Currently, little is known regarding the potential prognostic value of histologic features in primary cutaneous neuroendocrine (Merkel cell) carcinomas (MCC).
Currently, little is known regarding the potential prognostic value of histologic features in primary cutaneous neuroendocrine (Merkel cell) carcinomas (MCC).
In a retrospective review of the tumor histology and clinical outcome data (median follow-up, 51 months; range, 3-224 months) of 156 patients with a diagnosis of MCC, the following histologic features were evaluated: tumor thickness, tumor size (greatest dimension of the tumor), microanatomic compartment involved by tumor (dermis and/or subcutis and/or deeper), tumor growth pattern (nodular circumscribed vs infiltrative), lymphovascular invasion (LVI), tumor-infiltrating lymphocytes, tumor necrosis, ulceration, and solar elastosis.
The overall 5-year survival rate was 67.5%. On univariate analysis, parameters that were associated significantly with survival were tumor thickness (P = .001), tumor size (P = .0002), deepest anatomic compartment involved by tumor (P = .0003), tumor growth pattern (P = .003), LVI (P < .00001), tumor-infiltrating lymphocytes (P = .05), and solar elastosis (P = .04). On multivariate analysis, the presence of a nodular growth pattern, low tumor depth, and absence of LVI were associated with longer survival.
In addition to the known prognostic value of tumor stage, 3 histologic features were identified to have prognostic significance: tumor thickness (depth of tumor invasion), the presence of LVI, and tumor growth pattern. Cancer 2008. © 2008 American Cancer Society.
Primary cutaneous neuroendocrine (Merkel cell) carcinoma (MCC) was reported initially in 1972 by Toker1 under the heading trabecular carcinoma. Subsequent ultrastructural studies revealed the presence of dense-core neuroendocrine granules in the cytoplasm of tumor cells similar, to those observed in Merkel cells,2 which is why MCC has gradually replaced the original name. Whether MCC actually derives from Merkel cells or has a common origin with other epithelial cells, but follows a neuroendocrine differentiation pathway, is still being investigated.
MCC is an uncommon skin tumor with a reported age-adjusted incidence of 0.24 to 0.44 per 100,000 person-years.3–5 From 1986 to 2001, its incidence has tripled; however, it is unclear how much of this increase represents a true rise in tumor occurrence or is merely related to better reporting. MCC affects predominantly the Caucasian population and is more frequent in men.3, 6, 7 MCC tends to affect the elderly, with a median age at presentation between 67 and 76 years (range, 8-98 years).3, 7–13 It usually occurs in sun-damaged skin of the head and neck region or extremities, but it also involves sun-protected sites, such as the buttocks.3, 6, 7, 9, 11–19 Recently, a polyomavirus has been identified that is integrated into the genome of MCC and may play a role in the pathogenesis of these tumors.20
Among skin tumors, MCC is regarded as 1 of the more ‘aggressive’ cancers. Historic data typically derived from small case series have reported dismal 5-year survival rates ranging from 29% to 64%.7, 9, 10, 19, 21–23 However, studies involving larger number of patients have suggested up to 74% survival rates at 5 years.15, 24, 25 The most important prognostic determinant is tumor stage,3, 9, 10, 24–26 which is most accurate if information of the pathologically confirmed status of the regional lymph nodes is included (pathologic stage).15 A 3-tiered staging systems was proposed in 1980, with stage I coding for disease limited to skin, stage II for regional lymph node involvement, and stage III for distant spread.25 Currently, a 4-tiered system is widely used that separates the patients with localized cutaneous disease based on the largest clinically assessed diameter of the primary tumor (<2 cm or ≥2 cm).24 Lymph node-negative disease with a tumor <2 cm is classified as stage I, lymph node-negative disease with a tumor ≥2 cm is classified as stage II, lymph node-positive disease is classified as stage III, and the presence of distant metastatic disease is categorized as stage IV.
Although low-stage disease predicts a relatively good prognosis, it is not an optimal indicator, because approximately 20% of patients with clinical stage I MCC will die of disease within 5 years from diagnosis.15 Therefore, it would be beneficial to have independent pathologic predictors of tumor behavior in MCC, especially for low-stage disease. However, to our knowledge, the rarity of this entity has precluded to date the assembly of large series of primary MCCs with sufficient statistical power to systematically analyze the predictive value of various morphologic parameters. The objective of the current study was to analyze the clinical relevance of morphologic parameters as prognostic markers in a large series of MCCs.
This study was approved by the Institutional Review Board. Pathology reports and slides for all patients who had a diagnosis of MCC from 1980 to 2005 were retrieved. Only patients who had adequate pathologic material for assessment of histologic parameters (at least 1 representative slide of the primary tumor available for review) and verified clinical follow-up were included in the study. Because some of the older patients were diagnosed without the advent of immunohistochemistry, to ensure that our study included only primary cutaneous MCCs and not metastatic lesions, we selected only patients who had disease defined by clinical evaluation and follow-up as primary cutaneous carcinoma. In total, 156 patients were identified who fulfilled the criteria described above with a median follow-up of 51 months for the entire cohort (range, 3-224 months).
Hematoxylin and eosin-stained sections were reviewed, and the following morphologic parameters were recorded: tumor thickness (measured in mm from the granular layer of the epidermis to the deepest portion of the tumor), tumor size (the greatest tumor dimension in mm), deepest microanatomic compartment involved by tumor (dermis, subcutaneous tissue, or deeper planes, such as skeletal muscle or bone), lymphovascular invasion (LVI) (absent, present, or extensive), tumor-infiltrating lymphocytes (absent, present nondense, or dense; analogous to the absent, nonbrisk, and brisk categories for cutaneous melanoma),27–29 tumor necrosis (absent or present), ulceration (absent or present), and solar elastosis (absent or present). LVI was defined as the presence of tumor emboli within vascular spaces outside the tumor main boundaries. Extensive vascular invasion was defined as multiple tumor emboli (>2) readily identified within the sections. In addition, the tumor growth pattern was assessed for every patient in which full profile sections of the tumor were available for histologic evaluation. Tumors with a relatively well circumscribed interface with the surrounding tissue, typically composed of 1 nodule or multiple nodules, were classified as ‘nodular’ (Fig. 1a,b); whereas tumors that displayed an infiltrative growth pattern with single cells, rows, trabeculae, or strands of cells slicing through dermal collagen or deeper soft tissue were termed ‘infiltrative’ (Fig. 1c,d). A tumor that exhibited both nodular and infiltrative patterns was classified as infiltrative.
Univariate analysis of disease-specific survival was performed using Kaplan-Meier curves and the log-rank test. For multivariate survival analysis, a Cox regression model was used. To fit the multivariate models, a forward conditional method was used to introduce variables stepwise into the model. Statistical significance was defined as P < .05. For univariate analysis, tumor size and thickness were categorized in 4 groups: <5 mm, ≥5 and <10 mm, ≥10 and <20 mm, and ≥20 mm. For multivariate analysis, tumor size and thickness were treated as continuous variables. With regard to microanatomic skin compartment involved by tumor, cases were divided for multivariate analysis into those with dermal involvement only versus those with subcutaneous or deeper involvement.
The distribution of patients according to demographic parameters and the clinical data are presented in Table 1. The most common primary tumor site was the extremity area (42%), followed by head and neck (37%), buttocks (16%), and trunk (4.5%). The median age at presentation was 69.5 years (range, 39-88 years). Men were affected slightly more often than women (56% vs 44%, respectively). Approximately 52% of patients presented with pathologic stage I disease, 19% presented with stage II disease, 23% presented with stage III disease, and 6% presented with stage IV disease. Overall, there were 85 deaths recorded (54%), including 50 patients (32%) who died of disease at a median follow-up of 51 months (range, 3-224 months).
|Parameter||No. of Patients (%)|
|Location of primary tumor (N=156)|
|Head and neck||58 (37.2)|
|Stage of disease at presentation (N=156)|
|Status at last follow-up (N=156)|
|Follow-up, mo (N=156)|
|Age, y (N=156)|
|Tumor thickness, mm (N=149)|
|Mean (range) 12.3 (0.7-127)|
|≥5 and <10||38 (25.5)|
|≥10 and <20||49 (32.9)|
|Tumor size, mm (N=154)|
|Mean (range) 20.1 (1-140)|
|≥5 and <10||33 (21.4)|
|≥10 and <20||49 (31.8)|
|Anatomic compartment (N=155)|
|Subcutaneous tissue||111 (71.6)|
|Deeper fascia and skeletal muscle||7 (4.5)|
|Tumor growth pattern (N=108)|
|Tumor infiltrating lymphocytes (N=153)|
|Present, nondense||55 (35.9)|
|Tumor necrosis (N=95)|
|Solar elastosis (N=155)|
The distribution of histologic parameters in our series is illustrated in Table 1. The mean tumor thickness was 12.3 mm (range, 1-127 mm), and the mean tumor size was 20.1 mm (range, 1-140 mm). A direct correlation between tumor thickness and tumor size was observed, with larger tumors exhibiting deeper depths of involvement (Pearson correlation coefficient = 0.8; P < .0001).
Approximately 24% of all tumors were microanatomically confined to the dermis, 72% involved subcutaneous tissue, and 4.5% invaded the deeper planes (skeletal muscle or bone). Tumor growth pattern was evaluated reliably in 108 cases which had adequate full-profile sections available (required for this assessment). An infiltrative growth pattern was observed in the majority of tumors (71%), a nodular growth pattern in 20%, and a multinodular 1 in only 6%. Two tumors exhibited polypoid architecture, and 1 tumor displayed a pure trabecular pattern. For the statistical analysis of growth patterns, nodular, multinodular, and polypoid tumors were grouped together into the nodular growth pattern category. The single tumor with a trabecular pattern was included in the infiltrative group.
LVI was identified in 60% of tumors and was extensive in 4%. In 53% of tumors, there was no inflammatory infiltrate associated with the tumor. The inflammation was mild to moderate (nondense) in 36% of tumors and dense in 11%. Tumor necrosis was observed in 28% of patients. Ulceration of the epidermis overlying the tumor was present in 8% of patients. Moderate-to-severe solar elastosis was noted in the peritumoral stroma of 55% of patients.
The following histologic parameters demonstrated a statistically significant correlation with survival in univariate analysis: tumor thickness (P = .001), tumor size (P = .0002), deepest anatomic compartment involved by tumor (P = .0003), tumor growth pattern (P = 0.003), LVI (P < .00001), tumor infiltrating lymphocytes (P = .05) and solar elastosis (P = .04) (Table 2). Increased tumor thickness, tumor size, and deeper anatomic compartment were associated with decreased survival (Table 2, Fig. 2a-c). There was a significant difference in 5-year survival rates between patients who had tumors with a nodular versus infiltrative growth pattern (82% vs 55%) (Fig. 2d). The presence of LVI, the absence of tumor-infiltrating lymphocytes, and the lack of solar elastosis also were associated with significantly lower 5-year survival rates in univariate analysis (Fig. 2e-g). Ulceration and tumor necrosis were not associated with significant differences in survival.
|Parameter||2-Year Survival, %||5-Year Survival, %||P|
|Tumor thickness, mm||.001|
|≥5 and <10||86.5||77.2|
|≥10 and <20||85.7||61.3|
|Tumor size, mm||.0002|
|≥5 and <10||93.7||81.8|
|≥10 and <20||91.7||71.5|
|Deeper fascia and skeletal muscle||33.8||16.9|
|Tumor growth pattern||.003|
|Tumor infiltrating lymphocytes||.05|
The 7 histologic parameters that demonstrated a statistically significant association with survival in univariate analysis (tumor thickness, tumor size, deepest anatomic compartment involved by tumor, tumor growth pattern, LVI, tumor infiltrating lymphocytes, and solar elastosis) and the tumor stage were used to fit 2 multivariate Cox regression models (Table 3). On multivariate analysis only stage, tumor growth pattern, and LVI correlated independently with survival. The odds ratios of death from disease for infiltrative versus nodular tumor growth pattern and for presence versus absence of LVI were 6.85 (P = .001) and 3.84 (P = .007), respectively. The odds ratios for stage II, III, and IV disease versus I disease were 2.05 (P = 0.1), 2.68 (P = .02), and 33.66 (P < .00001), respectively (Table 3, Model 1).
|II vs I||2.05||0.78-5.36||.1|
|III vs I||2.68||1.12-6.41||.02|
|IV vs I||33.66||10.35-109.43||<.00001|
|Tumor growth pattern|
|Infiltrative vs nodular||6.85||2.11-22.19||.001|
|Present vs absent||3.84||1.43-10.28||.007|
|II vs I||1.28||0.47-3.45||.6|
|III vs I||3.76||1.76-8.03||.0006|
|IV vs I||17.79||6.84-46.26||<.00001|
|Present vs absent||3.19||1.38-7.37||.006|
Because the tumor growth pattern could be analyzed in only 108 patients, a second multivariate model was fitted that did not include this variable to take advantage of all patients. In this model (Table 3, Model 2), in addition to stage and LVI, tumor thickness emerged as an independent predictor of survival. The odds ratio for a 1-mm increase in tumor thickness was 1.02 (P = .005). If in the second multivariate model thickness was not analyzed as a continuous variable but, instead, was categorized in 2 groups (<2 and ≥2 cm), then the odds ratio of death from disease for tumors <2 cm versus tumors ≥ 2 cm was 2.26 (95% confidence interval, 1.11-4.58), P = .02.
Patients with stage I disease and patients with lymph node-negative (stage I + II) disease had 5-years survival rates of 86% and 82.5%, respectively. In an attempt to identify predictors of behavior that could better stratify this group of tumors better, we conducted a survival analysis that was restricted to stage I and stage I + II tumors (Table 4). In the stage I group (N = 81 patients), the only variables that were associated significantly with survival were the pattern of tumor growth (5-year survival rates for nodular vs infiltrative pattern: 100% vs 69%, respectively; P = .01) and the deepest anatomic compartment involved by tumor (5-year survival rates for tumor involving dermis, subcutis, and deeper planes: 96%, 85.5%, and 0%, respectively; P < .00001).
|Parameter||Pathologic Stage I Disease, N=81 Patients||Pathologic Stage I and II Disease, N=111 Patients|
|2-Year Survival, %||5-Year Survival, %||P||2-Year Survival, %||5-Year Survival, %||P|
|Tumor thickness, mm||.08||.1|
|Tumor size, mm||.4||.3|
|Deeper fascia and skeletal muscle||0||0||20||0|
|Tumor growth pattern||.01||.002|
|Tumor infiltrating lymphocytes||.1||.05|
|Associated solar elastosis||.5||.1|
The other parameters did not reach statistical significance; however, several trends were noted. None of the patients who had tumors that measured <5 mm in thickness died of disease at 5 years with a median follow-up of 51 months, compared with a 80% 5-year survival rate for patients who had tumors that measured >5 mm in thickness (P = .08). In addition, patients who had tumors that lacked LVI had a 5-year survival rate of 93% compared with 80% for patients who had documented LVI (P = .09).
For the lymph node-negative group (N = 111 patients), the variables that were associated significantly with survival were the pattern of tumor growth (5-year survival rates for nodular vs infiltrative patterns: 100% vs 66%, respectively; P = .002), the anatomic compartment involved by tumor (5-year survival rates for tumor involving dermis, subcutis. and deeper planes: 93%, 85%, and 0%, respectively; P < .00001), LVI (5-year survival rates for tumor with vs without LVI: 74% vs 93%, respectively; P = .006), and tumor infiltrating lymphocytes (5-year survival rates for tumor with vs without tumor infiltrating lymphocytes: 91% vs 74%, respectively; P = .05).
The main purpose of this study was to identify histologic parameters that might assist in improving the prognostic assessment of patients with MCC. Prior studies have examined various morphologic parameters, such as tumor thickness and depth of invasion (summarized in Table 5), but with conflicting or unconvincing results.6, 16–18, 30
|Study||No. of Patients||Tumor Size, mm||Tumor Thickness, mm||Anatomic Compartment*||Tumor Growth Pattern||LVI||TIL||Solar Elastosis||Necrosis||Ulceration||Cell Size||Mitoses||Ki-67 Index||Apoptosis|
|Current study, 2008||156||U||M||U||M**||M||U||U||N||N||—||—||—||—|
The average tumor thickness reported in MCC is close to 10 mm6, 16 (it was 12.3 mm in our study); however, previous studies have reported no correlation between tumor thickness and overall survival in this tumor.6, 16, 30 Mott et al17 observed that, when anatomic compartment involved by tumor was considered, invasion of the subcutis was correlated with decreased survival; however, another study by Skelton et al failed to confirm that finding.18 Tumor size in MCCs averages from 7 mm to 30 mm in most studies8, 9, 15–17, 24 (it was 20.1 mm in our study), and has been found to correlate with survival in univariate analysis and in multivariate analysis of patients with clinical stage I disease, but it loses significance in higher stages.9, 16–18, 24 A recent study by Sandel et al indicated that there was no correlation between tumor size and prognosis.6 Several histologic types are recognized in MCC based either on the size of the cells (small, intermediate, large)31–33 or on the architecture (solid, diffuse, or trabecular).16, 17, 34 Skelton et al observed a correlation between small cell type predominance and poor prognosis,18 and Mott et al found that a diffuse growth pattern is an adverse prognostic parameter,17 whereas 2 other studies failed to indicate an association between tumor pattern and prognosis.16, 34 LVI was encountered in MCC in approximately 30% to 60% of patients16–18 (60% in our data) and was not associated significantly with prognosis,17, 18 except in 1 study in which there was a trend toward worse survival.16 With respect to the presence of a tumor-associated inflammatory infiltrate, the findings from previous studies are contradictory and indicate that either the presence17 or the absence16 of inflammation correlates with a poor prognosis. Skelton et al18 reported that a high mitotic rate was associated with poor survival; however, Llombart et al failed to confirm that finding.16 Other potential histologic markers that were studied previously, including ulceration, tumor necrosis, and apoptosis, did not reveal a correlation with tumor behavior.16–18
In the current study, we examined 9 morphologic variables in a large series of MCCs. Seven of these parameters had been examined in prior, albeit much smaller studies: tumor thickness, tumor size, anatomic compartment involved by tumor, LVI, tumor infiltrating lymphocytes, necrosis, and ulceration. Two parameters had not been reported to our knowledge in any previously published study on this subject: the presence or absence of solar elastosis and the architectural growth pattern (nodular vs infiltrative).
On univariate analysis, 7 variables correlated in a statistically significant fashion with less favorable survival: increased tumor thickness, increased tumor size, involvement of subcutis or deeper planes, infiltrative growth pattern, presence of LVI, absence of tumor infiltrating lymphocytes and absence of solar elastosis. On multivariate analysis that included these 7 variables as well as the pathologic 4-tiered staging system, only tumor growth pattern and LVI, along with stage, were independent predictors of survival. If the tumor growth pattern was excluded from the model, then tumor thickness emerged as an independent determinant of survival. For patients with lymph node-negative disease, a nodular pattern of tumor growth, dermal location, absence of LVI, and presence of tumor-associated lymphocytes predicted excellent behavior.
How the differences between our findings (for example, with regard to tumor thickness and LVI) and findings from previous studies (which failed to document prognostic significance for these parameters) be explained? The sizes of patient samples in prior studies were smaller, typically in the range of 20 to 60 patients (Table 5).6, 16, 17, 30 Only 1 study reached a comparable sample size of 132 patients; however, histologic parameters in that study were correlated with outcome only in 85 patients.18 It is likely that many previous studies were statistically under powered for their scope. With 156 patients overall and 50 patients who died of disease, our analysis had superior statistical power compared with previously reported series. However, cognizant of limitations of any retrospective analysis, we believe that our findings need to be validated in a larger series, ideally in a prospective study, before prognostic value can be assigned firmly to histologic parameters that possibly may be used for revised staging models. Because of the rarity of MCC, such study cohorts likely will require multiinstitutional collaboration.
In conclusion, the current results indicate that tumor architecture (nodular or infiltrative), tumor thickness, and LVI are independent predictors of survival along with disease stage for patients with MCC. We encourage pathologists to document these parameters so that their significance and potential for clinical use can be studied further.
We thank Jason Bini for help with data retrieval and data entry.