Postoperative radiotherapy for primary mucosal melanoma of the head and neck


  • Presented at the Fifth International Conference on Head and Neck Cancer, San Francisco, California, July 29–August 2, 2000.



Primary head and neck mucosal melanoma (HNMM) has a poor prognosis with a low local control rate and frequent distant metastases. The objective of the current study was to determine the impact of postoperative radiotherapy on local control and survival.


One hundred forty-two patients with primary HNMM treated between 1979 and 1997 were reviewed. Of these, 69 patients with confirmed primary mucosal melanoma, absence of metastatic disease, and definitive management by surgery with or without postoperative radiotherapy and follow-up at the Institut Gustave-Roussy (Villejuif) were selected. The site of primary HNMM was sinonasal in 46 patients, oral in 19 patients, and pharyngolaryngeal in 4 patients. Twenty-two patients (32%) had a locally advanced tumor (T3–T4) and 17 patients had regional lymph node metastases after pathologic examination (pN > 0). Thirty patients underwent surgery alone and 39 received postoperative radiotherapy. Patients with locally advanced tumors had received postoperative radiotherapy more frequently than those with small tumors (P = 0.02).


Thirty-seven patients (54%) experienced local disease recurrence and 47 patients (68%) developed distant metastasis. The overall survival rates were 47% at 2 years and 20% at 5 years. In the Cox multivariate analysis, patients with early T-classification tumors who received postoperative radiotherapy had a better local disease-free survival (P = 0.004 and P = 0.05, respectively) compared with patients with late T-classification tumors who did not receive postoperative radiotherapy. Patients with advanced T-classification and pN > 0 stage had a shorter distant metastasis disease-free survival compared with patients with early T-classification and pN < 0 stage. Patients with advanced T-classification tumors had a shorter overall survival compared with patients with early T-classification tumors (P = 0.003).


The prognosis of patients with HNMM was poor. Patients had a high rate of distant metastasis and a low rate of local control. The current study suggested that postoperative radiotherapy increased local control even for patients with small tumors. Cancer 2005. © 2004 American Cancer Society.

Primary mucosal melanoma of the head and neck (HNMM) is a rare disease. In an exhaustive review, Manolidis et al.1 found 1000 cases reported in the literature up to 1997. Having studied a population of 2.5 million individuals in Denmark over a period of 30 years, Andersen et al.2 found a prevalence of 0.8% among all individuals with melanoma and 8% among all patients with melanoma of the head and neck. During 1985–1994, the U.S. cancer registry3 recorded 84,836 cases of melanoma, 0.7% of which were HNMM (611 cases). Most of the publications reported on small series of patients over long time periods. However, these reports did not lead to a consensual approach to treatment nor did the authors of the reports undertake the study of clinicobiologic factors specific to these aggressive tumors.

The overall 5-year survival rate is low,4–10 usually < 30%. Most of the patients died of their disease, with sometimes very late disease recurrences. Primary HNMM is commonly treated by primary comprehensive excision2, 11–18 followed by radiotherapy even though no defined benefit has been clearly demonstrated in terms of local control and survival with this adjuvant radiotherapy.8, 10, 15, 19. Recently, Owens et al.10 did a comparable study at The University of Texas M. D. Anderson Cancer Center (Houston, TX) and showed that for 44 patients with HNMM, the addition of radiotherapy decreased the rate of local disease recurrence (P = 0.13), but did not significantly improve survival (P = 0.73).

The aim of the current study was to review the experience of 142 patients with primary HNMM treated at the Institut Gustave-Roussy (Villejuif, France). We focused on defining the effect of postoperative radiotherapy on survival and local control rates.



Between 1979 and 1997, 142 patients with HNMM were treated at the Institut Gustave-Roussy. Histologic diagnosis was made with conventional Fontana-Masson and/or hematoxylin eosin safran staining. Tissue specimens from 72 patients underwent immunohistochemical analysis with S-100 (Dako, Glostrup, Denmark), HMB45 (Dako), and antikeratin antibodies (CK7, CK20, and vimentin; Novocastra, New Castle-Upon-Tyne, United Kingdom). A pathologist reviewed the records of 43 of the 45 patients who were diagnosed before 1985 and was unable to confirm the diagnosis or changed the initial diagnosis of melanoma to that of a poorly differentiated carcinoma, rhabdomyosarcoma, angiosarcoma, or lymphoma for 10 patients, after further histopathologic and immunohistochemical analysis.

Most of patients had undergone a dermatologic examination in search of primary skin melanoma. We excluded six patients who had a past history of skin or ocular melanoma and two patients with hypopharyngeal and nasopharyngeal mucosal distant metastasis whose examination revealed a primary skin tumor. We also excluded 49 patients who received exclusively radiotherapy or palliative chemotherapy because of unresectable primary tumor, distant metastasis, or comorbidity. Also excluded were two patients who died of complications just after surgery and four patients who did not receive follow-up care.

Of the initial group of 142 patients, 69 were included in the data analysis comparing surgery alone with surgery and postoperative radiotherapy.

The mean age of the patients at the time of the diagnosis was 58 years (range, 21–90 years). There were 36 men and 33 women (a male-to-female ratio of 1:1).

Tumor extent was evaluated by clinical examination. Forty-two patients received a computed tomography scan and 22 patients received a magnetic resonance imaging scan. A chest X-ray and a liver ultrasound had been performed systematically. Fourteen patients also underwent a bone scintigraphy. There were 46 (67%) sinonasal tumors (31 tumors in the nasal cavity and 15 tumors in the paranasal sinus), 19 (28%) oral cavity tumors (one-half of which were located in the upper jaw), and 4 (6%) pharyngolaryngeal tumors.

We retrospectively staged each patient by primary site using the criteria established by the 1997 TNM classification.20 Forty-seven (68%) patients had an early-stage tumor (T1–T2) and 22 patients (32%) had a locally advanced-stage tumor (T3–T4). Sixteen patients (23%) had clinical neck lymph nodes. Patients with sinonasal tumors had significantly less clinical neck lymph nodes (5 of 46) than patients with tumors at other primary sites (11 of 23 [47%]; P = 0.0014). Six patients had melanocytic lesions associated with tumors in the oral (n = 2) and nasal cavities (n = 4).

Of the 69 patients, 59 (86%) had undergone wide surgery defined as a large open surgical resection with macroscopic and microscopic clear surgical margins. Limited surgery, defined as an endonasal technique for sinonasal tumors or closed or positive margins as judged by the surgeon or the pathologist, had been performed in 15 patients. The surgical technique had been chosen according to tumor site and extension. Pathologic analysis of the surgical margins was not available for two patients with old medical records but these patients were judged to have received wide surgery in the surgical report. Twenty patients underwent neck dissection, bilateral in 3 patients (6 patients received selective lymph node dissections and 17 received radical modified lymph node dissections), including the 16 patients with clinical neck lymph nodes. Seventeen patients had lymph node metastasis after pathologic examination (pN > 0) and were studied in the statistical analysis.

Thirty patients (43%) received surgery alone and 39 patients (57%) received surgery followed by postoperative radiotherapy. Twenty-nine patients received 70 gray (Gy) to the tumor bed and 10 patients received 50 Gy. Conventional radiotherapy (2 Gy per fraction once a day) had been prescribed, except for 2 patients who had received hypofractionated radiotherapy. In 24 patients including the 17 patients with positive lymph nodes, radiotherapy had been delivered to the neck. There was a significant relation between T-classification and pN stage and postoperative radiotherapy (Table 1). Patients with locally advanced tumors had received postoperative radiotherapy more frequently than those with small tumors (P = 0.02). No correlation was found between the use of postoperative radiotherapy and other covariates, especially the tumor site (sinonasal vs. others sites) and the surgical procedure (wide vs. limited surgery; Table 1). No patient had received additional treatment after radiotherapy.

Table 1. Characteristics of 69 Patients with Head and Neck Mucosal Melanoma
CharacteristicsNo. of patients receiving surgery alone (n = 30) (%)No. of patients receiving surgery and postoperative radiotherapy (n = 39) (%)P valuea
  • a

    Chi-square tests.

  • b

    Retrospectively staged by primary tumor site using the 5th ed. of the TNM classification of malignant tumors.20

Median age (yrs)  0.38
 < 5817 (57)18 (46) 
 > 5813 (43)21 (54) 
Gender  0.7
 Male15 (50)21 (46) 
 Female15 (50)18 (54) 
Sites  0.7
 Sinonasal20 (67)26 (67) 
 Oral cavity9 (30)10 (25) 
 Pharyngolaryngeal1 (3)3 (8) 
Local tumor extensionb  0.017
 Early stage (T1–T2)25 (83)22 (56) 
 Advanced (T3–T4)5 (17)17 (44) 
Pathologic neck lymph nodes  0.05
 pN026 (87)26 (67) 
 pN > 04 (13)13 (33) 
Surgery  0.7
 Limited6 (80)9 (77) 
 Wide24 (20)30 (23) 

Statistical Methods

Data were obtained by analyzing the medical charts and records. Univariate analyses using the Fisher exact test and Student t test were performed for categoric and continuous data, respectively. Estimates of survival curves were calculated according to the Kaplan–Meier, product-limit method and were calculated from the time of surgery. Overall survival, local disease-free survival curves, and distant metastasis disease-free survival curves were compared for various prognostic groups using the log-rank test. Predictor variables with P < 0.20 in the univariate analyses were submitted to Cox regression multivariate analyses. P < 0.05 was significant. Statistical tests were performed using Statview 5.0 (SAS Institute Inc., Cary, NC).


Tumor Control and Survival

The median follow-up period was 3.8 years (range, 8–384 months). Thirty-seven patients (54%) had local disease recurrence including 7 patients with early disease recurrences within 2 months after the end of treatment. Local disease-free survival, which was defined as the time from surgery to the time of local disease recurrence, ranged from 1 month to 134 months. It was associated with T-classification (P = 0.01 by log-rank test) and was independent of age, gender, primary tumor site, and type of surgery (wide vs. limited; P > 0.05 by log-rank test).

Sixteen patients (23%) had disease recurrence in the neck lymph nodes, without local disease recurrence in 5 patients. Two patients who were classified initially as N0, who did no undergo neck lymph node dissection, had disease recurrence in the neck. Neck lymph node disease-free survival was not correlated with primary site, T-classification, or pN stage (P > 0.05 by log-rank test).

Of 69 patients, 47 (68%) had distant metastasis during the follow-up after curative treatment. For 30 patients, distant metastasis was diagnosed during the first year after the end of treatment. Distant metastases occurred in all patients and were located in the lungs in 53% (25 of 47) of patients, in the bones in 36% (17 of 47) of patients, in the liver in 20% (9 of 47) of patients, in the brain in 20% (9 of 47) of patients, and in the skin in 20% (9 of 47) of patients. For these patients, the distant metastasis disease-free survival period ranged from 2 months to 4 years. Sixteen of 17 patients with pN > 0 had distant metastasis during follow-up. Patients with pathologic neck lymph nodes and advanced T-classification had a shorter distant metastasis disease-free survival (P = 0.0005 and P = 0.0008, respectively) compared no neck metastasis and early T-classification. Patients died within a median of 4 months (range, 1–11 months) after a diagnosis of distant metastasis.

Overall survival rates were 47% at 2 years and 20% at 5 years. Age, gender, and primary tumor site were not prognostic factors (P = 0.22, 0.22, and 0.86, respectively, by log-rank test). Early T-classification (T1–T2) and pN0 were associated with a better overall survival (P < 0.0001 and P = 0.007, respectively, by log-rank test).

Role of Postoperative Radiotherapy

The local control rates were 26% with surgery alone (8 of 30) and 62% (24 of 39) with postoperative radiotherapy (Table 2), even though there were much more locally advanced tumors (T3–T4) in the radiotherapy group (P = 0.02; Table 1). Patients who received surgery alone had a median local disease-free survival period of 9 months (range, 1–344 months) and patients in the postoperative group had a median survival period of 33 months (range, 1–103 months). Univariate analysis showed that postoperative radiotherapy was associated with better local disease-free survival (P = 0.04 by log-rank test; Fig. 1), as did Cox multivariate analysis (relative risk = 0.4; confidence interval, 0.2–0.9; P = 0.05) and this result was independent of the T-classification (Table 2).

Table 2. Cox Multivariate Analysis of Local Disease-Free Survival
FactorsLocal disease recurrence
Yes (n = 37)No (n = 32)Relative risk (95% Confidence interval)P
T-classification  3.7 (1.5–9)0.004
Treatment  0.4 (0.2–0.9)0.05
 postoperative radiotherapy1524  
Figure 1.

Kaplan–Meier time to local disease recurrence curves according to treatment (surgery alone vs. surgery and postoperative radiotherapy) in 69 patients with primary head and neck mucosal melanoma.

Patients in the postoperative radiotherapy group developed distant metastasis more rapidly than patients in the surgery group (P = 0.03 by log-rank test; Fig. 2), but only advanced T-classification and pN > 0 were associated with a shorter distant metastasis disease-free survival period in the multivariate analysis (Table 3). In contrast, patients in the surgery group had a better overall survival (median, 30 months; range, 5–392 months) than patients in the postoperative radiotherapy group (median, 17 months; range, 4–109 months; P = 0.004 by log-rank test; Fig. 3). Cox regression multivariate analysis of primary tumor site, T-classification, pN stage, and postoperative radiotherapy indicated that advanced T-classification was the only predictor of a shorter overall survival period (P = 0.003; Table 4).

Figure 2.

Kaplan–Meier distant metastasis disease-free survival curves according to treatment (surgery alone vs. surgery and postoperative radiotherapy) in 69 patients with primary head and neck mucosal melanoma.

Table 3. Cox Multivariate Analysis of Distant Metastasis Disease-Free Survival
FactorsDistant metastasis
Yes (n = 47)No (n = 22)Relative risk (95% Confidence interval)P
Neck lymph node invasion  2.9 (1.2–7.3)0.02
 pN > 0161  
T-classification  2.5 (1.2–5.3)0.02
Treatment  1.3 (0.7–2.7)0.4
 Surgery alone1911  
 Postoperative radiotherapy2811  
Figure 3.

Kaplan–Meier overall survival curves according to treatment (surgery alone vs. surgery and postoperative radiotherapy) in 69 patients with primary head and neck mucosal melanoma.

Table 4. Cox Multivariate Analysis of Overall Survival
FactorsRelative risk (95% Confidence interval)P
T1–2 vs. T3–40.4 (0.2–0.8)0.003
Surgery vs. postoperative radiotherapy0.6 (0.3–1.1)0.07

Treatment of Disease Recurrence

Salvage treatments with a wide spectrum of modalities including salvage surgery and radiotherapy were performed in 17 of the 37 patients who had local disease recurrence. Ultimately, stable disease was achieved in only 3 patients (for a period of 23–183 months starting from the date of disease recurrence and ending with the date of the last follow-up visit). Most patients with distant metastasis or uncontrolled disease recurrences underwent systemic treatment with dacarbazine and/or interferon-alpha-2b without a significant response.


In this retrospective study of 69 patients with primary HNMM treated with surgery with curative intent in a single institution, postoperative radiotherapy appears to increase the local control rate, independent of primary tumor stage. Many studies2, 11–18, 21 report that wide surgical excision with clear margins as the primary treatment offers the best chances of obtaining local control and survival. The difference between survival rates for patients with cutaneous and mucosal melanoma is related to surgical margins. Surgery is often limited by anatomic and functional considerations. If surgery is not performed with wide margins, especially for patients with sinonasal tumors, postoperative radiotherapy could be useful.

As local control of HNMM was minimal,2, 11–16, 22–28 many patients were treated with postoperative radiotherapy even though melanoma is believed to be radioresistant.29–32 Complete or partial responses have been reported with exclusive radiotherapy.2, 13, 17, 19, 33–39 Gaze et al.33 obtained 8 complete clinical responses among 13 patients treated with external radiotherapy alone with a conventional fractionation schedule. At our institute and elsewhere, radiotherapy is commonly performed for patients with inoperable tumors or disease recurrences, which may partly explain the poor results.11, 15, 16, 22, 23, 27, 36, 40, 41

Radiobiologic and clinical studies32, 42 on cutaneous melanoma suggested that fractions of > 4 Gy delivered over a short period of time yielded better responses. Gilligan and Slevin34 reported a 3-year actuarial local control rate of 49% using radiotherapy alone that delivers doses of 50–55 Gy in 15–16 fractions in 28 patients with sinonasal melanoma. In another series,37 local control was achieved in 44% of 25 patients, with hypofractionated radiotherapy delivering 24 Gy in 3 fractions over 21 days. In our study, 24 patients received radiotherapy alone, 4 with a hypofractionated schedule, but none of them achieved long-term local control.

Postoperative radiotherapy was recommended for patients with high-risk locoregional cutaneous melanoma.31, 43–46 Lee et al.15 reported that “local control is the sine qua non condition for cure” (p. 124); and rightly so, because disease is usually localized at presentation and local disease recurrence often precedes distant metastasis. Therefore, adjuvant radiotherapy was adopted for patients with HNMM,15, 19 even in the absence of a demonstrated benefit for survival.8 Our retrospective study suggests that postoperative radiotherapy improves local control. This was particularly significant for small tumors for which the occurrence of metastases was related to local control and its corollary, survival. As reported in other studies,13, 15 local control seems to correlate with the development of secondary distant metastasis. Conversely, distant metastases occurred more frequently and earlier in patients with locally advanced tumors, even when local control was achieved. This could explain why we did not find any impact of postoperative radiotherapy on overall survival.

The current study shows that 47of 69 (68%) patients developed distant metastasis soon after the end of the treatment even though local control was achieved in 30 of these patients. Because wide surgical margins are often mutilating, this high rate of early distant metastasis underscores the need to detect lung, liver, and bone metastases before surgery is performed. Goerres et al.47 showed that positron emission tomography and 18F-fluorodeoxyglucose (FDG) may be suitable for the staging and/or restaging of these patients.

Given the severity of HNMM, adjuvant medical treatment is highly desirable. However, until now, no such treatment has proven to be of benefit in patients with skin melanoma. The current study suggests that postoperative radiotherapy could be of interest for HNMM, but these results will need to be confirmed in a large prospective randomized multicentric trial, given the retrospective nature of our investigation.


The authors thank Lorna Saint Ange for her assistance with article preparation and Dr. Gary Clayman (The University of Texas M. D. Anderson Cancer Center, Houston, TX) for reviewing the current report.