Assessment of the role of sentinel lymph node biopsy for primary cutaneous desmoplastic melanoma


  • The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.



The role of sentinel lymph node biopsy (SLNB) in the treatment of desmoplastic melanoma (DM) remains undefined. The purpose of this study was to evaluate the use of SLNB for DM.


In all, 1850 patients with cutaneous melanoma underwent wide local excision and SLNB. Patients with DM were identified and stratified as ‘pure’ DM or ‘mixed’ DM (i.e., DM associated with at least one other common histologic subtype).


Of the 1850 patients, 65 (3.5%) had DM. Of these, 46 (70.8%) had pure DM and 19 (29.2%) had mixed DM. Patients with pure DM had a median tumor thickness of 3.5 mm and 6.5% were ulcerated. Compared with patients with pure DM, patients with either mixed DM or non-DM (n = 1785) had thinner primary tumors (median, 1.7 mm and 1.5 mm, respectively, each P < 0.001 vs. pure DM) that were more likely to be ulcerated (27.7% and 21.3%, respectively, each P < 0.05 vs. pure DM). Although the incidence of a positive SLN was similar in patients with mixed DM (15.8%) and non-DM (17.5%), patients with pure DM were less likely to have a positive SLN (2.2%) (each P < 0.01 vs. non-DM and mixed DM). At a median follow-up of 2.9 years, no patient with pure DM had recurred.


Despite having thicker primary tumors, patients with pure DM have a lower incidence of positive SLNs compared with patients with non-DM. Whereas the treatment approach for patients with mixed DM should be similar to that of other melanoma patients, patients with pure DM are unlikely to have metastatic disease in regional lymph nodes and SLNB may not be warranted. Cancer 2006. © 2006 American Cancer Society.

In 1971, Conley et al.1 described desmoplastic melanoma (DM) as an uncommon histologic variant of melanoma that was characterized by an unusual spindle-cell tumor that produced abundant collagen. DM has also been characterized by the presence of fusiform melanocytes dispersed in a prominent collagenous stroma,2, 3 be present at a more locally advanced disease stage,4–6 and to have a higher incidence of local recurrence4, 6 than non-DM. Some authors have reported that DM rarely metastasizes to regional lymph nodes,6–8 whereas others have reported lymph node metastasis rates of 8–15%.3, 4, 9–15

Over the past decade, lymphatic mapping and sentinel lymph node (SLN) biopsy (SLNB) has been adopted as a minimally invasive method of identifying occult metastatic disease in the regional nodal basin in patients with primary cutaneous melanoma.16–18 Several studies have confirmed that when there are nodal metastases the SLNs are the first nodes to be involved, suggesting that the pathologic status of the SLNs reflects that of the entire regional nodal basin.18–21 Only a few studies,6–8, 15 however, have investigated the use of SLNB for patients with DM. Gyorki et al.7 and Thelmo et al.8 reported that in patients who undergo a successful SLNB, less than 4% of SLNs are positive for disease. On the contrary, Su et al.15 reported a substantially higher rate of positive SLNs (12%). The rate of a positive SLN in patients with DM, therefore, remains unclear.

Histologically, DM may display morphologic heterogeneity.22 Specifically, some DMs are characterized by a uniform desmoplasia that is prominent throughout the entire tumor (‘pure’ DM), whereas other DMs appear to arise in association with other histologic subtypes (‘mixed’ DM). Distinguishing the phenotypic heterogeneity of DMs has been reported to be important for stratifying patients with regard to prognosis.22–24 To date, no studies have assessed the risk for SLN involvement on the basis of primary tumor histologic subtype (i.e., pure DM vs. mixed DM).

In the current study, we analyze our experience with DM in patients treated at a large academic cancer center. The purpose of the study was to define the incidence of DM and evaluate the results of SLNB in patients with DM. Furthermore, we sought to analyze the effect of DM histologic subtype (pure vs. mixed) on the incidence of positive SLNs at the time of selective lymphadenectomy.


Data Collection

The medical records of 1850 patients with primary cutaneous melanoma were reviewed by database analysis. The patients had all undergone sentinel lymphadenectomy and wide local excision at The University of Texas M. D. Anderson Cancer Center between November 1992 and October 2003. The variables assessed were demographic information (e.g., age and gender), primary tumor characteristics (e.g., histologic subtype, site, tumor thickness, Clark level, and presence of ulceration), SLN pathologic status, and recurrence.

Clinical Management

Evaluation of the regional nodal basin was performed using the SLNB technique as previously described.19, 25 In general, most patients underwent preoperative lymphoscintigraphy in which technectium99m (99mTc)-sulfur colloid was intradermally administered to establish lymphatic drainage patterns and identify those basins at risk for metastatic melanoma. Intraoperatively, 1–3 mL of isosulfan blue dye (Lymphazurin 1%; Hirsch Industries, Richmond, VA) was injected intradermally around the intact tumor or biopsy site. An SLN was defined as one that localized blue dye and/or concentrated radiolabeled colloid within a regional nodal basin.

Local control of the primary melanoma was achieved using wide excision of the tumor down to the level of the deep fascia with a margin of normal-appearing skin. The size of the surgical margin was dictated by the tumor thickness of the primary tumor. For thin melanomas (less than 1 mm) a wide excision margin of 1 cm of normal-appearing skin was achieved, whereas a 2-cm margin was routinely excised for melanomas greater than 1 mm. Adjuvant radiation therapy to the primary site was administered using an appositional field of 6 MeV electrons to a total dose of 30 Gy in five fractions over 2.5 weeks at the discretion of the treating physician.

Pathologic Assessment

All pathology slides were reviewed by the Department of Pathology at the M. D. Anderson Cancer Center. Tumors were considered to be DM if there was a predominance of spindle cells within a fibromyxoid stroma. DMs were stratified into ‘pure’ and ‘mixed’ subtypes. Pure DMs were defined as tumors in which at least 90% of the invasive melanoma had desmoplastic features. Melanomas in which less than 90% of the invasive component had desmoplastic features were defined as mixed DMs. SLN specimens were examined following an institutional protocol.26 Until 1999, SLNs were breadloafed and submitted in their entirety, followed by examination of three serial hematoxylin and eosin-stained slides. Beginning in 1996, step-sectioning was also performed. When necessary, immunohistochemical studies with anti-S100 protein or HMB45 were performed on all suspicious nodes. After 1999, the protocol was changed to include two serial hematoxylin and eosin-stained slides and an immunohistochemical analysis with either HMB45 or anti-MART1.26

Statistical Analysis

Clinicopathologic factors, including age and gender, Breslow thickness, ulceration, Clark level, anatomic site, and lymph node involvement were assessed. The association of categorical clinicopathologic factors with melanoma histologic subtype was evaluated using the chi-square or Fisher exact test, as appropriate. Factors with a significant univariate association (P < 0.05) were further evaluated in multivariate models.27 Cox proportional hazards models were used for multivariate analysis with forward stepwise selection of the variables. In addition, Kaplan–Meier survival curves and the log-rank test were used to compare disease-specific survival among various DM subgroups. All reported P-values were two-sided.


Clinicopathologic Factors

Of the 1850 patients in the M. D. Anderson Cancer Center database who underwent wide local excision and SLNB, 65 (3.5%) were diagnosed with DM. There were 37 men and 28 women (1.3:1 ratio). The median age of the 65 patients with DM was 61 years (range, 17–90 yrs), and their primary tumors were more likely to be located on the head and neck than non-DM patients (P < 0.05) (Table 1). The median tumor thickness of the DM tumors was 2.9 mm (range, 0.6–28.0 mm) and the majority were either Clark level IV (55.4%) or V (35.4%); 12.3% were ulcerated.

Table 1. Clinicopathologic Factors of Patients (n = 1850)
CharacteristicHistologic subtypeP valuea
Non-DMMixed DMPure DM
n = 1785n = 19n = 46
  • a

    P values refer to comparison between pure DM versus other melanomas (i.e., non-DM and mixed DM).

  • b

    P value refers to comparison between non-DM versus other melanomas (i.e., mixed DM and pure DM).

  • DM: desmoplastic melanoma; NS: not significant.

Median age, yrs516461< 0.05
Male gender, %57.652.658.6NS
Primary tumor location, %    
 Head and neck5.127.821.3< 0.05b
Median Breslow depth, mm1.51.73.5< 0.001
 ≤ 1 mm, %27.621.10< 0.01
 > 1–2 mm. %38.336.823.9NS
 > 2–4 mm, %22.726.328.3NS
 > 4 mm, %11.415.847.8< 0.01
Clark level IV/V, %57.384.297.8< 0.01
Ulceration, %21.327.76.5< 0.05

Forty-six (70.8%) of the 65 patients with DM had pure tumor subtype, characterized by dense, fibrous stroma with a relatively hypocellular proliferation of spindle cells (Fig. 1). In 19 (29.2%) of the 65 DM tumors, the desmoplasia and fibrous stroma was only a partial or minor component of the invasive tumor—these tumors were classified as mixed DMs (Fig. 2).

Figure 1.

Low-power (×10) view of a pure desmoplastic melanoma (DM). Notice the pink, dense, fibrous stroma with a relatively hypocellular proliferation of spindle cells.

Figure 2.

Low-power (×10) view of a mixed histologic desmoplastic melanoma (DM) subtype. Notice the highly cellular aggregates of DM associated with an epithelioid and spindle non-DM.

Table 1 shows the clinicopathologic features of the patients with non-DM, mixed DM, and pure DM. The median age of patients with pure DM (61 yrs) and mixed DM (64 yrs) DM was higher than for non-DM tumors (51 yrs) (both P < 0.05). Patients with pure DM had a median tumor thickness of 3.5 mm and a 6.5% rate of primary tumor ulceration. In contrast, the mixed DM and non-DM tumors were thinner (median, 1.7 mm and 1.5 mm, respectively) than the pure DM tumors (each P < 0.001 vs. pure DM) and were more likely to be ulcerated (27.7% and 21.3%, respectively) than the pure DM tumors (each P < 0.05 vs. pure DM). Patients with mixed or non-DM melanoma were also significantly less likely to have melanomas that were Clark level V (16.5% and 6.3%, respectively) than were patients with pure DMs (48.4%) (each P < 0.01 vs. pure DM).

SLN Status

At least one SLN was identified in all 65 DM patients. Overall, 4 (6.2%) of 65 patients with a DM had a positive SLN. The pathologic status of the SLN was associated with the histologic subtype of the primary melanoma. Whereas patients with a non-DM or mixed DM histologic subtype had similar rates of positive SLNs (17.5% and 15.8%, respectively) (P = 0.43), the incidence of positive SLNs for patients with a pure DM was only 2.2% (each P < 0.01 vs. non-DM and mixed DM) (Table 2). Overall, only 1 of the 46 patients with pure DM had metastatic disease in the SLN.

Table 2. Sentinel Lymph Node Pathologic Status by Melanoma Histologic Subtype
SLN StatusHistologic subtypeP valuea
Non-DMMixed DMPure DM
n = 1785n = 19n = 46
  • a

    P values refer to comparison between pure DM melanoma versus other melanomas (i.e., non-DM and mixed DM).

  • SLN: sentinel lymph node; DM: desmoplastic melanoma.

Positive, %17.515.82.2< 0.01

Adjuvant Radiation Therapy

Adjuvant radiation therapy to the primary tumor site was used in 20.0% of patients with DM (mixed DM, 5 of 19 [26.3%] vs. pure DM, 8 of 46 [17.4%]; P = 0.41). Similarly, patients with tumors at least 4 mm thick received local radiation therapy more often than patients with tumors less than 4 mm. Specifically, 9 of 24 (37.5%) patients with a primary tumor at least 4 mm thick received local radiation compared with only 4 of 41 (9.5%) patients with tumors less than 4 mm thick (P = 0.01). None of the factors (age, gender, Clark level, site, or ulceration) was associated with having received radiation therapy.

Patterns of Recurrence

With a median followup of 2.9 years, 15.6% of patients with non-DM had experienced recurrence. The overall disease-free survival duration of patients stratified according to melanoma histologic subtype is shown in Figure 3. Patients with pure DM had a 3-year actuarial disease-free survival rate of 100% compared with 90.5% for patients with non-DM and 78.2% for patients with mixed DM (P = 0.004 compared with patients with pure DM).

Figure 3.

Disease-free survival rate based on the histologic subtype of melanoma. Patients with pure desmoplastic melanoma (DM) (n = 46) had a 3-year actuarial disease-free survival rate of 100% compared with 90.5% for patients with non-DM (n = 1785) and 79.1% for patients with mixed DM (n = 19) (P = 0.004 compared with patients with pure DM).

Thus far, of the 19 patients with mixed DM 4 (21.1%) have experienced recurrence (P = 0.39 vs. non-DM). No patient experienced a local recurrence. Two patients had a negative SLNB and subsequently recurred with lung metastases. Another patient with mixed DM had an initial negative SLNB in 1994 but subsequently recurred within the mapped nodal basin 8 years later. As a result, the patient's original SLN was examined more extensively using both immunohistochemistry and step-sectioning, which was not a standard practice in 1994. Subsequently, low-volume metastatic melanoma was observed. After this clinical recurrence, a formal lymphadenectomy was performed; 2 of 22 lymph nodes were positive for metastatic disease and extracapsular extension was also noted. Thirty months later, lung metastases developed. The final patient with mixed DM who experienced recurrence had an initially positive inguinal SLNB followed by a completion lymphadenectomy. Only 1 of 11 lymph nodes were positive for metastatic disease, and this patient subsequently developed in-transit metastases of the lower extremity.

In analyzing the 65 patients with DM, no patient with a pure DM has thus far recurred (P = 0.002 vs. non-DM). By univariate analysis, a Clark level greater than III (P = 0.03), positive SLN pathologic status (P = 0.02), and mixed DM subtype (P = 0.005) were significantly associated with a shorter disease-free survival in patients with DM (Table 3). By multivariate analysis, mixed DM subtype remained an independent predictor of recurrence (P < 0.001).

Table 3. Univariate Analysis of Characteristics Affecting Disease-Free Survival Rates of 65 Patients with DM
FactorP value
  1. SLN: sentinel lymph node; DM: desmoplastic melanoma.

Age, < 50 yrs vs. ≥ 50 yrs0.30
Thickness, < 4 mm vs. ≥ 4 mm0.08
Clark level, ≤ III vs. IV/V0.03
SLN pathologic status0.02
Radiation therapy0.18
DM histologic subtype, pure vs. mixed0.005


DM is a rare, atypical form of melanoma with a reported incidence of 2–4%.4, 22 In the current study, we found a similar incidence of 3.5% based on a large cohort of melanoma patients seen at The University of Texas M. D. Anderson Cancer Center. Because of the rarity of DM tumors, the natural history of DM has been difficult to define. However, several studies4, 6, 10, 14, 28, 29 have characterized the clinical behavior of DM as being distinct from non-DM tumors. Specifically, unlike non-DMs, DM tumors often present as innocuous, nonpigmented lesions11, 28 that are found in the head and neck region of individuals in their sixth or seventh decade.4, 11, 22, 28 In the current study, patients with DM tumors were more likely to be older than patients with non-DM tumors. Although head and neck tumors were more common among patients with DM tumors than non-DM tumors, a predilection for the head and neck region was not noted among patients with DM tumors. The lower incidence of head and neck tumors in the current study, however, may have been due to our institution's overall referral patterns.

DMs can represent a challenge to the pathologist, because they are often difficult to distinguish histologically from other benign and malignant neoplasms.5, 22, 30, 31 Busam et al.22 reported the importance of making a distinction between DM with prominent fibrosis (pure DM) from melanomas in which desmoplasia is only a partial component (mixed DMs). In the current study, we sought to examine DM with an emphasis on this histologic distinction.

Cumulative data in the literature suggest that DM commonly presents as a deeply invasive tumor at the time of diagnosis.4, 6, 7, 14 In general, our experience confirmed these findings. Patients with DM (pure or mixed) had a median tumor thickness of 2.9 mm, which was almost twice the median Breslow depth seen in patients with non-DMs (1.5 mm) (P < 0.001). DMs were also much more likely to present as Clark level IV or V compared with non-DM lesions (P < 0.01). Interestingly, when patients with DM were stratified as pure versus mixed, only pure DMs were more likely to be deeply invasive (Table 1). Patients with mixed DM tumors had a near identical median Breslow depth and Clark level as non-DM lesions. Non-DM and mixed DM tumors also had comparable rates of ulceration, in contrast to pure DM lesions, for which lower rates of ulceration were noted, which is consistent with previous reports in the literature.7

The incidence of regional lymph node involvement in patients with DM remains poorly defined, with reported rates of 0–15%.3, 4, 6–15, 23 Of these, a few series6–8, 15 have reported that the incidence of SLN metastasis is low in patients with DM. To our knowledge, however, no previous study of DM has examined SLN pathologic status in relation to the histologic subtype. In this study, the incidence of SLN positivity was much lower in DM than non-DM, but only for DMs that were classified histologically as pure. In contrast, the rate of SLN positivity associated with mixed DM (15.8%) was similar to that of non-DM (17.5%). The rate of positive SLNs for pure DM tumors was low (2.2%) despite the presence of at least one histologic feature typically associated with an increased risk of SLN involvement: tumor thickness. This may suggest that the behavior of pure DM is more akin to that of soft tissue sarcoma,6, 7 which has a low predilection for regional lymph node metastases.32, 33 The biology and natural history of mixed DMs, however, appear to be more analogous to non-DM lesions, as the rate of lymph node metastasis was comparable in these groups. The failure to make the distinction between pure and mixed DM may explain, in part, the seemingly incongruous reports in the literature on the biologic behavior and rate of lymph node metastases of DM.

Contrary to published reports detailing local recurrence rates of 11–55%,4–6, 10, 11, 14 no cases of local recurrence after resection of DM were observed thus far in our study cohort. This may be explained by the aggressive surgical resection that was routinely performed at our institution for all lesions 1 mm or thicker. Patients with these tumors all underwent a wide local excision with at least 2-cm margins. In an analysis of 280 patients with DM, Quinn et al.4 found that patients who underwent a wide local excision with a 2-cm margin had a significantly lower chance of local recurrence than patients with a 1-cm margin of resection. Gyorki et al.7 similarly demonstrated a low rate of local recurrence in their cohort of patients who had DM and who underwent wide local excision with at least 2-cm margins. Combined, these data suggest that, whenever possible, a margin of 2 cm or more should be attempted when resecting DM to maximize the chance for local control.

Some authors have advocated that adjuvant radiation therapy be used to assist in the local control of DM.34–36 These investigators have argued that DM is an aggressive local disease and that radiation therapy may reduce the risk for local recurrence. In the current study, because there were no local recurrences, it is impossible to be certain whether radiation therapy was helpful. The absence of local recurrences after appropriate surgical excision, however, argues against its routine application when wide local excision for the primary tumor is adequately performed.

Controversy also exists regarding the prognosis of patients with DM.1, 4, 6, 7, 22, 29, 37 Egbert et al.10 reported a recurrence rate of over 50% in their series of patients with DM, leading to the suggestion that DM is more aggressive than other types of melanoma. Whereas other authors1 have similarly reported a worse prognosis for patients with DM, the majority of studies14, 29, 38, 39 have described a better prognosis for DM compared with similarly staged conventional melanomas. Walsh et al.38 reported that DM may be associated with a more favorable outcome. Skelton et al.29 similarly reported longer disease-free survival duration for patients with DM compared with melanoma in general. In the few studies in which pure DM was differentiated from mixed DM, patients with mixed DM had a greater risk of death or metastatic disease than patients with pure DM.22, 23 The favorable prognosis for patients with pure DM was confirmed in the current study, as no patient with pure DM had recurred at the time of last followup. Only patients with a mixed DM histology had developed distant metastasis or locally advanced (in-transit) disease. The histologic subtype of DM, therefore, not only seems to be important with regard to the rate of lymph node metastasis, but overall disease-free survival duration as well.

Studies investigating the potential molecular differences among non-DM, mixed DM, and pure DM subtypes using proteomics and gene-expression profiling will be important in identifying the underlying biologic distinction of these histologic subtypes.40–42 Recently, investigators41, 42 have identified specific genetic differences in non-DM versus DM. By using an Affymetrix-based gene expression profiling platform and hierarchical clustering analysis, Busam et al.41 distinguished non-DM from DM among 10 primary cutaneous melanomas (5 non-DM, 1 mixed-DM, and 4 pure-DM). Genes up-regulated in DM included neurotrophic factors and genes involved in extracellular matrix production. Specifically, there was a high expression of clusterin in DM but not non-DM specimens. Interestingly, clusterin immunostaining revealed absent or weak staining for both non-DM and mixed DM, but strong staining for pure DM. Davison et al.42 recently found no evidence of a V599E BRAF mutation in 12 of 12 DM specimens, whereas the thymine to adenine missense mutation of the BRAF gene was detected in 23 of 57 non-DM specimens.

In summary, DM is associated with a low overall incidence of lymph node metastasis in patients undergoing SLNB. The histologic categorization of DM as pure or mixed is clinically relevant, because pure DM appears to have a distinct clinical biology (i.e., increased tumor thickness, advanced Clark level, low incidence of primary tumor ulceration, low incidence of SLN positivity, and recurrence) compared with mixed DM or non-DM melanoma subtypes. The results of the current study have important implications, as they suggest that the therapeutic strategies for treating DM should be modified based on whether the DM is a pure or a mixed variant. On the basis of our findings, the treatment approach for patients with mixed DM should be the same as that for patients with all other types of melanoma; patients with pure DM, however, are unlikely to have metastatic disease in regional lymph nodes and SLNB may not be warranted. Whereas a randomized trial of SLNB versus observation in patients with DM would be ideal, such a study is unlikely given the overall low incidence of DM. Further confirmation of our findings by other investigators is warranted to formalize definitive recommendations regarding SLNB in patients with DM.


The authors thank Christopher W. Schacherer, Ph.D., and Dana M. McClain for database development and support as well as the M. D. Anderson Cancer Center's Department of Scientific Publications for assistance with this article.