Histopathologic Characteristics, Recurrence Patterns, and Survival of 129 Patients With Desmoplastic Melanoma
- First Online:
- Cite this article as:
- Posther, K.E., Selim, M.A., Mosca, P.J. et al. Ann Surg Oncol (2006) 13: 728. doi:10.1245/ASO.2006.03.091
- 128 Views
Desmoplastic melanoma (DM) has been associated with higher local recurrence rates than other types of cutaneous melanoma. Current controversies regarding locoregional treatment strategies warrant further investigation.
Retrospective review of a prospectively maintained melanoma database identified 129 patients with DM out of >12,500 melanoma patients referred for treatment from 1980 to 2003. Clinical and histopathologic characteristics, recurrence, and survival were analyzed.
The median follow-up was 4.0 years. Of the 129 patients identified, 82 (63.6%) were male, and the median age was 55.2 years. American Joint Committee on Cancer staging was I, II, and III in 25.6%, 68.0%, and 6.4% of patients, respectively, and the mean tumor thickness was 4.42 mm. Overall survival was 76% at 5 years and 64% at 10 years; median survival was 13.0 years. A total of 51 patients (39.5%) experienced disease recurrence, with a median time to recurrence of 1.3 years. The first recurrence was local in 18 patients (14.0%), nodal in 18 patients (14.0%), and distant in 15 patients (11.6%), with median survivals of 6.7, 7.8, and 1.8 years, respectively. Statistically significant predictors of recurrence were a final positive margin status and stage, and predictors of overall survival were patient age and stage.
Compared with other types of melanoma, DMs do demonstrate a tendency toward local recurrence, thus suggesting that narrower excision margins may not be appropriate in this population. Scrutiny of final surgical margins is critical to the local management of DM. In addition, the potential for regional nodal involvement must be considered at the time of diagnosis and during surveillance for disease recurrence.
First described by Conley et al.1 in 1971, desmoplastic melanoma (DM) is a rare variant of malignant melanoma and represents <1% of all cutaneous melanomas. Since 1971, several single-center series have sought to elucidate the unique histopathologic features and clinical behavior of DM.2–12 As described in these series and reviews, DMs tend to present with advanced Breslow thickness and demonstrate higher rates of local recurrence, lower rates of nodal involvement, and similar rates of survival compared with other cutaneous melanomas.4,5,7,13–15
Clinically, DMs present as slow-growing, indurated, painless, and often amelanotic palpable nodules.13 Like all melanomas, they tend to appear in more sun-exposed regions of skin and are commonly identified in the head and neck location. The high local recurrence rates associated with DM have been attributed to misdiagnosis of the original lesion and inadequate surgical excision.4,12 Because of its aggressive local behavior, previous reports have advised that DM be treated with wide local excision, with routine surgical margins of at least 2.0 cm.4,5,15 In addition to the proven risk factors defined for cutaneous melanoma, such as stage, tumor thickness, and depth of invasion, a poorer prognosis for patients with DM has been associated in some studies with male sex, older age, head and neck location, high mitotic index, and neurotropic features.4
The largest series of DM in the literature to date was reported by the Sydney Melanoma Unit and included 280 patients.10 The authors found that the survival of DM was similar to that of other cutaneous melanomas, with a lower rate of nodal involvement both at diagnosis and at first recurrence. Recurrence was higher for DM resected with less than a 1.0-cm excision margin, compared with resection with greater than a 2.0-cm margin. Presentation at a later stage of disease suggested a delay in diagnosis and, possibly, more rapid disease progression.
Several recent series have observed less than a 10% rate of pathologically confirmed regional nodal metastases at the time of primary diagnosis of DM.5,7,10,12 Most DM patients with regional nodal involvement have their nodal disease identified first on clinical examination, with confirmation of metastases by fine-needle aspiration or biopsy to determine the need for formal lymph node dissection. For patients without clinically suspicious lymph nodes, regional management strategies and sentinel lymph node biopsy (SLNB) are controversial.23–25
Duke University Medical Center (DUMC) has prospectively maintained a comprehensive database of melanoma patients for several decades. We conducted a retrospective review of this database, including analysis of histopathologic features, recurrence, and survival, for all patients with a diagnosis of DM after its initial description in the literature in 1971. For this review, we specifically examined disease characteristics, recurrence patterns, nodal metastases, and correlation with survival to provide additional guidance for the surgical management of primary and recurrent DM.
The Duke Comprehensive Cancer Center has prospectively maintained a database of melanoma patients since 1979, and the database also includes data retrospectively collected for patients from 1970 to 1979. After institutional review board approval, this database was retrospectively examined for cases of DM. For this review, DM was pathologically defined as loosely arranged, atypical spindle cells invading the dermis; staining for S-100 and HMB-45, neurotropic features, and the presence or absence of pigmentation also contributed to diagnosis. In all cases, patient records maintained in outpatient clinic files were then examined, including outside pathology reports, DUMC dermatopathologist reviews of outside pathology slides, and follow-up reports for recurrence and survival information.
Database variables included patient name, medical record number, date of birth, sex, and age at diagnosis. Information about the primary lesion included the date of original clinical diagnosis of cutaneous melanoma, diagnostic procedure (type of biopsy or excision performed), primary site of diagnosis, American Joint Committee on Cancer (AJCC) stage, definitive surgical procedure performed for the primary lesion, Breslow thickness, Clark level, and evidence of ulceration. AJCC staging (6th edition) was updated for cases, using current criteria (6th edition), for the purpose of this review.26–28
For cases referred to DUMC from other institutions, the date of first diagnosis was based on the original diagnostic procedure performed at the outside institution. The definitive surgical procedure for the primary melanoma was defined as the most extensive operation performed within 4 months of the initial diagnosis to achieve negative margins. Cases in which pathology revealed positive margins, without re-excision within a 4-month time period, were considered to have final positive margins. Alternatively, if re-excision was later performed at the same site as the original lesion, after prior excision with negative margins, this was considered to be a local recurrence.
At the original time of referral, outside pathology slides and reports were individually reviewed by a dedicated DUMC dermatopathologist. If the local pathology and DUMC pathologic reviews disagreed regarding tumor thickness or margin involvement, results of the DUMC review were used for purposes of this analysis. When available, data regarding tumor ulceration, perineural involvement, pigmentation, S-100 and HMB-45 staining, and number of mitotic figures per high-power field were extracted from pathology reports. The number of mitotic figures was categorized as low (<1/mm2), moderate (1–4/mm2), or high (>4/mm2). The presence or absence of pigment was determined both from the gross descriptions on physical examination and from pathology reports.
Where specifically noted, surgical margins were determined from dictated operative reports according to surgeon measurement. In the absence of information about surgical margins, pathologic margins were obtained from the pathology report by using the measurement of the margin closest to the primary lesion or residual scar. If pathologic margins were not provided in the original report, they were calculated by subtracting the greatest diameter of the malignant lesion from the diameter of the entire resected specimen and dividing this number by 2.
Tumor thickness was based on pathology reports of completely excised lesions. In the case of incompletely excised lesions with positive deep margins, tumor thickness was estimated by the thickest measurement provided by the pathologist. For the purpose of this analysis, this margin (either surgical or pathologic) was collectively referred to as the clinical margin.
Lymph nodes were defined as clinically positive or negative on the basis of physical examination and suspicion of malignant involvement. For patients who underwent diagnostic evaluation of regional lymph nodes, the following information was recorded: dates of procedures, location of regional nodal basin, diagnostic and final surgical procedures, total number of nodes identified in the pathology specimen, and number of histopathologically positive nodes. Patients with identification of enlarged lymph nodes at the time of primary diagnosis, followed by positive pathologic assessment within 4 months of the original excision, were considered to have AJCC stage III disease at primary diagnosis. Patients with clinically negative lymph nodes at the time of primary excision and subsequent pathologically positive lymph nodes identified >60 days from the time of the original operation were considered to have a nodal recurrence.
Time to recurrence was calculated from the date of definitive surgical excision to the date of documented recurrence, either by date of biopsy (for locoregional disease) or date of radiographical evidence (for metastatic disease). Recurrence was defined as local if described by the surgeon to require excision at or near the previous excision site. Regional nodal recurrence was defined as pathologically confirmed metastatic disease involving the first nodal basin to drain the site of primary tumor involvement. In cases of multiple sites of recurrence identified within the same 30-day time period, the most advanced recurrence determined the type of recurrence.
Vital status was captured according to the date of the most recent follow-up or date of death, and survival was calculated from date of diagnosis to the date of the most recent follow-up or date of death. Cause of death was recorded as melanoma related, cardiac, or unknown.
For selected cases, pathology reports and specimens were re-reviewed by a dedicated dermatopathologist and discussed at a multidisciplinary melanoma conference. Cases for review were selected on the basis of initial misdiagnosis, conflicting outside hospital and DUMC pathology reports, multiple local recurrences, and node-positive cases identified at either diagnosis or recurrence.
The mean and median duration of disease-free survival (time to first recurrence) and overall survival were determined by Kaplan-Meier (product-limit) estimates. 2-, 5-, and 10-year disease-free and overall survival estimates were also calculated. Predictors of disease-free and overall survival were identified by using the log-rank test, and P < .05 was used to determine statistical significance. Logistic regression was used to identify additional predictors of patterns of recurrence (local, nodal, and distant). The Cox proportional hazards model was used to fit multivariate overall survival and disease-free survival models. Because AJCC stage, Breslow thickness, and Clark level are highly correlated, only the AJCC stage was included of these three variables into the multivariate model.
Of >12,500 cases of cutaneous melanoma registered in the Duke Melanoma database, 129 were recorded as having DM. The first case of DM recorded was diagnosed in January 1980, well after the initial 1971 description was published by Conley et al.1 The most recent case included in this series was diagnosed in October 2002.
No. of patients
Characteristics of primary tumor
No. of patients
Location (n = 129)
Breslow thickness (n = 125)
Clark level (n = 125)
Mitoses (n = 55)
Neurotropism present (n = 55)
Ulceration present (n = 129)
S-100 positive (n = 60)
HMB-45 positive (n = 36)
Surgery for primary lesion and stage
No. of patients
Surgery (n = 129)
Wide local excision
Clinical margins (n = 75)
Final margins (n = 129)
AJCC stage (n = 124)
Adjuvant therapy, recurrence, and survival
No. of patients
No adjuvant therapy
A total of 51 patients (39.5%) experienced a recurrence of their DM after surgery, with 18 cases recurring locally, 18 cases recurring first in the regional nodal basin, and 15 cases presenting with distant metastasis at first recurrence. At follow-up, 103 patients (79.8%) were alive, 24 patients (18.6%) had died as a result of their melanoma, and 2 patients (1.6%) had died secondary to cardiac events. Of the 24 patient deaths attributed to DM, 5 patients died with lung metastases, 4 with liver metastases, 3 with bony metastases, 1 with gastrointestinal involvement, 7 with multiorgan involvement, and 4 with unknown site of distant metastasis in the absence of any other attributable cause of death.
When the first recurrence was local (n = 18), the median time to recurrence was 1.0 years, and the median survival was 6.7 years from the time of recurrence. Nine of the 10 subjects with positive final pathologic margins had additional local recurrences, and 1 patient died 20 months after excision and was assumed to be dead as a result of melanoma, although no follow-up information about recurrence was available. Twelve of these 18 patients had subsequent recurrences: local (n = 5), nodal (n = 1), or distant (n = 6).
For the 15 patients whose first recurrence appeared in a distant location, the median time to recurrence was 1.7 years, and the median survival was 1.8 years from the time of recurrence. Sites of distant first recurrence were lung (n = 10), liver (n = 2), bone (n = 1), gastrointestinal (n = 1), and distant (n = 1).
Predictors of overall and disease-free survival: Kaplan-Meier univariate analyses
No. of patients (% total)
5-y disease-free survival
5-y overall survival
Breslow thickness (mm)
In multivariate analysis, older patients were more likely to have head and neck primary tumors, and they were more likely to be male. For overall survival, only age and AJCC stage remained significant predictors (P = .0212 and P = .0082, respectively). For disease-free survival, only AJCC stage and positive final margin remained significant predictors (P = .0046 and P = .0001, respectively).
Patterns of recurrence were also subjected to analysis with logistic regression. Features predictive of local recurrence included a positive final pathologic margin (P < .0001), more than four mitoses per square millimeter (P = .0031), a clinical margin ≤1.0 cm (P = .0137), and Clark level V (P = .0076). Age ≥60 years also approached statistical significance for correlation with local recurrence (P = .0553). For nodal recurrence, only male sex (P = .0569) and Breslow thickness >4.00 mm (P = .0850) approached statistical significance. Only AJCC stage (P < .0001) was found to predict distant recurrence.
Although the original slides from primary DM biopsy samples and excisions were not available for re-review, original pathology reports from outside hospital reviews, as well as reports from re-review of specimens after referral to our institution, were reexamined by a dermatopathologist (M.A.S.) for notable features. For the eight patients with positive lymph nodes at the time of primary diagnosis, four cases described neurotropic features, six cases described an epithelial component (“classical DM” by the Jain and Allen classification), and three cases demonstrated both epithelioid and spindle cells invading the dermis. For the 18 patients with lymph node involvement as the first site of recurrence, review of reports from the primary DM specimen revealed 7 cases with perineural invasion and 12 cases with an invasive epithelial component. Four reports described a mixed epithelioid and spindle cell pattern of dermal invasion.
Comparison with previous desmoplastic melanoma series and reviews
Current series (2005)
No. of patients
Median age (y)
Head and heck
High mitotic rate
Tumor thickness (mm)
81% at 2.5 y
72.2% at 5 y
73% at 1.5 y
76% at 5 y
59% at 3.3 y
68% at 5 y
62% at 5 y
55% at 5 y
The proportion of DM patients in this series who presented first with a local recurrence was 14%, which was lower than found in some series5,12 but still higher than that observed in other cutaneous melanomas of comparable thickness. Consistent with previous studies,4,10 we found the probability of local recurrence to be higher in patients with narrower excision margins (≤1.0 cm) by univariate analysis, but we did not find other factors, such as neurotropism, to be predictive of local recurrence. Patients with unrecognized microscopically positive margins at the time of primary excision were likely to require repeat wide local excision for disease recurrence, and microscopic margin positivity remained predictive of recurrence on multivariate analysis.
Also consistent with previous reports,10 the rates of nodal involvement at diagnosis and nodal metastases at recurrence were lower than historically observed for other cutaneous melanomas. However, in our cohort of patients, the development of nodal metastases after primary excision (14%) was equal to the rate of local recurrence (14%) as the site of first recurrence.
In our univariate analysis based on Kaplan-Meier survival estimates, we found tumor thickness, AJCC stage, Clark level, and age >60 years all to be predictive of overall survival. We did not find a high mitotic rate to be predictive of overall survival for patients with DM, and this finding was inconsistent with previous studies.6,10 In addition, neurotropism, male sex, and head and neck location were not factors that influenced overall survival, as were suggested in other articles.10
Comparison of desmoplastic melanoma with all stage I and II cutaneous melanomas from the Duke melanoma database (unpublished data)
Current series (%)
Cutaneous melanoma (%)
Overall survival (y)
Lung as site of distant metastasis
The explanations for the locally aggressive behavior observed in DM remain unclear. Previous studies have pointed to initial misdiagnosis, unrecognized positive margins, failure to re-excise positive margins, perineural extension, and insufficiently wide local excisions as possible factors involved in the high local recurrence rate of DM. Novel immunohistochemical stains for markers such as p75 nerve growth factor receptor, which has demonstrated high identification rates of DM and desmoplastic neurotropic melanoma, may become clinically useful for the assessment and identification of obscure positive margins.29–32 Immunohistochemical staining for expression of CD34, in addition to S-100, has also been suggested to confirm the diagnosis of DM.33 In addition, complementary DNA arrays may become clinically useful in the identification of gene expression patterns that could identify tumors more likely to demonstrate local aggression, nodal metastasis, or recurrence.34
DM presents as thicker lesions that may not be as amenable to narrow excision as other cutaneous melanomas as a result in a difference in local behavior and unique tumor biology. Because many of these lesions appear on the head and neck, it may be tempting for surgeons to use narrower margins of excision, even <1.0 cm, to achieve primary closure and an aesthetically superior result for patients. However, the results of our analyses suggest that narrower excision margins may increase the probability of local recurrence. Although excision margins did not predict decreased overall survival in our study, the additional morbidity of re-excision for local recurrence and the effect on patient quality of life should not be overlooked. In an era of narrowing excision margins for thicker cutaneous melanomas,35 it is important to examine the specific implications for the more locally aggressive DMs and apply these data to the management of DM with caution. In addition, the role of radiotherapy in these patients is being explored.11,36
A recent review of DM examined the experience with SLNB in 27 patients with DM of median thickness 2.2 mm at Memorial Sloan-Kettering Cancer Center. Out of 24 successfully performed SLNBs, there were no patients with a positive SLN by H&E or immunohistochemistry by using S-100 and HMB-45. Five of these patients presented later with local or distant recurrence; none of the subjects experienced regional nodal recurrence. The investigators suggested that because DM does not appear to be nodotropic, surgical staging of clinically negative regional lymph nodes at the time of the initial diagnosis and surgical excision of the primary lesion may be unnecessary.23 The earlier report by Thelmo et al.24 of SLNB in 16 patients with DM (mean tumor thickness, 3.9 mm) also identified no patients with a positive SLN.
However, investigators at the University of Michigan have reported their experience with SLNB in 33 clinically node-negative patients with DMs of median thickness 2.8 mm (mean, 4.0 mm). In this series, four patients (12%) were identified to have a positive SLN by using H&E and immunohistochemical stains (for S-100 and melan-A). Clinicopathologic factors such as tumor thickness, mitotic rate, head and neck location, ulceration, and neurotropism were compared between SLN-positive and SLN-negative patients, without statistically significant findings. The investigators suggested that in cases of DM, SLNB can provide important information for treatment decisions and can provide regional control when positive SLNB is followed by completion lymphadenectomy.25
Although our experience with SLNB has not identified any positive SLNs, we take caution in drawing conclusions from these negative cases about the application of SLNB to all patients with DM. Rates of nodal metastases at the time of primary diagnosis may be lower than those of other cutaneous melanomas, even though DMs tend to present with thicker lesions. However, the rate of nodal metastasis at the time of first recurrence is not negligible (14% of all patients in our series), and this suggests that the spread of micrometastatic disease to the lymph nodes is more common than we are observing at primary diagnosis by clinical assessment alone. For the patients who did not undergo previous examination of clinically negative lymph nodes at the time of diagnosis, it is uncertain what the rate of positive SLN would have been. However, 17% of patients in our series who were observed for clinically node-negative disease went on to have nodal metastases as the site of first recurrence. On the basis of the results of our experience, it may not therefore be appropriate to omit SLNB for clinically node-negative DM. It is also possible that the use of more sensitive assays (such as novel immunohistochemical stains and reverse transcriptase-polymerase chain reaction) in the examination of nodal tissue may identify more nodal micrometastases; whether these micrometastatic lesions would be clinically significant is unclear. In addition, new investigations in the field of tumor lymphangiogenesis may shed more light on the distinct regional behavior of DM and identify new predictors of lymphatic metastasis in these patients.37–41 The absence of in-transit disease at diagnosis and recurrence suggests that this tumor may interact differently within the lymphatic system than other cutaneous melanomas.
As with any retrospective review of patients from a single institution, the results of our analyses may not be representative of DMs in the general population. Although many patients were referred after diagnosis of melanoma, the dates of first diagnosis were confirmed from historical data, and pathologic re-reviews of available biopsy and surgical specimens were performed by our institution at the time of referral. Unlike other reported series,3,5 we did not distinguish the degree of neurotropism in our review of DM cases, nor were we able to draw conclusions about the prognostic significance of the presence or absence of pigmentation in our series.
Because many of the patients included in this study were referred from outside institutions for further management of their melanoma after initial excision, re-review of all primary specimens by using the original H&E slides was not possible. All the original pathology reports, as well as the reports from the examination of outside pathology materials by DUMC upon patient referral, were available for review. In addition, specific cases were selected for reexamination by a dedicated dermatopathologist for this study. However, we acknowledge that in the absence of formal histopathologic slide review of each of the primary DM lesions, the strength of our conclusions from pathology reports alone may be limited. In a similar context, it is difficult to draw conclusions from our review of operative notes and pathology reports for specific mention of surgical excision margins and microscopic tumor margins. Also, specimen shrinkage and microscopic extension of tumor cells beyond the gross borders of the DM lesion may have led us to underestimate the original excision margins taken by the surgeon.
In conclusion, DM is a distinct form of cutaneous melanoma with unique histopathologic features and clinical behavior. These characteristics should be considered when decisions are made about locoregional treatment strategies at the time of initial diagnosis and disease recurrence.
The authors thank Evelyn Nicholson, RN, and Nancy J. Pickett, RN, for their dedicated care of melanoma patients at Duke University Medical Center, for their significant contributions to the maintenance of the melanoma database, and for their assistance with this study. Supported, including statistical support and medical photography services, by the Duke Melanoma Research Fund.