–29.0%)
93.9% (89.1–97.0%)
58.3% (36.7–77.9%)
71.3% (64.8–77.2%)
Grey circles
4.6, 95% CI: 2.2–9.7
26.5% (15.0–41.1%)
94.2% (89.9–97.1%)
54.2% (32.8–74.4%)
83.3% (77.7–88.0%)
Rhomboids
2.0, 95% CI: 0.7–5.3
18.2% (5.3–40.3%)
90.8% (86.2–94.3%)
16.7% (4.8–37.4%)
91.7% (87.2–95.0)
Circle in a circle/Double circle
2.0, 95% CI: 0.3–12.3
20.0% (3.3–71.2%)
90.2% (85.7–93.7%)
4.2% (0.7–21.2%)
98.1% (95.3–99.5%)
Basal cell carcinoma Any grey structure
7.7, 95% CI: 1.1–56.6
11.6% (7.2–17.3%)
98.5% (91.9–99.8%)
95.2% (76.1–99.2%)
30.1% (24.1–36.7%)
Branched vessels
17.8, 95% CI: 10.5–30.3
100.0% (62.9–100.0%)
94.4% (90.6–97.0%)
38.1% (18.1–61.6%)
100.0% (98.3–100.0%)
Serpentine vessels
12.2, 95% CI: 6.7–22.1
83.3% (36.1–97.2%)
93.2% (89.1–96.0%)
23.8% (8.3–47.2%)
99.5% (97.5–99.9%)
Ulceration
11.7, 95% CI: 6.0–22.7
66.7% (35.0–89.9%)
94.3% (90.5–96.9%)
38.1% (18.2–61.6%)
98.2% (95.4–99.5%)
Pigmented actinic keratosis/Bowen’s disease White circles
3.0, 95% CI: 2.2–4.3
68.8% (50.0–83.9%)
77.4% (71.1–82.9%)
31.9% (21.2–44.2%)
94.2% (89.5–97.2%)
Sharply demarcated border
1.7, 95% CI: 1.2–2.6
44.2% (29.1–60.1%)
74.6% (68.0–80.5%)
27.5% (17.5–39.6%)
86.0% (79.8–90.8%)
Incomplete circles
1.7, 95% CI: 1.1–2.5
39.5% (28.5–51.4%)
76.2% (69.0–82.5%)
43.5% (31.6–56.0%)
73.1% (65.8–79.6%)
Rhomboids
2.3, 95% CI: 1.5–3.5
59.1 (36.4–79.3%)
74.3% (68.0–80.0%)
18.8% (10.4–30.1%)
94.7% (90.2–97.6%)
Four-dot clod
2.1, 95% CI: 1.0–4.5
60.0% (15.4–93.5%)
71.9% (65.7–77.6%)
4.3% (1.0–12.2%)
98.8% (95.8–99.8%)
Scale
3.4, 95% CI: 2.5–4.8
72.2% (54.8–85.8%)
78.9% (72.7–84.3%)
37.7% (26.3–50.2%)
94.2% (89.5–97.2%)
NPV, negative predictive value; PPV, positive predictive value; RR, relative risk.
(a)
(b)
Figure 4 Dermatoscopic structures predicting melanoma (representative examples of structures referred to are indicated by arrows): Dermatoscopic images show flat pigmented melanomas on the face with the feature of ‘grey structures’ (a) and ‘grey circles’ (b).
benign lesions the sensitivity was 85.1% and the specificity 39.7%.
Discussion The major differences between this and other studies of the dermatoscopy of pigmented lesions on the face are that we included only flat lesions and that we did not select the lesions based on their diagnosis. Another important difference is that our sample of facial melanomas includes mainly small and early lesions of LM while advanced cases have not been observed. This mirrors the situation in Australia where melanomas are recognized at an early stage. The focus on early lesions explains why the frequencies of classic criteria of facial melanomas such as
JEADV 2014
rhomboids3 or circle within a circle6 was much lower in our sample than in other studies. In our series the dermatoscopic appearance of melanoma on the face was contrasted with a variety of other pigmented lesions on chronic sun-damaged skin such as pAK, SL, LPLK, nevi and BCCs. We found that dermatoscopic criteria that have been reported to be specific for facial melanoma are not as specific as previously thought. Asymmetric pigmented follicular openings3,4,14 (incomplete circles) are not very specific for melanoma. They can be found in a significant portion of solar lentigines and pigmented actinic keratoses. Due to the prospective nature our study (lesions were not selected because of the diagnosis) we focused on the practical significance of dermatoscopic patterns and clues and calculated
© 2014 European Academy of Dermatology and Venereology
Dermatoscopy of flat pigmented facial lesions
positive and negative predictive values taking into account disease prevalence. We demonstrated that it is important to differentiate between different types of dermatoscopic patterns on the face because patterns are to some extent associated with specific diagnoses. We do not recommend the use of the term ‘pseudo-pigmentnetwork’ because it lumps together three different types of patterns: (i) circles; (ii) structureless and (iii) dots (Figs 1,2). Although these three patterns are not 100% specific they are of some significance. A pattern of clods (‘globules’) in a flat pigmented lesion is highly predictive for BCC especially if other clues are present too. Although other patterns are less specific it is useful to recognize them because some diagnoses become more likely or unlikely depending on the pattern observed. Although a pattern of reticular or curved lines does not exclude any diagnosis except BCC, it most likely indicates a SL especially if other criteria of SL, such as a sharply demarcated border, are present, and clues of melanoma or pAK are missing. With the exception of a pattern of clods, early facial melanoma may present with any pattern, but a pattern of pigmented circles is more likely than others. According to our analysis, a pattern of pigmented circles is the most common pattern in early LM and in our multivariate logistic model, we found that grey circles had a high association with melanoma while white circles, scale and a sharply demarcated border were more in favour pAK/pBD. It has to be emphasized, however, that a pattern of pigmented circles may not be the most common pattern in advanced cases of lentigo maligna. In advanced cases, when pigmentation has become more prominent, the interfollicular space will get pigmented and a structureless pattern will prevail, which is in line with the progression model of Stolz et al.3,4 On the other hand, we have shown that a pattern of circles can be found in other diagnoses too, chief among them being pAK, and thus is not highly specific. We confirmed that pAK is the great masquerader, it is not characterized by one specific pattern: Circles, dots and structureless patterns are equally likely. This is the reason why LM and pAK/pBD may appear with the same pattern. The differential diagnosis is difficult because even some dermatoscopic clues such as rhomboids and incomplete asymmetric circles can be found in both. Akay et al.8 were the first to notice that LM had striking dermatoscopic similarities to pAK. In our series we found that white circles, scale and a sharply demarcated border are suggestive of pAK, while grey circles suggest melanoma. Four-dot clods (‘rosettes’) were not only found in pAK/pBD but also in one melanoma and one BCC. This structure might not be as specific for pAK as previously thought. It has to emphasized that most dermatoscopic images were taken with non-polarized dermatoscopes. Because four-dot clods are mainly visible with polarized light, we might have underestimated the presence of this structure. Rhomboids, previously described to
JEADV 2014
7
be diagnostic of LM,3 were not only found in melanomas but also in pAK/pBD. In this series of consecutively collected flat and pigmented skin lesions, about half were malignant. This represents the selection bias when specimens are selected in routine practice as requiring biopsy (such a selection bias is unavoidable when histology is required for the analysis). Predictive values and diagnostic measures therefore are biased and even small correlations with malignant nature may represent a strong correlation in daily practice with unselected lesions. The overall high percentage of ‘any grey structure’ in this series may be explained by the fact that the author providing the cases (and also the other authors) sees the presence of grey colour in lesions on the face as a strong clue to a malignant process, if no confident benign diagnosis can be made otherwise. Even through this bias the feature ‘any grey structure’ remained an independent significant coefficient in multivariate analysis, underlining the importance of this feature. A high percentage of LPLKs also shows grey structures because of its regressive nature. By relying on ‘grey structures’ as a predictor of a malignant process, more of those lesions may be excised for diagnostic reasons, which, in the eyes of the authors, is necessary to not miss a potentially deadly melanoma.
References 1 Kittler H, Pehamberger H, Wolff K, Binder M. Diagnostic accuracy of dermoscopy. Lancet Oncol 2002; 3: 159–165. 2 Rosendahl C, Tschandl P, Cameron A, Kittler H. Diagnostic accuracy of dermatoscopy for melanocytic and nonmelanocytic pigmented lesions. J Am Acad Dermatol 2011; 64: 1068–1073. 3 Schiffner R, Schiffner-Rohe J, Vogt T et al. Improvement of early recognition of lentigo maligna using dermatoscopy. J Am Acad Dermatol 2000; 42: 25–32. 4 Stolz W, Schiffner R, Burgdorf WH. Dermatoscopy for facial pigmented skin lesions. Clin Dermatol 2002; 20: 276–278. 5 Tanaka M, Sawada M, Kobayashi K. Key points in dermoscopic differentiation between lentigo maligna and solar lentigo. J Dermatol 2011; 38: 53–58. 6 Pralong P, Bathelier E, Dalle S, Poulalhon N, Debarbieux S, Thomas L. Dermoscopy of lentigo maligna melanoma: report of 125 cases. Br J Dermatol 2012; 167: 280–287. 7 Pock L, Drlik L, Hercogova J. Dermatoscopy of pigmented actinic keratosis–a striking similarity to lentigo maligna. Int J Dermatol 2007; 46: 414– 416. 8 Akay BN, Kocyigit P, Heper AO, Erdem C. Dermatoscopy of flat pigmented facial lesions: diagnostic challenge between pigmented actinic keratosis and lentigo maligna. Br J Dermatol 2010; 163: 1212–1217. 9 Mahendran RM, Newton-Bishop JA. Survey of U.K. current practice in the treatment of lentigo maligna. Br J Dermatol 2001; 144: 71–76. 10 Wong JG, Toole JW, Demers AA, Musto G, Wiseman MC. Topical 5% imiquimod in the treatment of lentigo maligna. J Cutan Med Surg 2012; 16: 245–249. 11 Kittler H. Dermatoscopy: introduction of a new algorithmic method based on pattern analysis for diagnosis of pigmented skin lesions. Dermatopathol Pract Concept 2007; 13: 3. 12 Zalaudek I, Giacomel J, Schmid K et al. Dermatoscopy of facial actinic keratosis, intraepidermal carcinoma, and invasive squamous cell carcinoma: a progression model. J Am Acad Dermatol 2012; 66: 589–597.
© 2014 European Academy of Dermatology and Venereology
Tschandl et al.
8
13 Rosendahl C, Cameron A, Argenziano G, Zalaudek I, Tschandl P, Kittler H. Dermoscopy of squamous cell carcinoma and keratoacanthoma. Arch Dermatol 2012; 148: 1386–1392. 14 Sahin MT, Ozturkcan S, Ermertcan AT, Gunes AT. A comparison of dermoscopic features among lentigo senilis/initial seborrheic keratosis, seborrheic keratosis, lentigo maligna and lentigo maligna melanoma on the face. J Dermatol 2004; 31: 884–889.
JEADV 2014
Supporting information Additional Supporting Information may be found in the online version of this article: Table S1. Predefined dermatoscopic features assessed within this study.
© 2014 European Academy of Dermatology and Venereology
DOI: 10.1111/jdv.12483
ORIGINAL ARTICLE
Dermatoscopy of flat pigmented facial lesions P. Tschandl,1 C. Rosendahl,2,* H. Kittler1 1
Department of Dermatology, Division of General Dermatology, Medical University of Vienna, Vienna, Austria School of Medicine, The University of Queensland, Australia *Correspondence: C. Rosendahl. E-mail: [email protected]
2
Abstract Background The diagnosis of flat pigmented lesions on the face is challenging because of the morphologic overlap of biologically different lesions and the unknown significance of dermatoscopic patterns. Objective To better characterize dermatoscopic patterns of various types of flat pigmented facial lesions and to analyse their significance by calculating their relative risks and diagnostic values. Methods We prospectively analysed dermatoscopic images of 240 flat pigmented facial skin lesions collected consecutively from 195 patients (41.5% females, mean age: 61 14 years) between 2007 and March 2012 in a primary skin cancer practice situated in Queensland, Australia. Results Histopathologically 114 (47.5%) lesions were malignant (24 lentigo maligna, 21 basal cell carcinomas and 69 pigmented actinic keratoses). Compared with all other patterns the positive predictive value for lentigo maligna was highest for a pattern of circles (31.3%, 95% CI: 11.1–58.6%). A pattern of clods was associated with basal cell carcinoma. If grey structures were present the relative risk for malignancy was 2.2 (95%CI: 1.4–3.4). The best clues to differentiate pigmented actinic keratosis from other lesions were the presence of scale (positive predictive value: 72.2%, specificity: 94.2%), white circles (positive predictive value: 68.8%, specificity: 94.2%) and a sharply demarcated border (positive predictive value: 44.2%, specificity: 86.0%). Conclusions In flat lesions a pattern of circles bears the highest risk for facial melanoma but other patterns do not exclude it. Scale, white circles and a sharply demarcated border are clues to pigmented actinic keratoses. The presence of grey colour is a clue to malignancy regardless of pattern. Received: 26 November 2013; Accepted: 20 February 2014
Conflicts of interests None declared.
Funding sources None declared.
Introduction Dermatoscopy is a widely accepted method for the early diagnosis of skin cancer and has been proven to be accurate for melanoma and epithelial neoplasms.1,2 Dermatoscopy of melanoma may vary depending on anatomic site and whether the lesion is flat or raised. Dermatoscopy characteristics of melanoma in situ on the face have been described before by Schiffner,3,4 Tanaka5 and Pralong.6 However, the knowledge about the significance of dermatoscopic patterns with regard to the differentiation of lentigo maligna (LM) from other flat pigmented facial lesions is limited. In particular, the differentiation of LM from facial epithelial neoplasms is mostly based on case reports.7 Akay et al., who published the largest series of consecutive cases, found that the value of dermatoscopy to differentiate between melanocytic lesions and pigmented actinic keratosis (pAK) is limited.8 Caucasian skin chronically exposed to sun is susceptible to
JEADV 2014
benign and malignant pigmented facial lesions. Pigmented facial lesions frequently present to physicians and are challenging diagnostically because the differential diagnosis includes a variety of benign and malignant conditions with similar clinical appearances. In many cases, diagnostic uncertainty is not resolved by clinical inspection or by dermatoscopy. This usually leads to biopsy or excision to rule out melanoma in situ. On the other hand, flat, pigmented, facial lesions are often treated by cryotherapy9 or other local treatment modalities,10 sometimes without a definite histopathologic diagnosis. This procedure bears the risk of inappropriate treatment of melanoma and delay in diagnosis. The significance of dermatoscopic patterns has not been analysed prospectively and no clear guidelines exist to help the physician to select facial lesions for additional in vivo examinations like confocal microscopy, biopsy, local treatment or monitoring. Some dermatoscopic clues like asymmetric
© 2014 European Academy of Dermatology and Venereology
Tschandl et al.
2
pigmented follicular openings, rhomboids and circles within a circle that are used to differentiate LM from other flat pigmented lesions have not been tested prospectively and their specificity is unknown. In this study, we sought to identify dermatoscopic structures aiding in diagnosis and management of pigmented skin lesions on the face by blinded evaluation of a consecutive series of dermatoscopic images.
Methods We included dermatoscopically documented pigmented skin lesions on the face excised between January 2007 and March 2012 in a primary skin cancer clinic situated in Queensland, Australia. We excluded lesions without unequivocal histopathologic reports as well as raised lesions. The gold standard for diagnosis was the histopathologic report. All dermatoscopic images before October 2010 were taken with a Dermlite Fluid (non-polarized) dermatoscope and after that date with the same device and also in most cases with a Dermlite DL3 (3Gen, San Juan Capistrano, CA, USA) dermatoscope in both polarized and non-polarized modes. Always, some kind of immersion fluid, either 70% ethanol handwash gel, or ultrasound gel, was used when taking the photographs. Images were assessed digitally by author PT with no time limit for predefined patterns (Figs 1,2) and structures (Table S1) based on modified pattern analysis.11 A pattern consists of multiple basic elements (lines, dots, clods, circles or pseudopods) and has to take up a substantial part (20%) of the lesion, whereas ‘clues’ can be present just once, e.g. a single grey circle. If more than one pattern was present, we classified the lesion according to the most specific pattern. Patterns from most
(a)
(b)
(c)
(d)
specific to least specific were as follows: lines (reticular/ branched/radial/curved), circles, clods, dots and structureless. We added the following additional structures not previously described in modified pattern analysis (Fig. 3): ‘Four-dot clods’ are defined as four white dots, arranged in a square (Fig. 3a), this structure has been mentioned in the literature before as ‘rosettes’.12 A ‘double circle’6 was defined as a grey circle within another circle (Fig. 3b). Incomplete circles, also known as asymmetric pigmented follicular openings (APFO),3 were defined as brown or grey semicircles. A sharply demarcated scalloped border (‘moth-eaten border’) was defined as convex outline with abrupt cessation of pigmentation, white circles (Fig. 3c) were assessed as described before.13. ‘Rhomboids’ (Fig. 3c,d) are interfollicular lines that form a polygon (most commonly a rhomboid). Interfollicular reticular lines are reticular lines like that seen on non-facial skin. ‘Scale’ (Fig. 3d) was evaluated from the dermatoscopic, not from the clinical, image after application of fluid or gel. All features were treated as binary values (present or absent). Ethics statement
Ethics approval for this study was granted by the Ethics Committee of The University of Queensland, Australia. Statistical analysis
Differences in the presence of dermatoscopic features between groups were evaluated with the chi-squared test or Fisher’s exact test as appropriate. A logistic regression model with forward conditional entering of variables was used for the multivariate analysis. Sensitivity, specificity, positive and negative predictive values and their 95% confidence intervals were calculated
Figure 1 Sketches of basic dermatoscopic patterns as assessed within this study, namely, lines reticular, branched or curved (a, curved), circles (b), dots (c) and structureless (d).
JEADV 2014
© 2014 European Academy of Dermatology and Venereology
Dermatoscopy of flat pigmented facial lesions
3
(a)
(b)
(c)
(d)
(e)
(f)
Figure 2 Examples of dermatoscopic patterns as assessed in this study (representative examples of structures referred to are indicated by arrows): Interfollicular reticular lines (a; solar lentigo), grey circles (b; lentigo maligna), brown circles (c; lentigo maligna), grey dots (d; solar lentigo), clods (e; dermal naevus) and structureless (f; seborrhoeic keratosis).
according to standard formula. All statistical calculations were performed with IBMâ SPSSâ Statistics 20 software (IBM Corp., New York, NY, USA). A P-value < 0.05 was regarded statistically significant. ROC curves were calculated and fitted with a semiparametric estimation approach with proper binormal curve modelling using the ‘ROC-KIT’ software provided by the Department of Radiology of the University of Chicago (URL: http://metz-roc.uchicago.edu/MetzROC/software).
lesions, 50 (20.8%) were small in diameter (2 cm). All melanomas except one (Breslow 0.3 mm) were in situ. About a third of the lesions occurred on the cheek (32.3%), second most on the forehead (29.7%), followed by the nose (14.4%), malar region (11.4%), ear (9.6%) and chin and eyelid (1.3% respectively). Frequency and significance of patterns
Results General data
A total of 240 consecutive lesions (195 patients; 41.5% females, mean age: 61 14 years) were collected including 114 (47.5%) malignant (24 lentigo maligna (LM), 21 basal cell carcinomas (BCC) and 69 pigmented actinic keratosis or pigmented Bowens disease (pAK/pBD)) and 126 (52.5%) benign lesions (33 melanocytic nevi, 71 solar lentigines/seborrhoeic keratoses (SL/SebK), 19 lichen planus like keratoses (LPLK), three other skin lesions (vascular, inflammatory)). Of the 240
JEADV 2014
Frequencies and positive predictive values of patterns for different diagnoses are given in Table 1. A pattern of lines was the most common pattern (33%, n = 78) and SL/flat SebK was the most common diagnosis when this pattern was present (n = 35, 44.9%). A pattern of circles was found in 7% (n = 16) of lesions. More than two thirds of lesions with a pattern dominated by circles were malignant on histopathology (68.8%, n = 11), either pAK (n = 6) or LM (n = 5). Among benign lesions, the highest positive predictive value of circles was found for SL/SebK (PPV: 25.0%, 95% CI: 7.4–52.4%).
© 2014 European Academy of Dermatology and Venereology
Tschandl et al.
4
(a)
(c)
(b)
(d)
A pattern of clods was found in 8% (n = 20) of lesions and had the highest positive predictive value and specificity for BCCs (PPV 55.0%, 95% CI: 31.6–76.9%; specificity 95.9%, 95% CI: 92.3–98.1%). A pattern of dots occurred in 30% (n = 73) of all lesions with a similar frequency among all diagnoses (Table 1). The highest PPV and the highest specificity were found for pAK (PPV 37.0%, 95% CI: 26.0–49.1%,specificity73.1%,95%CI:65.8–79.6%). The structureless pattern was present in 22% (n = 53) of lesions, of which 43% (n = 23) were malignant. The highest PPV and the highest specificity were found for pAK (PPV 37.7%, 95% CI: 24.8–52.1%, specificity 80.7%, 95% CI: 74.0– 86.3%). Clues to malignant lesions (Table 2) Melanoma/Lentigo maligna (Fig. 4) Grey structures (grey dots, clods, circles or lines) were present in 95.8% (n = 23) of LM. The relative risk for LM when a grey structure was present was 8.9 (95% CI: 1.2–64.7). In 18.4% of lesions with incomplete circles (asymmetric pigmented follicular openings) the diagnosis was melanoma. The sensitivity and specificity for diagnosing LM were 58.3% and 71.3% respectively. Grey circles were present in 54.2% of LM, their PPV was 26.5% (95%CI: 15.0–41.1%) with a specificity of 83.3% respectively. The clue of a ‘circle within a circle’ was only found in one LM and therefore had a low sensitivity (4.2%). Rhomboids had a high specificity (91.7%) but their PPV was low (18.2%). Basal cell carcinoma Grey structures were found in 95.2% (n = 20) of BCCs. Dermatoscopic clues almost exclusively found
JEADV 2014
Figure 3 Clues for facial pigmented lesions not previously described in modified pattern analysis (representative examples of structures referred to are indicated by arrows): (a) Dermatoscopic image of a pigmented actinic keratosis with ‘four-dot clods’; (b) The dermatoscopic clue of ‘double circles’, histopathology revealed a seborrhoeic keratosis; (c) Dermatoscopic example of ‘rhomboids’ and ‘white circles’ in a pigmented actinic keratosis; (d) ‘Scale’(right arrow) as seen by dermatoscopy after appliance of immersion fluid on a pigmented actinic keratosis and ‘rhomboids’(left arrow).
in BCC were branched vessels [38.1%, n = 8, PPV 100.0% (95% CI: 62.9–100.0%), specificity 100% (95% CI: 98.3–100.0%)], serpentine vessels [(23.8%, n = 5, PPV 83.3% (95% CI: 36.1– 97.2%), specificity 99.5% (95% CI: 97.5–99.9%)] and ulceration [38.1%, n = 8, PPV 66.7% (95% CI: 35.0–89.9%), specificity 98.2 (95% CI: 95.4–99.5%)]. Pigmented actinic keratosis/pigmented Bowen’s disease (pAK/ pBD) Grey structures were also common in pAK/pBD (78.3%).
Incomplete circles were found in 43.5% (n = 30) of cases. The most specific clues were scale [PPV 72.2% (95% CI: 54.8– 85.8%) and specificity 94.2% (95% CI: 89.5–97.2%)], white circles [PPV 68.8% (95% CI: 50.0–83.9%) and specificity 94.2% (95% CI: 89.5–97.2%)] and four-dot clods [PPV 60.0% (95% CI: 15.4–93.5%) and specificity 98.8% (95% CI: 95.8–99.8%)]. Rhomboids had a high specificity [94.7% (95% CI: 90.2– 97.6%)] but a low PPV [59.1% (95% CI: 36.4–79.3%)]. Clues to benign lesions
Melanocytic nevus. The best clue for nevi was the presence of interfollicular reticular lines [PPV 18.1% (95% CI: 10.0–28.9%) and specificity 71.5 (95% CI: 64.8–77.5%)] which were present in 39.4% (n = 13) of all nevi. Grey structures were less common than in malignant lesions (66.7%, n = 22). Solar lentigo/seborrhoeic keratosis (SL/SebK) Interfollicular reticular lines were also common in SL/SebK. They occurred in 39.4% of SL/SebK. Their PPV was 38.9% (95% CI: 27.6–51.1%) and the specificity was 74.0% (95% CI: 66.7–80.4%). Incomplete circles showed a similar PPV (34.2%) and specificity (70.4%). A sharply demarcated border was found in 23.9% of lesions (PPV
© 2014 European Academy of Dermatology and Venereology
5
39.5% (95% CI: 25.0–55.6%), specificity 84.6 (95% CI: 78.3– 89.7%)). White dots/clods (‘milia’) and orange clods (‘comedo like openings’) were present in 15.5% (n = 11) of cases. Compared with other lesions on the face, grey structures were less common in SL/SebK (49.3%, n = 35).
LPLK, Lichen planus-like keratosis; PAK/pBD, pigmented actinic keratosis/pigmented Bowen’s disease; SL/SebK, Solar lentigo/seborrhoeic keratosis.
n.a.
13.7% (6.8–23.8%)
39.6% (26.5–54.0%) 37.7% (24.8–52.1%) 1.9% (0.3–10.1%) Structureless (n = 53)
3.8% (0.6–13.0%)
13.2% (5.5–25.4%)
n.a.
12.3% (5.8–22.1%) 15.1% (7.8–25.4%) 37.0% (26.0–.49.1%) 9.6% (4.0–18.8%) Dots (n = 73)
11.0% (4.9–20.5%)
10.0% (1.5–31.7%)) 30.0% (12.0–54.3%)
6.3% (1.0–30.3%) 25.0% (7.4–52.4%) n.a.
5.0% (0.8–25.0%)
37.5% (15.3–64.5%)) n.a.
55.0% (31.6–76.9%) n.a.
31.3%(11.1–58.6%) Circles (n = 16)
Clods (n = 20)
PPV for LPLK PPV for SL/SebK
44.9% (33.6–56.6%)
PPV for Naevus
11.5% (5.4–20.8%)
PPV for pAK/pBD
19.2% (11.2–29.7%)
PPV for basal cell carcinoma
14.1% (7.3–23.8%) Lines (n = 78)
n.a.
PPV for melanoma Pattern
Table 1 Positive predictive values and their 95% confidence intervals of dermatoscopic patterns according to different diagnoses JEADV 2014
10.3% (4.6–19.2%)
Dermatoscopy of flat pigmented facial lesions
Lichen planus-like keratosis (LPLK) Lichen planus-like keratoses on the face are dermatoscopically characterized by a pattern of dots [52.6%, n = 10, PPV 13.7% (95% CI: 6.8–23.8%), specificity 71.5% (95% CI: 65.1–77.4%)] or lines [42.1%, n = 8, PPV 10.3% (95% CI: 4.6–19.2%), specificity 68.3% (95% CI: 61.8– 74.4%)]. A sharply demarcated border was found in 26.3% (n = 5). Clues that are associated with malignant lesions were also common in LPLK: incomplete circles were found in 21.1% (n = 4) and grey structures in 94.7% (n = 18). Diagnostic models Lentigo maligna vs. pAK/pBD Since LM and facial pAK/pBD may have similar patterns on dermatoscopy, we built a multivariate diagnostic model including the most specific clues for each diagnosis. In the multivariate logistic model grey circles were associated with melanoma (odds ratio: 5.99, 95% CI: 1.27–28.18, P = 0.02). White circles (odds ratio for melanoma = 0.07, 95% CI: 0.01–0.70, P = 0.02), scale (odds ratio for melanoma = 0.09, 95% CI: 0.01–0.73, P = 0.02) and sharply demarcated border (odds ratio for melanoma = 0.09, 95% CI: 0.01–0.78, P = 0.03) were associated with pAK/pBD. Based on this model we created a simple score that adds one point if grey circles are present and subtracts one point if scale, white circles or a sharply demarcated border is present. The higher the score the higher is the risk for melanoma the lower the score the higher is the risk for pAK/ pBD. The diagnostic accuracy of this score expressed as area under the receiver operating characteristics curve was 0.87 (95% CI: 0.79–0.94). Lentigo maligna vs. solar lentigo/SebK/LPLK In a multivariate model including the most specific clues for LM and SL/SebK/ LPLK, the presence of any grey structure was significantly associated with melanoma (odds ratio: 16.13, 95% CI: 2.97–125.92, P = 0.008). A sharply demarcated scalloped border was in favour of SL/SebK/LPLK (odds ratio: 0.13, 95% CI: 0.02–1.08), but did not reach statistical significance (P = 0.06). If the presence of any grey structure was used as the sole criterion to differentiate melanoma from SL/SebK/LPLK the sensitivity was 95.8% and the specificity 41.1%. If LPLKs, which accounted for most false-positive cases, were excluded, the specificity of ‘any grey structure’ rose to 50.7%. Differentiation of benign from malignant lesions If the presence of any grey structure was used as the sole criterion to differentiate malignant lesions (melanoma, pAK/pBD and BCC) from
© 2014 European Academy of Dermatology and Venereology
Tschandl et al.
6
Table 2 Diagnostic indices of dermatoscopic clues to flat malignant facial lesions. Numbers in braces depict 95% confidence intervals RR
PPV
NPV
Sensitivity
Specificity
Melanoma Any grey structure
8.9, 95% CI: 1.2–64.7
13.3% (8.6–19.3%)
98.5% (91.9–99.8%)
95.8% (78.8–99.3%)
30.6% (24.5–37.2%)
Vessels as dots
3.5, 95% CI: 1.1–11.8
33.3% (5.3–77.3%)
90.6% (86.1–94.0%)
8.3% (1.3–27.0%)
98.1% (95.3–99.5%)
Incomplete circles
3.0, 95% CI: 1.4–6.5
18.4% (10.5–29.0%)
93.9% (89.1–97.0%)
58.3% (36.7–77.9%)
71.3% (64.8–77.2%)
Grey circles
4.6, 95% CI: 2.2–9.7
26.5% (15.0–41.1%)
94.2% (89.9–97.1%)
54.2% (32.8–74.4%)
83.3% (77.7–88.0%)
Rhomboids
2.0, 95% CI: 0.7–5.3
18.2% (5.3–40.3%)
90.8% (86.2–94.3%)
16.7% (4.8–37.4%)
91.7% (87.2–95.0)
Circle in a circle/Double circle
2.0, 95% CI: 0.3–12.3
20.0% (3.3–71.2%)
90.2% (85.7–93.7%)
4.2% (0.7–21.2%)
98.1% (95.3–99.5%)
Basal cell carcinoma Any grey structure
7.7, 95% CI: 1.1–56.6
11.6% (7.2–17.3%)
98.5% (91.9–99.8%)
95.2% (76.1–99.2%)
30.1% (24.1–36.7%)
Branched vessels
17.8, 95% CI: 10.5–30.3
100.0% (62.9–100.0%)
94.4% (90.6–97.0%)
38.1% (18.1–61.6%)
100.0% (98.3–100.0%)
Serpentine vessels
12.2, 95% CI: 6.7–22.1
83.3% (36.1–97.2%)
93.2% (89.1–96.0%)
23.8% (8.3–47.2%)
99.5% (97.5–99.9%)
Ulceration
11.7, 95% CI: 6.0–22.7
66.7% (35.0–89.9%)
94.3% (90.5–96.9%)
38.1% (18.2–61.6%)
98.2% (95.4–99.5%)
Pigmented actinic keratosis/Bowen’s disease White circles
3.0, 95% CI: 2.2–4.3
68.8% (50.0–83.9%)
77.4% (71.1–82.9%)
31.9% (21.2–44.2%)
94.2% (89.5–97.2%)
Sharply demarcated border
1.7, 95% CI: 1.2–2.6
44.2% (29.1–60.1%)
74.6% (68.0–80.5%)
27.5% (17.5–39.6%)
86.0% (79.8–90.8%)
Incomplete circles
1.7, 95% CI: 1.1–2.5
39.5% (28.5–51.4%)
76.2% (69.0–82.5%)
43.5% (31.6–56.0%)
73.1% (65.8–79.6%)
Rhomboids
2.3, 95% CI: 1.5–3.5
59.1 (36.4–79.3%)
74.3% (68.0–80.0%)
18.8% (10.4–30.1%)
94.7% (90.2–97.6%)
Four-dot clod
2.1, 95% CI: 1.0–4.5
60.0% (15.4–93.5%)
71.9% (65.7–77.6%)
4.3% (1.0–12.2%)
98.8% (95.8–99.8%)
Scale
3.4, 95% CI: 2.5–4.8
72.2% (54.8–85.8%)
78.9% (72.7–84.3%)
37.7% (26.3–50.2%)
94.2% (89.5–97.2%)
NPV, negative predictive value; PPV, positive predictive value; RR, relative risk.
(a)
(b)
Figure 4 Dermatoscopic structures predicting melanoma (representative examples of structures referred to are indicated by arrows): Dermatoscopic images show flat pigmented melanomas on the face with the feature of ‘grey structures’ (a) and ‘grey circles’ (b).
benign lesions the sensitivity was 85.1% and the specificity 39.7%.
Discussion The major differences between this and other studies of the dermatoscopy of pigmented lesions on the face are that we included only flat lesions and that we did not select the lesions based on their diagnosis. Another important difference is that our sample of facial melanomas includes mainly small and early lesions of LM while advanced cases have not been observed. This mirrors the situation in Australia where melanomas are recognized at an early stage. The focus on early lesions explains why the frequencies of classic criteria of facial melanomas such as
JEADV 2014
rhomboids3 or circle within a circle6 was much lower in our sample than in other studies. In our series the dermatoscopic appearance of melanoma on the face was contrasted with a variety of other pigmented lesions on chronic sun-damaged skin such as pAK, SL, LPLK, nevi and BCCs. We found that dermatoscopic criteria that have been reported to be specific for facial melanoma are not as specific as previously thought. Asymmetric pigmented follicular openings3,4,14 (incomplete circles) are not very specific for melanoma. They can be found in a significant portion of solar lentigines and pigmented actinic keratoses. Due to the prospective nature our study (lesions were not selected because of the diagnosis) we focused on the practical significance of dermatoscopic patterns and clues and calculated
© 2014 European Academy of Dermatology and Venereology
Dermatoscopy of flat pigmented facial lesions
positive and negative predictive values taking into account disease prevalence. We demonstrated that it is important to differentiate between different types of dermatoscopic patterns on the face because patterns are to some extent associated with specific diagnoses. We do not recommend the use of the term ‘pseudo-pigmentnetwork’ because it lumps together three different types of patterns: (i) circles; (ii) structureless and (iii) dots (Figs 1,2). Although these three patterns are not 100% specific they are of some significance. A pattern of clods (‘globules’) in a flat pigmented lesion is highly predictive for BCC especially if other clues are present too. Although other patterns are less specific it is useful to recognize them because some diagnoses become more likely or unlikely depending on the pattern observed. Although a pattern of reticular or curved lines does not exclude any diagnosis except BCC, it most likely indicates a SL especially if other criteria of SL, such as a sharply demarcated border, are present, and clues of melanoma or pAK are missing. With the exception of a pattern of clods, early facial melanoma may present with any pattern, but a pattern of pigmented circles is more likely than others. According to our analysis, a pattern of pigmented circles is the most common pattern in early LM and in our multivariate logistic model, we found that grey circles had a high association with melanoma while white circles, scale and a sharply demarcated border were more in favour pAK/pBD. It has to be emphasized, however, that a pattern of pigmented circles may not be the most common pattern in advanced cases of lentigo maligna. In advanced cases, when pigmentation has become more prominent, the interfollicular space will get pigmented and a structureless pattern will prevail, which is in line with the progression model of Stolz et al.3,4 On the other hand, we have shown that a pattern of circles can be found in other diagnoses too, chief among them being pAK, and thus is not highly specific. We confirmed that pAK is the great masquerader, it is not characterized by one specific pattern: Circles, dots and structureless patterns are equally likely. This is the reason why LM and pAK/pBD may appear with the same pattern. The differential diagnosis is difficult because even some dermatoscopic clues such as rhomboids and incomplete asymmetric circles can be found in both. Akay et al.8 were the first to notice that LM had striking dermatoscopic similarities to pAK. In our series we found that white circles, scale and a sharply demarcated border are suggestive of pAK, while grey circles suggest melanoma. Four-dot clods (‘rosettes’) were not only found in pAK/pBD but also in one melanoma and one BCC. This structure might not be as specific for pAK as previously thought. It has to emphasized that most dermatoscopic images were taken with non-polarized dermatoscopes. Because four-dot clods are mainly visible with polarized light, we might have underestimated the presence of this structure. Rhomboids, previously described to
JEADV 2014
7
be diagnostic of LM,3 were not only found in melanomas but also in pAK/pBD. In this series of consecutively collected flat and pigmented skin lesions, about half were malignant. This represents the selection bias when specimens are selected in routine practice as requiring biopsy (such a selection bias is unavoidable when histology is required for the analysis). Predictive values and diagnostic measures therefore are biased and even small correlations with malignant nature may represent a strong correlation in daily practice with unselected lesions. The overall high percentage of ‘any grey structure’ in this series may be explained by the fact that the author providing the cases (and also the other authors) sees the presence of grey colour in lesions on the face as a strong clue to a malignant process, if no confident benign diagnosis can be made otherwise. Even through this bias the feature ‘any grey structure’ remained an independent significant coefficient in multivariate analysis, underlining the importance of this feature. A high percentage of LPLKs also shows grey structures because of its regressive nature. By relying on ‘grey structures’ as a predictor of a malignant process, more of those lesions may be excised for diagnostic reasons, which, in the eyes of the authors, is necessary to not miss a potentially deadly melanoma.
References 1 Kittler H, Pehamberger H, Wolff K, Binder M. Diagnostic accuracy of dermoscopy. Lancet Oncol 2002; 3: 159–165. 2 Rosendahl C, Tschandl P, Cameron A, Kittler H. Diagnostic accuracy of dermatoscopy for melanocytic and nonmelanocytic pigmented lesions. J Am Acad Dermatol 2011; 64: 1068–1073. 3 Schiffner R, Schiffner-Rohe J, Vogt T et al. Improvement of early recognition of lentigo maligna using dermatoscopy. J Am Acad Dermatol 2000; 42: 25–32. 4 Stolz W, Schiffner R, Burgdorf WH. Dermatoscopy for facial pigmented skin lesions. Clin Dermatol 2002; 20: 276–278. 5 Tanaka M, Sawada M, Kobayashi K. Key points in dermoscopic differentiation between lentigo maligna and solar lentigo. J Dermatol 2011; 38: 53–58. 6 Pralong P, Bathelier E, Dalle S, Poulalhon N, Debarbieux S, Thomas L. Dermoscopy of lentigo maligna melanoma: report of 125 cases. Br J Dermatol 2012; 167: 280–287. 7 Pock L, Drlik L, Hercogova J. Dermatoscopy of pigmented actinic keratosis–a striking similarity to lentigo maligna. Int J Dermatol 2007; 46: 414– 416. 8 Akay BN, Kocyigit P, Heper AO, Erdem C. Dermatoscopy of flat pigmented facial lesions: diagnostic challenge between pigmented actinic keratosis and lentigo maligna. Br J Dermatol 2010; 163: 1212–1217. 9 Mahendran RM, Newton-Bishop JA. Survey of U.K. current practice in the treatment of lentigo maligna. Br J Dermatol 2001; 144: 71–76. 10 Wong JG, Toole JW, Demers AA, Musto G, Wiseman MC. Topical 5% imiquimod in the treatment of lentigo maligna. J Cutan Med Surg 2012; 16: 245–249. 11 Kittler H. Dermatoscopy: introduction of a new algorithmic method based on pattern analysis for diagnosis of pigmented skin lesions. Dermatopathol Pract Concept 2007; 13: 3. 12 Zalaudek I, Giacomel J, Schmid K et al. Dermatoscopy of facial actinic keratosis, intraepidermal carcinoma, and invasive squamous cell carcinoma: a progression model. J Am Acad Dermatol 2012; 66: 589–597.
© 2014 European Academy of Dermatology and Venereology
Tschandl et al.
8
13 Rosendahl C, Cameron A, Argenziano G, Zalaudek I, Tschandl P, Kittler H. Dermoscopy of squamous cell carcinoma and keratoacanthoma. Arch Dermatol 2012; 148: 1386–1392. 14 Sahin MT, Ozturkcan S, Ermertcan AT, Gunes AT. A comparison of dermoscopic features among lentigo senilis/initial seborrheic keratosis, seborrheic keratosis, lentigo maligna and lentigo maligna melanoma on the face. J Dermatol 2004; 31: 884–889.
JEADV 2014
Supporting information Additional Supporting Information may be found in the online version of this article: Table S1. Predefined dermatoscopic features assessed within this study.
© 2014 European Academy of Dermatology and Venereology
