Desmoplastic neurotropic melanoma

A clinicopathologic analysis of 128 cases


  • James Y. Chen MB, BS,

    Corresponding author
    1. Department of Radiation Oncology, Sydney Cancer Centre, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    • Department of Radiation Oncology, Royal Prince Alfred Hospital, Missenden Road, Camperdown NSW 2050, Australia===

    Search for more papers by this author
    • Fax: (011) 612-95158115

  • George Hruby MBChB,

    1. Department of Radiation Oncology, Sydney Cancer Centre, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    Search for more papers by this author
  • Richard A. Scolyer MB, BS,

    1. Department of Anatomical Pathology, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    2. Sydney Melanoma Unit, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    3. Discipline of Pathology, Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Rajmohan Murali MB, BS,

    1. Department of Anatomical Pathology, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    2. Sydney Melanoma Unit, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    3. Discipline of Pathology, Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Angela Hong PhD,

    1. Department of Radiation Oncology, Sydney Cancer Centre, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    2. Sydney Melanoma Unit, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    Search for more papers by this author
  • Patrick FitzGerald PhD,

    1. National Health and Medical Research Council Clinical Trials Centre, Camperdown, New South Wales, Australia
    Search for more papers by this author
  • Trang T. Pham MB, BS,

    1. Department of Radiation Oncology, Liverpool Hospital, Liverpool, New South Wales, Australia
    Search for more papers by this author
  • Michael J. Quinn MB, BS,

    1. Sydney Melanoma Unit, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    2. Discipline of Surgery, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • John F. Thompson MD

    1. Sydney Melanoma Unit, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    2. Discipline of Surgery, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author



Several studies have suggested that desmoplastic neurotropic melanoma (DNM) is associated with higher local recurrence rates than other types of melanoma. The authors investigated the local recurrence rates for patients with DNM after surgery alone or surgery followed by radiotherapy (RT).


One hundred twenty-eight patients with DNM were treated at the Sydney Melanoma Unit and the Sydney Cancer Center from 1996 to 2007. All patients underwent local excision, 27 patients also received RT. For both groups, clinical and pathologic features, treatment details, and local recurrence data were analyzed.


The median age at diagnosis was 65.5 years. The ratio of men to women was 2.7:1. The head and neck was the most common location (51%). The median Breslow thickness was 4 mm, and 99% of patients had Clark Level IV or V primary tumors. Patients who received adjuvant RT had thicker tumors (P = .003), deeper Clark level invasion (P < .001), and narrower excision margins (P < .001). There were 8 local recurrences, including 6 (6%) in the surgery only group and 2 (7%) in the adjuvant RT group. A positive margin (P < .001) and head and neck location (P = .03) were significant predictors of local recurrence.


The local recurrence rate in this series was lower than the rates reported in historic control groups and in the authors' previous temporal cohort. The results indicated that clear surgical margins are of paramount importance in minimizing local recurrence; when margins are compromised, the addition of RT may reduce local recurrence rates compared with historic controls. A prospective randomized trial is needed to quantify the risk reduction with adjuvant RT. Cancer 2008. © 2008 American Cancer Society.

Desmoplastic melanoma (DM) is a rare histologic subtype of melanoma that originally was described by Conley et al1 in 1971. DM accounts for approximately 1% of all melanomas and is characterized histologically by variably pleomorphic, spindle-shaped cells with associated collagen production. The tumor usually includes some hyperchromatic and elongated nuclei; however, they often are paucicellular with scattered, mildly atypical, spindle-shaped cells in abundant collagen. Under light and electron microscopy, the tumor cells resemble fibroblasts observed in a scar or fibrous tissue. Consequently, DM usually presents clinically as a nondescript plaque, nodule, or thickening that may or may not be pigmented. For these reasons, they often are misdiagnosed clinically and pathologically. In fact, many cases of DM are not diagnosed until they recur. Unrecognized DM was identified as a recurrent problem in a recent review of melanoma-related medical malpractice claims.2

Subsequent to the original description, Reed and Leonard3 reported 19 cases of DM with nerve infiltration, which they termed ‘desmoplastic neurotropic melanoma’ (DNM). This even rarer variant of melanoma accounts for 30% to 40% of DM.4, 5 In DNM, tumor cells may surround nerves (perineural invasion), directly invade nerve fibers (intraneural invasion), and may even exhibit the histologic appearance of nerve-like structures or nerve sheath tumors (neural transformation).6, 7 Although they often are aggressive locally, the prognosis for DNM appears to be better ‘stage for stage’ than for other subtypes of melanoma.8 Therefore, local control is particularly important. The reported local recurrence rates for DNM of 20% to 50% are higher than for nondesmoplastic melanomas.4, 9, 10 This may be related to the biology of the tumor, possible failure to recognize this entity early, or inadequate treatment once diagnosed.

Recently, it was suggested that DM can be divided into pure and mixed subtypes on the basis of the degree of desmoplasia within the tumor.5, 11, 12 Desmoplasia may be prominent throughout the entire tumor (‘pure’ DM) or may represent a portion of an otherwise nondesmoplastic melanoma (‘mixed’ or ‘combined’ DM).5 DMs of the ‘pure’ subtype reportedly have a higher rate of local recurrence yet superior overall survival compared with DMs of the mixed subtype and nondesmoplastic melanomas, whereas ‘mixed’ DMs have local recurrence rates and prognosis intermediate between ‘pure’ DMs and nondesmoplastic melanomas.5, 12

The previous largest single-institution review of DNM was published by Quinn et al from the Sydney Melanoma Unit (SMU) in 1998. For that study, 90 patients with DNM were accrued over the 10 years between 1985 and 1995. The local recurrence rate was 20% with surgery alone, higher than for other histologic types of melanoma. The role of adjuvant radiotherapy (RT) was not reported in that study. Although some of the published series discussed the use of adjuvant RT in DM,13 to our knowledge, none specifically addressed its role in the neurotropic subtype. In view of this, a single-institution, retrospective analysis was performed in a more contemporary cohort to describe the clinicopathologic features of DNM, to update outcomes, and to explore the role of adjuvant radiation treatment in the management of this entity.


One hundred twenty-eight patients who were diagnosed with DNM between 1996 and 2007 were identified by searching the databases of the Sydney Cancer Center Radiation Oncology Department and the SMU. Clinical records were reviewed to obtain patient demographics, primary tumor characteristics, radiation treatment details, and follow-up information. For patients who were referred to the SMU from other institutions, pathology slides and reports from external sites were reviewed individually by SMU pathologists who had expertise in melanoma (R.A.S., R.M.). If there was disparity between the external pathology report and the SMU pathologists' review, then the results of the SMU review were used for the purposes of this analysis.

Where possible, DM was classified into pure versus mixed subtypes according to the criteria proposed by the Memorial Sloan-Kettering Cancer Center (MSKCC).12 Ninety percent or more of the invasive tumor had to exhibit the typical paucicellular, fibrosing appearance for it to be classified as pure DM. Greater than 10% of desmoplasia was required for it to be classified as mixed DM. Neurotropism was defined as 1 or more of 1) tumor extension along nerves either perineurally or endoneurally, 2) formation within the tumor of structures resembling nerves, or 3) a change in the morphology of the tumor cells to resemble neural tissue. Neurotropic melanomas without desmoplasia were excluded from our series. Surgical margins were determined from pathology reports by using the measurement of the closest lateral or deep margin.

Recurrence was defined as local if it was described by the surgeon at or near the previous excision site (arbitrarily defined as <5 cm at the SMU). In transit metastasis was defined as any cutaneous or subcutaneous metastasis that occurred >5 cm from the original excision site, but not beyond the regional lymph node basin. Regional lymph node recurrence was defined as pathologically confirmed metastatic disease involving the draining lymph node basin. To determine the efficacy of adjuvant RT for DNM, the patterns of recurrence both within and outside the radiation field were determined. All RT treatment fields were documented using digital or film photography.

Of 177 patients who presented with stage I through III disease, 22 patients had a diagnosis of ‘probable’ DNM after SMU pathology review and were not included in the current analysis. An additional 27 patients who either had <6 months of follow-up or were lost to follow-up also were excluded.

The Wilcoxon rank-sum test, the Fisher exact test, and chi squared tests were used to estimate whether there was a statistically significant difference between the 2 subgroups in our cohort and to determine the predictors of local recurrence. A P value <.05 was considered statistically significant. The risk of local recurrence was estimated by odds ratio (OR) with 95% confidence interval (95% CI). This study was approved by the institutional Human Research Ethics Review Committee.


The median age at diagnosis was 65.5 years (range, 22-93 years). There were 94 men and 34 women, and the ratio of men to women was 2.7:1. The tumor location and staging are reported in Table 1. The head and neck region was the most common site of primary tumors, representing 51% of patients, followed by upper limbs (21.1%), and thorax (19.5%). Twenty-seven of 128 patients (21%) were referred for adjuvant RT after they underwent initial local excision. The majority of patients (91%) presented with American Joint Committee on Cancer (AJCC) stage I or II disease. Only 8% of patients had regional lymph node or in transit disease at presentation (ie, stage III melanoma). One patient could not be staged fully; he had a T2 tumor with unknown ulceration status.

Table 1. Patient Characteristics (N=128)
CharacteristicNo. of Patients (%)
  1. AJCC indicates American Joint Committee on Cancer.

Age, y 
 ≥6081 (63.3)
 <6047 (36.7)
 Men94 (73.4)
 Women34 (26.6)
Tumor location 
 Head and neck65 (50.8)
 Upper limb27 (21.1)
 Thorax25 (19.5)
 Lower limb9 (7)
 Abdominal wall and lower back2 (1.6)
AJCC stage 
 I10 (7.8)
 II107 (83.6)
 III10 (7.8)
 Unable to stage1 (0.8)

The median Breslow thickness was 4 mm (range, 0.9-45 mm). All tumors for which the Clark level was reported (127 of 128 patients) already had invaded the reticular dermis at the time of diagnosis, and 60% were Clark Level V tumors. Twenty-four patients had ulcerated tumors, 94 tumors were nonulcerated, and this feature was not recorded in 10 patients. The dermal mitotic rate was assessed in 126 patients and was low (<1/mm2) in 14 patients, intermediate (1-4/mm2) in 73 patients, and high (≥5/mm2) in 39 patients.14 The closest excision margin was documented in 123 patients. The margin was ≤1 mm in 13 patients, from 1.1 mm to 5 mm in 25 patients, from 5.1 mm to 9.9 mm in 36 patients, >1 cm in 41 patients, and positive in 8 patients.

Four patients had had DM at the same site and subsequently recurred with neurotropism. They all were treated with wide local excision followed by adjuvant RT.

There were several significant differences between patients who underwent surgery alone and those who also received adjuvant RT. The latter group had tumors with worse prognostic features (see Table 2). The surgery-alone group had a younger median age than the RT group (63 years [range, 22-93 years] compared with 74 years [range, 33-90 years], respectively). The ratio of men to women was much higher in the RT group (12:1). In the surgery-alone group, 39% of patients had DNM on the head and neck; whereas, in the RT group, all but 1 patient (thorax) presented with DNM involving the head and neck. The lip, cheek, and scalp predominated, constituting approximately 66% of the RT group. There were more early-stage tumors in the surgery-alone group. No patients in the RT group had stage I disease at presentation, although 10% in the surgery-alone group did. There was a significant difference in the median Breslow thickness between patients who underwent surgery alone and patients who also received RT (3.5 mm [range, 0.9-16 mm] compared with 7 mm [range, 1.6-45 mm], respectively). There were more T4 lesions in the RT group (80% compared with 41%), a higher percentage of ulcerated lesions (33% compared with 15%), and a high mitotic rate (≥5/mm2) was reported in 41% of patients in the RT group compared with 28% of patients in the surgery-alone group.

Table 2. Details of Patients Who Underwent Surgery and Adjuvant RT Versus Surgery Alone
CharacteristicNo. of Patients (%)P
Adjuvant RT, N=27Surgery Alone, N=101
  • RT indicates radiotherapy.

  • *

    Includes 12 patients who had margins ≥2 cm and 2 patients who underwent amputations.

Age, y  .06
 ≥6021 (77.8)60 (59.4) 
 <606 (22.2)41 (40.6) 
Sex  .011
 Men25 (92.6)69 (68.3) 
 Women2 (7.4)32 (31.7) 
Tumor location  <.001
 Head and neck26 (96.3)39 (38.6) 
 Limbs0 (0)36 (35.6) 
 Trunk1 (3.7)26 (25.8) 
Breslow thickness, mm  .003
 ≤4 (T1-T3)5 (18.5)60 (59.5)<.001
 >4 (T4)22 (81.5)41 (40.5) 
Clark level  <.001
 IV3 (11.1)48 (47.5) 
 V24 (88.9)52 (51.5) 
 Unknown0 (0)1 (1) 
Ulceration  .032
 Present9 (33.3)15 (14.9) 
 Absent16 (59.3)78 (77.2) 
 Unknown2 (7.4)8 (7.9) 
Mitotic rate, per mm2  .271
 <11 (3.7)13 (12.9) 
 1-415 (55.6)58 (57.4) 
 ≥511 (40.7)28 (27.7) 
 Unknown0 (0)2 (2) 
Closest excision margins  <.001
 ‘Clear’ but not quantified1 (3.7)4 (4) 
 Positive6 (22.2)2 (2) 
 ≤1 mm10 (37)3 (3) 
 1.1-5 mm4 (14.8)21 (20.8) 
 5.1-10mm6 (22.2)30 (29.7) 
 >1 cm0 (0)41 (40.6)* 

Finally, not only did patients in the RT group tend to have more advanced tumors, they also had less adequate surgical margins despite undergoing more surgery. Only 4 patients (4%) who underwent surgery alone required more than 1 wide local excision after initial excision biopsy to achieve the desired margins. In contrast, 9 patients (33%) in the RT group underwent more than 1 definitive excision, although the margin still was positive or was <0.5 mm in 3 of the 9 patients. Furthermore, 41% of patients had margins >1 cm in the surgery-alone group compared with no patients in the RT group.

The RT dose-fractionation schedules, treatment volumes, and modality depended on the treating radiation oncologist (see Table 3). Most patients who were treated earlier in the series (1996-2000) received 33 grays (Gy) in 6 fractions twice weekly, reflecting the popularity of hypofractionated regimes for melanoma at that time.15 Patients who were treated after 2000 generally received more conventional schedules using 2 to 2.4 Gy per fraction per day.

Table 3. Radiotherapy Details (N=27)
Dose, Fields, and ModalityNo. of Patients (%)
  1. Gy indicates grays; LNs, lymph nodes; MV, megavolt.

  2. *Did not receive the last fraction because of comorbidities.

Dose/fractionation schedule 
 28.5-40 Gy/5-10 fractions6 (22.2)
 48-50 Gy/20-25 fractions13 (48.1)
 54 Gy/27 fractions4 (14.8)
 60-64 Gy/30-32 fractions4 (14.8)
Treatment fields 
 Excision bed only15 (55.6)
 Excision bed and draining LNs8 (29.6)
 Including nerve to base of skull4 (14.8)
 Orthovoltage alone9 (33.3)
 MV photons alone7 (25.9)
 Electrons alone8 (29.6)
 Combined modality3 (11.1)

The median duration of follow-up was 40.5 months (range, 8 months-141 months; 48 months in the surgery-alone group vs 37 months in the RT group). The pattern of recurrence is recorded in Table 4. Eight local recurrences (6.3%) were observed, including 6 (5.9%) that in the surgery-alone group. The median time to local recurrence in this group was 10 months (range, 5 months-16 months). Two patients (7.4%) in the adjuvant RT group developed local recurrences within the radiation field. The first patient had a recurrence at the field edge 10 months after RT. The second patient had an in-field recurrence 5.5 years after RT, he developed an out-of-field cutaneous metastasis before the subsequent (late) ‘in-field’ recurrence. Five patients developed in-transit metastases, all in the surgery-alone group.

Table 4. Patterns of Recurrence
Site of RecurrenceNo. of Patients (%)
All Patients, N=128Surgery Only, N=101RT Group, N=27
  • RT indicates radiotherapy; NA, not available.

  • *

    Two patients in the RT group had recurrences outside the radiation field that were neither local recurrence nor in transit metastases.

Local recurrence8 (6.3)6 (5.9)2 (7.4)
In transit metastases5 (3.9)5 (5)NA*
Lymph node metastases20 (15.6)15 (14.9)5 (18.5)
Distant metastases20 (15.6)17 (16.8)3 (11.1)

The overall incidence of lymph node metastases was 15.6%. This included 10 patients who presented with lymph node disease (7 had positive sentinel lymph nodes [SLNs], and 3 had clinically palpable lymph nodes) and 10 patients who subsequently developed recurrent disease in regional lymph nodes. Twenty patients developed distant metastases, including 17 patients (16.8%) in the surgery-alone group and 3 patients (11.1%) in the adjuvant RT group. Lung and bone were the most common sites of distant diseases.


Our cohort was from the 10-year period after the original report by Quinn et al at the same center. The median patient age in the current series was 65.5 years, which was similar to the mean age of 63 in a previous literature review of 387 patients with DM and DNM. This confirms that DNMs occur in an older population when compared with a median age of 46 years for other melanomas. The ratio of men to women was approximately 2.7:1 in our series and was much higher (12:1) in the subgroup of patients who received adjuvant RT. The reason for the greater preponderance of men compared with the 1.3:1 to 2:1 ratios quoted in previous reports is unclear. Nearly half of our patients had tumors of the head and neck region. This is similar to previous reports in which head and neck was the most common site of DM and DNM. In contrast, only 19% of conventional melanomas were located in the head and neck region.16 The finding that, in the RT group, there was only 1 lesion outside the head and neck may reflect the greater difficulty in achieving adequate resection margins in the head and neck; thus, surgeons may be more inclined to refer patients for adjuvant RT.

In the previous study from the SMU, DMs were diagnosed if lesions were entirely desmoplastic or if they contained a recognizable desmoplastic component. Subsequently, both Payne et al17 and McCarthy et al8 applied a cutoff value when they described DMs as tumors with a desmoplastic component that represented >50% of the invasive dermal tumor. This definition was criticized later by Busam et al5 at MSKCC, who proposed a separation of DM into pure versus mixed subtypes because of differences in patient outcome. We were able to apply this classification to 50% of our patients, because the MSKCC criterion was introduced only in 2004. Of the 64 patients who were classified, 31 had mixed DM, and 33 had pure DM. There appeared to be a higher proportion of pure DNMs in the RT group (13 of 18 reported) than in the surgery-alone group (18 of 46 reported).

Not all neurotropic melanomas (NMs) are desmoplastic, and a subset of NMs are solid spindle or epithelioid cell tumors with minimal or no intratumoral fibrosis.18 Patients with these tumors were excluded from our study, because it is difficult to separate the prognostic significance of neurotropism versus that of desmoplasia, in that most series combined NM together with DM.19 We did not distinguish the degree of neurotropism in our review of DNMs. Only Smithers et al studied the prognostic importance of the degree of neurotropism. Neurotropism and neural transformation were associated with local recurrence on univariate analysis, but not on multivariate analysis. In another small series of 11 DMs, 11 DNMs, and 5 NMs, neurotropism was associated with statistically significantly worse survival and a higher local recurrence rate in melanomas with desmoplasia (60% 8-year survival in DNM vs 90% in DM). In other series,20–22 neurotropism was not associated with a worse outcome or a higher local recurrence rate.

Tumor Characteristics

In contrast to the previous report by Quinn et al, in which no neurotropism was observed in tumors that measured <1.5 mm in thickness, 1 patient in our cohort had a T1 tumor, and another 2 patients had T2 tumors <1.5 mm thick, perhaps related to better recognition and earlier diagnosis by the pathologists. The median Breslow thickness in this study was 4mm, which was typical in other reports.10, 22 In contrast, the mean Breslow thickness was 2.5 mm in the previous SMU series, which included both DM and DNM. Furthermore, the median Breslow thickness was significantly greater in our RT subgroup (7 mm) than in any other series. This again suggests selection bias, in which more advanced tumors are most likely to be referred for adjuvant RT because of suboptimal excision margins and a presumed higher risk of local recurrence. Approximately 60% of our patients had Clark Level V invasion, which was higher than most other reports that combined DMs and DNMs. Tumor ulceration and the dermal mitotic rate are documented in Table 5, which provides a summary of this series compared with the literature.

Table 5. Literature Summary
VariableCurrent StudyQuinn 19984Busam 20045Posther 200620Livestro 200521Literature Review10
  • H & N indicates head and neck, NA, not available.

  • *

    Calculated from an entire cohort of 290 patients, including 190 patients with desmoplastic melanoma (DM) and 90 patients with desmoplastic neurotropic melanoma (DNM).

  • Calculated from an entire cohort of 92 patients, including 40 patients with DM and 52 patients with DNM.

  • Calculated from an entire cohort of 128 patients, including 73 patients with DM and 55 patients with DNM.

  • §

    Calculated from an entire cohort of 89 patients, including 39 patients with DM and 50 patients with DNM.

No. of patients12890525550387
Median age, y65.559.6*635564§63
Sex (men:women)2.7:11.75:1*1.7:11.8:11.7:1§1.2:1
Location, % H & N5141*58476666
Mean Breslow thickness, mm42.5*§4.71
Clark level, %      
Ulceration present, %20.420.8*2022NANA
Mitotic rate, %      
Recurrence rate, %      
 Regional15.6  1412§ 
 Distant15.6  1214§ 

Named Nerve Involvement

Because DNMs frequently are located in the head and neck region, such neurotropism can give rise to cranial neuropathies because of direct tumor extension. In our series, there were 4 patients who had DNM that involved named nerves, and they all received adjuvant RT covering the primary excision site as well as the involved nerve, including its route to the skull base. Three patients achieved local control and were alive at last follow-up, and 1 patient died of metastatic disease.

Excision Margins

In this study, the surgical margin was described in detail, because its importance was highlighted previously in 2 other Australian series, which reported increased local recurrence when excision margins <1 cm were used and a nonstatistically significant, 2-fold increase in local recurrence when comparing a 1-cm margin with a 2-cm margin.4, 9 To achieve this margin, a significant proportion of our patients underwent re-excision, including 33% of patients in the RT group versus 4% of patients in the surgery-alone group.21 This reflected the predominant head and neck location of our DNMs, in which margins were limited by anatomic and cosmetic constraints. Sixty percent of patients who were referred for RT had either a positive or very close margins (≤1 mm) compared with only 5% in the surgery-alone group. Because of the similar features of DM and scar tissue on light microscopy, the dermatopathologists at SMU routinely use S-100 immunohistochemistry to help identify residual tumor and determine the exact surgical margin with greater confidence.23 However, this is not always reliable, because benign spindle cells within dermal scar tissue also may stain positively for S-100.24 Arora et al22 studied a more contemporary cohort (which included 44 pure DMs and 21 DNMs) in which the margin was ≥1 cm in 90% of patients and 2 cm in 58% of patients, respectively. The local recurrence rate after surgery alone was only 4% (2 of 49 patients). In our study, 40% of patients in the surgery-alone group had margins >1 cm, and 70% had margins >5 mm. Although these margins are less generous than those used by Arora et al, our local recurrence rate was only slightly higher at 5.9%. None of our patients who underwent surgery-alone and had margins >1 cm developed a local recurrence. However, because of the small number of events, it was not possible statistically to determine what constitutes an adequate surgical margin in our study. The wide confidence interval reflects the small number of recurrences observed and confers uncertainty regarding the level of risk.

Local Recurrence

Patients in our study had a local recurrence rate of 6.3% (clinical details are documented in Table 6). The rate was 7.4% in the adjuvant RT group and 5.9% in the surgery-alone group. It is probable that patients in the RT group were at higher risk of local recurrence than patients in the surgery-alone group, because patients in the RT group generally had head and neck primary tumors with thicker tumor and advanced Clark level. Most noteworthy, the excision margin was considerably less robust in the RT group (60% of RT patients had margins ≤1 mm). Despite these adverse features, adjuvant RT appeared to produce local control similar to that produced by wide surgical excision.

Table 6. Details of Patients Who Developed Local Recurrence
Patient No.DNM LocationBreslow Thickness, mmClark LevelUlcerationDMR, per mm2Excision Margin, mmAdjuvant RT
  • DNM indicates desmoplastic neurotropic melanoma; DMR, dermal mitotic rate; RT, radiotherapy; NA, not available.

  • *

    This patient had biopsy-proven inferior alveolar nerve involvement.


Smithers et al analyzed 45 DNMs for factors that influenced local recurrence after surgery alone. These included incorrect pathologic diagnosis, excision margins <1 cm, head and neck primary, level V invasion, and thickness >4 mm. In Twenty-one of 129 DMs, published by Posther et al, positive margins or margins ≤1 cm, Clark Level V invasion, and tumor stage were factors predictive of local recurrence (Table 7). In our series, positive margins (OR, 30; 95% CI, 5.2-169.4 [P < .001]) and head and neck location (OR, 8.7; 95% CI, 1.1-71.8 [P < .033]) were significant predictors of local recurrence (Table 7). Patients with Clark Level V invasion (OR, 6.3; 95% CI, 0.76-51.8), tumors >4 mm thick (OR, 4.23; 95% CI, 0.84-21.2), and ulcerated tumors (OR, 1.33; 95% CI, 0.25-7.1) also appeared to have a higher risk of local recurrence, although this was not statistically significant (P = .079, P = .085, and P = .664, respectively). To our knowledge there is little information in the literature regarding the local control rate after RT, especially in the setting of suboptimal excision, to allow comparison. Vongtama et al13 reported no local recurrences among 15 patients with DM who received adjuvant RT compared with a rate of 48% for patients who underwent with surgery alone. However, it was not clear how many of those 15 patients had neurotropism, and surgical margins were poorly quantified. These high rates of recurrence raise concerns regarding the adequacy of their surgery and/or pathologic assessment. In our study, 6 patients with who had microscopically positive margins and 10 patients with margins ≤1 mm were referred for RT, and only 1 of those patients (1 of the 6 with positive margins) developed a local recurrence after adjuvant RT. It appears that adjuvant RT produces local control rates similar to those produced by adequate surgical excision when the latter cannot be achieved. However, as mentioned above, because of the small number of local recurrences in both groups and the significant disparity in risk features between the 2 groups, we were not able to make a better quantification of the risk reduction from adjuvant RT.

Table 7. Analysis of Factors Predictive of Local Recurrence
VariableOR (95% CI)P
  1. OR indicates odds ratio; 95% CI, 95% confidence interval.

Positive excision margin29.6 (5.2-169.4)<.001
Head and neck location8.7 (1.1-71.8).033
Clark level V invasion6.3 (0.76-51.8).079
Breslow thickness >4 mm4.23 (0.84-21.2).085
Ulceration1.33 (0.25-7.1).664

Regional Lymph Node Involvement

There are conflicting reports regarding the risk of regional lymph node metastases in patients with DM. Even less information is available for patients with DNM, because most series have combined DM and DNM. The incidence of lymph node metastases reportedly is lower in DMs compared with other thickness-matched cutaneous melanomas.4, 25, 26 When comparing case-matched controls, in which the incidence was 20%, the overall incidence was 12% in Livestro's cohort of 89 DMs (including 50 DNMs).21 The lower incidence of lymph node metastasis suggests a different biology and natural history for DM compared with conventional melanoma,27 in which thicker tumors were more likely to be associated with regional lymph node involvement. Approximately 8% of patients in the current study presented with positive regional lymph nodes, which was higher than the 4% to 5% in the cohorts reported by Quinn et al and Livestro et al, reflecting the thicker tumors in our series. Our overall incidence of lymph node metastases was 15.6%, which was similar to that reported in the literature.

Currently, it is common practice for patients with AJCC stage IB or higher melanoma to undergo a SLN biopsy.28 Because of the low incidence of lymph node involvement in DM, the benefit of SLN biopsy has been questioned by the group from MSKCC,25 whose cohort of 24 patients with DM had no positive SLNs. A similar observation was made in the Mayo Clinic29 series of 59 patients with DM. Consequently, Busam et al no longer recommend performing SLN biopsies for patients with pure DM. The SMU policy was to offer SLN biopsy to all patients with primary melanomas ≥1 mm in Breslow thickness, mostly in the context of a clinical trial protocol. This included patients with DMs and DNMs. Some patients with tumors <1 mm in Breslow thickness also were offered SLN biopsy if there were unfavorable features, such as ulceration, Clark Level IV invasion, a high dermal mitotic rate, or young age (<45 years), because the rate of SLN positivity is considerably higher in younger patients. Some patients declined SLN biopsy, whereas others were not offered it because of medical comorbidities, advanced age, clinically involved lymph nodes, or elective treatment of the lymph node basin (either surgery or RT). Some patients were participants of the Multicenter Selective Lymphadenectomy Trial 1 (MSLT-1) trial and were randomized to undergo local excision alone. Rarely, the SLN could not be identified.

Overall, SLN biopsy was performed in only 55% of patients in the current study, 10% of whom had positive SLNs and proceeded to surgical dissection. Those with negative SLNs were observed, and 4.3% of these patients subsequently developed a recurrence in the same lymph node basin during follow-up. This rate is similar to the rates observed in MSLT-130 and in the MSKCC series.31 Eighteen patients with ‘pure’ DNM underwent SLN biopsy with only 1 positive result. Although pure DNM may carry a lower risk of nodal disease than other cutaneous melanomas, some would regard a 1 in 18 risk (5.6%) as sufficient to justify SLN biopsy. We do not entirely agree with Busam et al that SLN biopsy is not needed in patients who have pure DNM, but we do agree that the likelihood of a positive SLN is much lower than it is for other histologic subtypes of cutaneous melanoma. This needs to be explained to patients with DNM who are offered SLN biopsy.

Distant Metastasis

The incidence of distant metastases was 15.6% in the entire cohort; 16.8% in the surgery-alone group, and 11.1% in the RT group. The median time to distant failure was 18 months (range, 7-74 months). Lung, bone, and liver were the most common sites of distant metastases, similar to other melanomas.32 Although our overall incidence of distant failure was comparable to that reported in published series, the lower rate of distant disease noted in the RT group was surprising, because adjuvant local treatment was not expected to influence distant failure. Furthermore, the RT patients had thicker and more ulcerated DNMs with a higher dermal mitotic rate and should have been at greater risk of distant failure. The discrepancy may be attributed to the significantly older age in the RT group, in which patients may have died from other, competing medical comorbidities (2 of 27 died of nonmelanoma causes) before they developed distant metastases from their melanoma. The median duration of follow-up in the RT group also was shorter than that in the surgery-alone group. In the series reported by Busam et al,5 patients who had pure desmoplasia had better survival compared with patients who had mixed DM. The difference in prognosis was most pronounced among patients who had thick, Clark Level V tumors, which represented the majority of DNMs in the RT group. A significant proportion of pure DMs (13 of 18 reported) in the RT group also may explain the lower rate of distant metastases.

In conclusion, DNMs often are amelanotic and cytologically bland, leading to initial clinical and pathologic misdiagnosis in approximately 33% of patients and advanced lesions at presentation.9 The difficulty in the examination of margins without the use of immunohistochemistry and the neurotropic nature of infiltration can lead to inadequate surgical excision. These factors contribute to the historically high local recurrence rates of up to 50%.10 The local recurrence rates are lower in more contemporary series,20–22, 25 presumably because of better pathologic assessment and more adequate surgical excision. However, the cohorts in these reports were quite heterogeneous and included both DNMs and DMs.

To our knowledge, the current study represents the largest single-institution review of DNM alone published to date, with detailed assessment of margins and a strict definition of the pathologic features of desmoplasia and neurotropism. Our local control rate in the surgery-alone cohort was excellent and was similar to recent reports that included both DMs and DNMs. It is clear that patients who were referred for adjuvant RT had significantly more advanced tumors and narrower margins. Despite these adverse features, the local control rate for the RT group approximated that of the surgery-alone group. However, because of the imbalance of the 2 groups, it was not possible to quantify the efficacy of adjuvant RT accurately in this report. Although our results do not support the routine use of adjuvant RT, prospective data are needed to clarify the role of RT in circumstances such as close surgical margins (<1 cm) or tumor in the head and neck location, where the risk of local recurrence is high. Such a randomized, controlled trial to evaluate the role of adjuvant RT in this subtype of melanoma is being proposed currently by the Trans Tasman Radiation Oncology Group33 and the Australia-New Zealand Melanoma Trials Group.