Desmoplastic Melanoma: A Pathologically and Clinically Distinct Form of Cutaneous Melanoma
Desmoplastic melanoma (DM) is a rare variant characterized by the presence of fusiform melanocytes in a sclerotic stroma. Pathologic heterogeneity within DM may account for the controversy regarding the clinical presentation and prognosis of DM compared with conventional melanoma (CM).
We identified 131 patients with a diagnosis of DM seen between 1979 and 2002. Tumors were categorized as either pure DM (pDM; n = 92), if desmoplasia was prominent throughout the entire invasive tumor, or mixed DM (mDM; n = 39), if fibrosis was well developed in only parts of an otherwise non-DM. Differences in clinical behavior among pDM, mDM, and CM (n = 3976) were examined.
Seventy-three percent of patients with DM had tumors >2 mm in depth, compared with 31% of patients with CM (P < .001). Regional nodal metastasis was uncommon in patients who presented with clinically localized pDM (1%) compared with those with mDM (10%) or CM (6%) (P < .05, pDM vs. CM). Five-year melanoma-specific mortality was lower for patients who presented with pDM compared with mDM (11% vs. 31%; P < .01). Patients with pDM and CM had a similar melanoma-specific mortality despite a 3-fold difference in median tumor depth (3.6 vs. 1.2 mm, respectively).
DMs can be divided into two subtypes based on a histological quantification of desmoplasia. Tumors with prominent fibrosis (pure subtype) are unlikely to disseminate to regional lymph nodes and are associated with a favorable outcome when compared with those with mixed desmoplasia or CM.
KeywordsDesmoplastic melanomaLymph nodeSurvivalRecurrence
Desmoplastic melanoma (DM) is a rare variant of invasive cutaneous melanoma that was first described by Conley et al.1 in 1971. In the original description of seven cases, five of the seven patients developed a local recurrence, and three of the seven patients developed lymph node metastases. These findings led the authors to characterize DMs as “...highly malignant stubbornly recurring and often metastasizing neoplasms.”1 In the last 30 years, investigators have reported more than 600 additional patients with DM,1–8 and several features of the tumor are consistently agreed upon. A high proportion of patients with DM have tumors of the head and neck region.1–8 In addition, the mean age of onset is consistently 10 years older than it is for patients with conventional melanoma (CM).1–8 Controversy remains, however, with regard to the risk of regional nodal metastasis in patients with DM. In addition, these tumors have traditionally been associated with a higher risk of local recurrence.1–8 Whether this is due to a true difference in tumor biology or is a consequence of misdiagnosis at presentation remains unclear. Most importantly, the implications of a diagnosis of DM compared with CM for patient survival are controversial.1–8
We describe our experience at Memorial Sloan-Kettering Cancer Center (MSKCC) with patients whose melanoma was reported to be desmoplastic, focusing on presentation, natural history, and clinical outcome. It became apparent that some melanomas showed a purer phenotype than others. In some, desmoplasia was prominent throughout the tumor (pure DM, or pDM). In others, it constituted only a part of an otherwise nondesmoplastic invasive melanoma (combined tumors are herein described as mixed DM, or mDM). We examined differences in clinical behavior and patient outcome between these two groups. We used our entire experience (n = 131) to describe the clinical characteristics of patients with DM, as well as pathologic features of the primary tumor. A smaller group of patients with DM who presented to our institution with a clinically localized lesion (n = 95) were used to evaluate differences in survival among patients with pDM (n = 69), mDM (n = 26), and CM (n = 1973). The second subgroup consisted of 88 of the 95 patients who presented with a clinically localized lesion and negative margins of resection at operation. These patients were used to evaluate the risk of local recurrence among those with pDM (n = 67) and mDM (n = 21) and to compare this risk with our experience with patients with CM. In this way, we avoided the confounding factor of referral bias when describing clinical outcome that is present in many prior reports.
The study was approved by the Institutional Review Board of MSKCC. We identified 137 patients seen at our center with a diagnosis of DM between 1979 and 2002. Four patients were eliminated because the pathologic specimen was no longer available for review, and two patients were eliminated secondary to an inability to obtain sufficient clinical records, leaving a total study group of 131 patients. Since 1995, all patients with melanoma seen at MSKCC have been entered into a prospective melanoma database. Most patients were identified from this database (n = 90). Forty-one additional patients were identified through our institutional pathology database (1979–1995). Data for analysis included computerized or archived patient records. Invasive cutaneous non-DMs were designated as CM and were obtained from the prospective institutional database between 1995 and 2002 (n = 3976).
Treatment for the primary melanoma over the study period consisted of a wide local excision. Most patients underwent a wide local excision with at least a 2-cm radial margin. One half of the DM patients in this study presented with a primary tumor on the head or neck. Given this tumor location, it was often not possible to achieve a circumferential margin of 2 cm, and an optimal surgical margin with preservation of essential structures was pursued. Adjuvant radiotherapy was not routinely used, but it was used in a small number of selected patients with close or positive margins. These patients were excluded from the analysis of local recurrence. Too few patients received radiotherapy to make any meaningful comparisons.
For the purpose of this study, margins were considered negative if the pathologist reported that the microscopic margins were clear. In several cases, a diagnosis of melanoma was not made at the initial excision, and patients underwent inadequate local excision. Most of these patients presented to our institution after re-resection of the local recurrence revealed a melanoma.
The management of the regional lymph node basin evolved over the 23 years of the study period. For this study, patients reported as having a positive regional lymph node included those with pathologically involved nodes on elective or therapeutic lymph node dissection or a pathologically involved node on sentinel lymph node biopsy.
To avoid referral bias, survival and recurrence analyses were performed only on the subgroup of patients who presented for management of primary melanoma without local recurrence or evidence of regional lymph node involvement within 3 months of diagnosis. These patients were defined as “primary presenters” and consisted of 95 DM patients and 1973 CM patients. Within the DM primary presenters, seven patients had a very close or positive margin after definitive resection at MSKCC. Although these seven patients were included in the survival analysis, they were excluded from the local-recurrence analysis. In this series, the margin in question was always the deep margin, and no patient was left with a positive radial margin. Associations between groups were assessed by using χ2, trend, or t-tests where appropriate. Differences were considered significant at the P < .05 level. Melanoma-specific mortality (MSM) and recurrence were estimated with a competing risk approach.9,10 DMs occur in an older population than CMs. To account for the difference in ages, we used MSM for all statistical comparisons. Comparing MSM ensures that only deaths attributed specifically to melanoma are used, whereas deaths due to other causes are censored. MSM was compared among pDM, mDM, and CM by using a modified χ2 test.9,10 All survival curves and recurrence data are calculated from initial presentation at MSKCC.
Unless otherwise specified, all results reported refer to the entire group of patients (n = 131) with DM.
Clinical and pathologic characteristics of patients with pure and mixed desmoplastic melanoma and conventional melanoma (all patients)
No. of patients
Age, y, mean (median)a
Sex, n (%)
Primary tumor site, n (%)
Head and neck
Tumor depth, n (%)a
Tumor depth, mm, mean (median)
Nodal status, n (%)
Systemic metastasis, n (%)
Primary Tumor Characteristics
DMs (pDM and mDM) were much more likely than CM to arise in the head and neck (50% vs. 17%; P < .001). Median tumor depth was significantly greater in patients with DM compared with patients with CM. This was true for both pDM (3.6 mm) and mDM (5.5 mm) compared with CM (2.1 mm). Most patients (73%) with DMs presented with T3 or T4 lesions11, compared with only 31% of those with CM (P < .001). There was no significant difference in the incidence of ulceration, which ranged from 20% to 40%, between patients with pDM and mDM. Other features of the primary tumors were considered in an attempt to identify additional factors that influence clinical behavior. Neurotropism was identified in 55% of DMs. There was no difference in the incidence of neurotropism between patients with pDM and mDM. There was a high incidence of clinically amelanotic primary tumors in patients with DM. The incidence of an amelanotic primary lesion, as documented in the clinical record, was not significantly different when pDM (43%) was compared with mDM (38%), but it was higher than the reported incidence of 2% to 8% for CM.12–14 Patients with pDM were incorrectly diagnosed on initial biopsy or excision 28% of the time. This was significantly higher than the misdiagnosis rate of 10% in patients with mDM (P < .01). When initially misdiagnosed, DMs were most often considered benign and confused with blue nevi, hypertrophic scars, or “fibromas.” Occasionally they were confused with soft tissue sarcoma or squamous carcinoma. In many cases, the correct diagnosis was not made until resection of a recurrence.
Regional and Systemic Metastasis:All Patients (n = 131)
Local recurrence occurred at some point in the clinical course in 23 (26%) of 92 patients with pDM and 15 (39%) of 39 patients with mDM. This includes patients who experienced local recurrence after adequate wide excision but also includes patients who presented to MSKCC at the time of recurrence and those who were previously misdiagnosed and inadequately treated for their initial lesion. At presentation to MSKCC, there was a significantly lower rate of lymph node involvement in patients with pDM (1%) compared with those with mDM (18%; P < .001) or CM (30%; P < .001) (Table 1). In the entire group of patients with pDM (n = 92), only 2 patients (2%) either presented with or developed a regional nodal basin recurrence, compared with 17 (44%) of 39 patients in the mDM group. The high rate of lymph node involvement at presentation in the CM group is consistent with the tertiary cancer referral status of MSKCC. Despite this, lymph node involvement was extremely uncommon in patients with pDM. It should be noted that the single patient in the pDM group with a positive regional lymph node at presentation to MSKCC had stage IV melanoma with pulmonary metastasis. Only two patients with DM—one in the pDM group (as noted previously) and one in the mDM group—presented to MSKCC with systemic disease (Table 1). This is significantly different from our experience with CM, for which 13% of patients presented to MSKCC with stage IV disease (P < .001).
Patterns of Recurrence in DM PatientsWho Presented With Clinically Localized Disease (Primary Presenters) Who Had NegativeMargins of Resection (n = 88)
Only patients who presented to MSKCC with clinically localized primary melanoma within 3 months of diagnosis and who underwent excision with negative margins were analyzed for patterns of recurrence. This consisted of 67 patients with pDM, 21 patients with mDM, and 1973 patients with CM (Table 2). The incidence of local recurrence was not significantly different between the two DM groups: it occurred in 5 (7%) of 67 patients with pDM and 1 (5%) of 21 patients with mDM. The incidence of local recurrence for DM (pDM and mDM) was significantly higher (6 of 88; 7%), however, than the 2% (43 of 1973) incidence observed in CM patients who presented with clinically localized disease (P < .001; Table 2).
Among primary presenters with pDM (n = 67), only 1 patient developed a regional nodal basin recurrence, and this coincided with the development of pulmonary metastasis. No patient who presented to our institution with pDM and clinically localized disease developed an isolated regional nodal recurrence during the follow-up period. In the 21 primary presenters with mDM, 2 (10%) of 21 patients developed regional nodal recurrence during the follow-up period. Regional nodal basin recurrence occurred in 123 (6%) of the 1973 primary presenters with CM. The difference in regional nodal basin recurrence between patients with pDM and CM was significant (P < .05).
Recurrence patterns in patients with clinically localized desmoplastic and conventional melanoma after optimal surgical resection
Type of melanoma
No. of patients
Pure desmoplastic melanomaa
Mixed desmoplastic melanomaa
Ninety-five patients who presented to MSKCC with localized DM were analyzed for survival. This included seven patients who where not analyzed for patterns or recurrence because although they had clinically localized disease, they had close or positive margins of resection. In this group were 69 patients with pDM and 26 patients with mDM. These patients were compared with 1973 patients with CM who presented with localized primary disease. Survival among primary presenters with DM was principally determined by the disease, as demonstrated by the 5-year overall survival of 78% and melanoma-specific survival of 82%. The median follow-up of survivors who presented with primary disease was 2.5 years for pDM patients, 3.3 years for mDM patients, and 1.9 years for CM patients. The MSM of primary presenters with pDM (n = 69) was compared with that of mDM (n = 26) and CM (n = 1973). MSM for patients who presented with pDM was 0% at 2 years and 11% at 5 years and was significantly more favorable compared with that for patients with mDM (13% at 2 years and 31% at 5 years; P < .01; Figure 2A). The MSM of primary presenters with pDM was similar to that of primary presenters with CM (Figure 2B), even though the mean depth of the primary tumor in the pDM group was 4.5 mm, compared with a mean of 2.1 mm in the CM group.
DM is a rare variant of invasive cutaneous melanoma; it represents well under 1% of patients referred to MSKCC. It most commonly arises in chronically sun-damaged skin. Rarely, DM may present on a mucosal surface. We did not include patients with mucosal presentation in this report because patients with mucosal melanoma have an outcome distinctly different from that of those with cutaneous disease.15
DM is characterized by the presence of fusiform melanocytes in a densely sclerotic stroma. Within this group there is a spectrum of desmoplasia. Some tumors are associated with prominent fibrosis throughout the entire invasive component, whereas others show only partial or minor desmoplasia. To determine whether tumors with prominent desmoplasia were associated with a clinical course different from that of tumors in which desmoplasia is only partial or minor, we classified them into pDM and mDM and compared them with each other, as well as with CM.
Consistent with previously published reports, patients with DM presented at an older age (median, 65 years) and were more likely to have tumors of the head and neck (50%) compared with patients with CM (17%). Other investigators reported that 64% to 67% of patients with DM were male.1–8 In both subgroups in our series, men were more likely than women to present with DM. This pattern was much more pronounced in mDM (90%) compared with pDM (61%).
Other investigators have reported that patients with DM were more likely than patients with CM to present with a deep primary lesion, have amelanotic tumors, and be initially misdiagnosed. Our experience confirmed these previous reports. Patients with both pDM and mDM presented with deep primary tumors (median, 4.1 mm), had a high incidence of misdiagnosis (15%), and were more likely to be amelanotic (40%). Given the clinical scenario described previously (an older patient population with an amelanotic lesion and common misdiagnosis), it is not surprising that patients with DM were more likely to have a local recurrence and present with a deep primary tumor as compared with patients with CM. The percentage of patients initially misdiagnosed may be partially due to referral bias, because several patients presented with a local recurrence after misdiagnosis at another facility. There was a declining trend in the incidence of misdiagnosis over the duration of this study. Although it is impossible to estimate the true incidence of misdiagnoses from our experience, it is likely to be significantly higher than that for CM.
In an attempt to define the clinical behavior of DM, several investigators have sought to subdivide these patients into groups, hoping to identify those with a better or worse prognosis. For example, Skelton et al.,3 using the Armed Forces Institute of Pathology database, found that the presence of stromal mucin was associated with a small survival advantage. In addition, they reported an improved overall survival for patients with DM compared with CM on the basis of a comparison with historical controls.3 In the largest experience published, Quinn et al. subdivided DMs into two groups based on neurotropism but found no differences in survival between those with and without perineural invasion.2 Compared with a similarly collected cohort of CM patients, the patients in this study did not seem to have a survival advantage.2 Quinn et al. reported fewer lymph node metastases compared with the group of patients with CM.2 This is consistent with our recently reported experience with sentinel lymph node biopsy in patients with DM.4 In this series, we included only patients with pDM. No patient was found to have a positive sentinel node out of the 27 who underwent sentinel lymph node mapping.4 In contrast, Su et al.16 reported 4 patients (12%) with a positive sentinel lymph node in 33 patients with DM. The criteria used by the authors to classify patients as having DM are not reported in this article, but representative pathologic sections are shown. It is quite possible that patients with a positive lymph node would have been classified as having mDM with our criteria.
Patients with pDM were extremely unlikely, in our experience, to present with or develop involved regional lymph nodes. The single patient with a positive lymph node at presentation also had lung metastases. During surveillance, one additional patient developed a clinically positive lymph node and did so in the presence of disseminated disease. In patients with mDM, lymph nodes were more commonly involved at presentation (7 of 39 patients) and more commonly developed during surveillance (10 additional patients). When pathologically involved lymph nodes in patients with mDM were examined, they often revealed only the nondesmoplastic component from the primary tumor.
Patients with pDM melanoma had a statistically lower MSM than those with mDM in our series. MSM was similar when pDM was compared with CM despite the more than 2-fold difference in the Breslow depth of the primary tumor. There were no deaths due to melanoma in the pDM group when the primary was <2 mm. In the patients with pDM who developed a recurrence, the first site of recurrence was most likely to be distant. Our experience with patients with pDM is consistent with the survival advantage suggested by Skelton et al.3 It is probable that the disparate reports regarding DM’s ability to metastasize to lymph nodes and its relative survival advantage or disadvantage may be explained by the heterogeneity within this group of tumors.
We have previously suggested that sentinel lymph node biopsy may be unnecessary in patients with pDM.4 Although some may argue that the morbidity of sentinel lymph node biopsy is low, half of all pDMs occur in the head and neck, where sentinel node biopsy carries an increased risk of nerve injury. Our experience suggests that identification of patients with pDM will facilitate more appropriate clinical management and identify a subgroup of patients with relatively favorable survival.
Patients with pDM are unlikely to develop regional lymph node metastasis and seem to have a favorable prognosis despite advanced Breslow depth. Mixed DMs retain the ability to metastasize to regional lymph nodes and should continue to be treated like CM.