Biology of Desmoplastic Melanoma: A Case-Control Comparison With Other Melanomas
http://www.100md.com
《临床肿瘤学》
the Departments of Surgery, Biostatistics, Dermatology, and Pathology, Massachusetts General Hospital, Boston, MA
the Faculty of Medicine, University Medical Center, Utrecht, the Netherlands
ABSTRACT
PURPOSE: Previous studies have established that patients with desmoplastic melanoma (DM) have thicker primary tumors. Consequently, comparisons with other forms of melanoma have been strongly biased by differences in Breslow stage. This is the first case-matched control study comparing DM with other forms of melanoma.
PATIENTS AND METHODS: From a database of 3,202 melanoma patients treated at one institution, 89 patients with DM and 178 case-matched control patients (2:1) were identified by matching for tumor thickness, age, sex, and year of diagnosis. Clinical, pathologic, and outcome information was obtained from chart review.
RESULTS: Controls were matched successfully to patients for tumor thickness, age, sex, and year of diagnosis. Presentation with American Joint Committee on Cancer stage III or IV disease is less common in patients with DM compared to case-matched control patients (5% v 21%; P < .001). Re-excisions to obtain clear surgical margins are required more often in patients with DM compared to case-matched control patients (21% v 6%; P < .001). Risk of positive sentinel nodes is lower in patients with DM compared to case-matched control patients (8% v 34%; P = .013). Despite these differences, survival rates of patients with DM are the same as case-matched control patients.
CONCLUSION: Use of case-matched control patients matched for tumor thickness avoids biases introduced by the advanced Breslow stage of DMs. DMs are more locally aggressive than thickness-matched controls, and positive sentinel nodes are limited to patients with thick primary tumors. Importantly, patients with DM have survival rates similar to patients with other melanomas of similar thickness.
INTRODUCTION
Desmoplastic melanoma (DM) is a histological subtype of melanoma that has been reported to account for approximately 1% of all cases of melanoma1 and accounts for 2.8% of all cases in our database of patients with cutaneous melanoma seen at the Massachusetts General Hospital (MGH). DM was described first by Conley et al2 in 1971 as being characterized by a dermal spindle cell population of malignant melanocytes in a background of abundant collagen. DM may occur as a separate histological subtype or as a feature of a pre-existing pigmented lesion, most commonly lentigo maligna melanoma.
The largest series of patients with DM published thus far came from the Sydney Melanoma Unit.1 Published in 1998, this study of 280 patients found that DM was more common in males, occurred at a relatively late age, and was usually in the head and neck area. The median tumor thickness at presentation was 2.5 mm, which was much higher than the median thickness of 1.0 mm in the general Sydney Melanoma Unit population of patients with melanoma.
In general, series published by numerous other institutions support these findings.3-9 It has also been reported that the occurrence of lymph node involvement in patients with DM, both at presentation and as regional lymph node recurrence, is lower than that generally seen in the melanoma population as a whole.1,4,10,11 For this reason, some authors have suggested that sentinel lymph node biopsy may not be indicated for this group of patients.11 Although patients with DM usually present with thicker tumors, survival is thought to be better than for other forms of melanoma with the same thickness.6,12
However, a case-matched control study has never been performed. Thus, comparison with other forms of melanoma with respect to lymph node involvement and outcome has been influenced heavily by a significant difference in tumor thickness. In an effort to draw conclusions solely on the basis of the difference in histology, we compared the population of patients with DM seen at our institution with a group of controls matched for thickness, age, and sex. Because views on melanoma management have changed over the study period,13,14 we also matched for year of diagnosis to compare patients treated with similar standards of care.
PATIENTS AND METHODS
We identified 89 patients with DM from our database that contains information on 3,202 patients who presented to the MGH with melanoma between 1970 and 2002. Patients were classified as having DM if the pathology report, signed by a staff dermatopathologist, indicated there was at least a partial desmoplastic component. Because slides were often not available in our archives, the lesions were not rereviewed systematically. However, primary lesions of patients with DM with positive lymph nodes were rereviewed by one of our staff dermatopathologists (T.J.F.).
Subsequently, the two closest matches in the database were identified for each DM patient with respect to sex, tumor thickness, year of diagnosis, and age. If more than two matches for a DM patient were found in the initial query, two controls were selected by using computerized randomization. If the initial query did not yield two or more matches, the age range was broadened to a maximum of 5 years' difference. If this did not yield enough matches, the range for year of diagnosis was broadened to a maximum of 2 years' difference. For tumor thickness, although an exact match was not always possible, all matches were within the same American Joint Committee on Cancer (AJCC) staging thickness subgroup. All cases were sex-matched. Thus, a study population was formed and consisted of 89 patients with DM and 178 controls.
Information on patient characteristics, pathology, recurrences, and cause of death was subsequently collected from medical records, the social security death index, and the Massachusetts Department of Public Health Registry of Vital Records and Statistics. In case of conflicting sources, medical-record review was determined to be the most accurate. Patients were staged according to the criteria of the latest AJCC staging system for melanoma.15
The date of first recurrence or of last follow-up was used as the end point for determination of disease-free survival.
Statistical methods used included Student's t test for comparing age, the Mann-Whitney test for comparing tumor thickness, and Pearson's 2 test for comparing categoric variables such as sex, stage, and mitotic rate between patients with DM and controls. Survival curves were produced on the basis of the Kaplan-Meier method, and patients with DM were compared with controls by using a Cox proportional-hazards model. If cause of death was unknown, patients were censored in cause-specific survival analysis. The effects of clinical and pathologic characteristics on cause-specific survival were assessed with a Cox multivariate stepwise-regression model. P < .05 was considered statistically significant. SAS statistical software (SAS Institute, Cary, NC) was used for statistical analyses.
Surgical management of primary melanoma in general changed during the study period, as did surgical management of the lymph nodes. The extent of surgical margins was generally reduced during the study period to the current goal of resection of at least a 1-cm margin of normal skin on the face and at least a 2-cm margin in other anatomic locations, restricted only by functional or cosmetic limitations. The anticipated need for a skin graft or flap for adequate closure generally did not reduce the planned surgical margin. Elective lymph node dissection was used to stage lymph nodes before 1994, after which time sentinel node mapping was used to stage lymph nodes.
This retrospective study was carried out under a waiver of patient consent granted by the institutional review board of MGH.
RESULTS
Finding Case-Matched Controls
To identify criteria for matching a control group, the DM group was first compared to the general database of 3,202 melanoma patients seen at MGH between 1970 and 2002. Sex and age were examined in addition to aspects of the primary lesion that, according to the current staging system, affect outcome, such as Breslow thickness and ulceration.
The median Breslow thickness was 2.60 mm in the DM group and 0.96 mm in the overall database (P < .001). Thus, tumor thickness was selected as a matching criterion. The male-to-female ratio was 1.7:1 in the DM group and 1.1:1 in the overall database (P = .067). Because the difference approached statistical significance and it has been shown that males with melanoma have a poorer prognosis than females, 16 sex was used to select case-matched controls. The mean age was 62.3 years in the DM group and 52.7 years in the overall database (P < .001). Therefore, age was also chosen as one of the matching criteria. There was no significant difference in ulceration between the DM group and the overall database. The size of the general database would only allow us to match for one additional criterion, and we therefore preferred to match for year of diagnosis to ensure that control patients had the same diagnostic and therapeutic opportunities as patients with DM. This was necessary because the views on melanoma management evolved over the study period.13,14
Subsequently, 178 controls were identified for 89 patients with DM. A 2:1 matching ratio was chosen to obtain greater statistical power. After the matching procedure, there were no significant differences between the two groups with respect to sex, age, or tumor thickness (Table 1).
General Patient Characteristics
General patient and histology characteristics for the DM and case-matched control groups are listed in Table 1.
The most common site of the primary melanoma in the DM group was head and neck, whereas the trunk was the most common site in the control group. There was no difference in the incidence of nonpigmented melanomas between the groups.
Although data on neurotropism were incomplete, this histological feature was observed much more commonly in patients with DM. Clark's level of invasion was significantly higher and mitotic rate was significantly lower in the DM group.
Patients with DM presented with an earlier stage of disease than did control patients, although it should be noted that patients were matched for tumor thickness, which thus was not factored into this stage comparison. Five percent of the patients with DM versus 21% of control patients presented with stage III or IV disease (P < .001).
Sixteen patients (17.9%) with DM had more than one primary melanoma, whereas only 12 controls (6.7%) had more than one primary melanoma (P = .004). This was an unexpected finding, and we therefore examined patients with DM with multiple primary melanomas more closely. Second primaries were only scored as such if they were invasive; thus, none of the second primaries were in situ melanomas. Thirteen of these 16 patients had another primary melanoma before diagnosis of DM. None had more than one primary DM; six patients had lentigo maligna melanoma as a second primary, and five had superficial spreading melanoma. Family history was positive in five of 16 patients with DM with multiple primary melanomas and negative in nine of 16 patients with DM.
Patients with DM also had significantly more nonmelanoma skin cancers, which additional analysis showed to be caused by a higher prevalence of basal cell carcinoma. No difference was observed in the prevalence of squamous cell carcinoma between the two groups. Two patients with DM had a leiomyosarcoma, one in the stomach before diagnosis of the melanoma and one in the spinal cord after the diagnosis of melanoma. There was no significant difference in prevalence of other cancers, dysplastic nevi, immune suppression, or a positive family history.
Treatment of Primary Lesion
Standard treatment of the primary lesion was wide local excision, usually after an initial biopsy. We found that patients with DM required significantly more re-excisions than control patients to obtain clear margins. More than 20% of the patients with DM had microscopically positive margins after wide local excision, which required at least one re-excision, compared to only 6.4% of the controls (Table 2).
Lymph Node Evaluation
There was no difference between the two groups with respect to the number of sentinel lymph node biopsies or elective lymph node dissections. Table 3 shows the number of sentinel lymph node biopsies and lymph node dissections without lymphatic mapping for both groups, as well as the number of positive results. Although the proportions of DM and control patients who underwent sentinel node biopsy were similar, patients with DM had a significantly lower rate of positive sentinel lymph nodes compared with control patients (8.0% v 33.8%, respectively; P = .013). After having a positive sentinel lymph node, one of two patients with DM and four of 20 control patients had at least one additional positive node at completion of lymph node dissection (P = NS).
No positive lymph nodes were identified in patients who underwent elective lymph node dissection, either in the DM group or the control group. Therapeutic lymph node dissection was performed more often in the control group, but the difference did not reach statistical significance (P = .054). The maximum number of positive nodes in an individual patient was two in the DM group and six in the control group. Overall, 5% of the patients with DM had positive lymph nodes at the time of presentation, compared with 20% of the control patients (P = .005).
After closer analyses of the four patients with DM who had positive nodes, we found that the thicknesses of their primary lesions were 2.9, 3.0, 3.8, and 7.0 mm, respectively. All four primary lesions had evidence of neurotropism, and three of the four were ulcerated. The primary lesions of two of these patients showed a mitotic rate between one and four mitoses per 10 high-power fields; the other two had more than four mitoses per 10 high-power fields.
Busam et al17 recently described a distinction between "combined DM" (in which a desmoplastic component is seen in a lesion that is primarily a nondesmoplastic "conventional" invasive melanoma) and pure DM (in which the lesion is completely desmoplastic). We re-examined the primary lesions of three of the four patients with positive lymph nodes according to their criteria. One of two patients who underwent therapeutic lymphadenectomy for positive nodes had a primary melanoma that could be classified as pure DM. Neither of the two lymphatic mapping patients with positive sentinel nodes had pure DM. The slides of the primary lesion were irretrievable in one patient who underwent therapeutic lymphadenectomy for positive nodes.
Recurrences
Of all patients with DM, 24.7% had at least one recurrence, in comparison with 29.8% of the controls, a difference that is not statistically significant (Table 4). There was no significant difference in the number of local, regional, or distant recurrences, either as first or as any recurrence, between patients with DM and controls. The lungs were the most common site of distant metastasis in both groups, and there was no significant difference in the occurrence of metastases to lungs, bone, liver, CNS, or gastrointestinal tract. The sites of distant metastases are listed in Table 5.
Survival
Compared with the entire melanoma database, overall survival was significantly worse in the DM group (P < .001). Having a case-matched control group, however, allowed for more accurate comparison of survival between DM and other forms of melanoma. No difference in overall, cause-specific, or disease-free survival was observed between DM and the control group (Figs 1, 2, and 3, respectively). Thirty-three patients with DM (16 related, two of unknown cause) and 56 controls (35 related, five of unknown cause) died during the study period. Mean follow-up time was 5.5 years in the DM group and 6.0 years in the control group (median, 4.3 and 4.8 years, respectively). Five-year overall survival was 72.6% in the DM group and 76.9% in the control group, and 5-year disease-free survival was 65.7% and 67.4%, respectively.
In univariate analysis, Breslow thickness, ulceration, mitotic rate, and presentation with AJCC stage III (as opposed to stage I or II) were of significant negative influence on cause-specific survival of patients with DM. Sex, age, site of primary melanoma, Clark's level, neurotropism, and the necessity of multiple re-excisions had no statistically significant effect on cause-specific survival. In multivariate analysis, tumor thickness and presentation with AJCC stage III remained as independent significant predictors of poor outcome (Table 6).
DISCUSSION
This is the first study on DM to include a case-matched control group for direct comparison with other forms of melanoma. A comparison of patients with DM with a control group matched for sex, age, tumor thickness, and year of diagnosis provides interesting insights into the clinical and biologic behavior of this rare form of melanoma.
In general, the presented data are consistent with several previously described characteristics of the initial presentation of DM. Our study supports previous observations that DM is locally invasive. The average thickness of the primary lesions of patients with DM was greater than that of the other patients in our database of 3,288 patients with melanoma, which is in accordance with results from other studies on DM.1,18 In addition, Clark's level of invasion was higher in the DM group than in the case-matched control group. Obtaining clear surgical margins in patients with DM was significantly more difficult than in control patients. Looking more closely at the characteristics of patients with DM who required multiple re-excisions, we noted that 12 of these 19 patients had their primary lesion in the head and neck area, where anatomic and cosmetic considerations often limit surgical options. Furthermore, 10 of these 19 primary tumors showed neurotropism, which has been shown to be associated with local aggressiveness of DM.1,19 Regional or distant metastases at time of presentation, however, were much less frequent in the DM group compared to case-matched controls.
To our knowledge, this is the first study on DM to look at other diseases occurring in these patients. It is noteworthy that patients with DM have significantly more other primary melanomas, as well as basal cell carcinomas, than do control patients. The prevalence of second primary melanomas in patients with DM is surprisingly high, and we could not find a clear explanation for this observation. None of these patients had multiple DMs, and they did not have a positive family history more often than other patients. A possible explanation could be increased sun exposure in these patients, but we could not verify this retrospectively.
We also found that two patients with DM had leiomyosarcoma. Thus, the prevalence of leiomyosarcoma in our group of patients with DM is substantially greater than would be expected for the general population. The incidence of soft-tissue sarcomas is approximately two to four cases per 100,000, with leiomyosarcoma only constituting approximately 10% to 25% of them.20,21 Although this is an interesting observation, our study was not initially designed to address the relationship between leiomyosarcoma and DM. The fact remains that we observed only two cases of leiomyosarcoma.
Our results also confirm the lower incidence of positive sentinel node biopsies among patients with DM reported in literature. The 8% rate of positive sentinel node biopsies in our series is in accordance with a recently published study by Su et al,10 who reported a rate of 12%. The mean tumor thickness in their study was higher than in our series, which may explain their slightly higher fraction of patients with positive sentinel nodes.
Gyorki et al11 found no positive sentinel lymph nodes in their study of 24 patients with DM undergoing sentinel lymph node biopsy and concluded that sentinel lymph node biopsy might not be indicated in this subgroup of patients with melanoma. In our series, 8% of the patients who underwent mapping had positive nodes, and presentation with AJCC stage III was an independent predictor of cause-specific survival in multivariate analysis. It is our impression, therefore, that although the likelihood of a positive lymph node at time of diagnosis is lower in patients with DM, the prognostic value of such a finding is of no less significance. Considering the nature of the primary lesions of the four patients with DM in our series that had positive nodes, we suspect that sentinel node biopsy might be particularly useful in subgroups with advanced primary lesions (ie, tumors that are thicker than 2.0 mm, ulcerated, or have neurotropism).
We should emphasize that desmoplastic primary lesions are located significantly more often on the head and neck, because sentinel node biopsy has been reported to be more difficult in this area than in other parts of the body. A recent large study on this topic reported a lower number of positive sentinel nodes and higher false-negative rates for sentinel lymph nodes biopsies in the head and neck area than in other body parts.22
In contrast with suggestions from previous studies,1,23 we observed no significant difference in recurrence patterns between patients with DM and case-matched controls. Previous studies reported high local recurrence rates in patients with DM, although the most recent studies show somewhat lower local recurrence rates than earlier series. In this study we found a local recurrence rate of 7.9% as first recurrence and 11.2% as any recurrence. Although these rates are higher than for melanoma in general, they are not significantly higher than the local recurrence rates in the thickness-matched control group. However, initial local control, in terms of clear surgical margins, was significantly harder to obtain in the DM group; patients with DM required significantly more re-excisions.
The difference in regional recurrence between the two groups also did not reach statistical significance, which is contrary to results from prior studies.1,11 However, although not statistically significant, the local recurrence rate is twice as high in the DM group compared with the control group, and the regional recurrence rate was approximately twice as low. Despite this large absolute difference and the fairly large study population of 267 patients, the frequency of these events was too low for this difference to be statistically significant. Because this is the first study to include a control group for statistical comparison, it was not possible to combine these data with those from previous publications to perform a meta-analysis.
In a review of their experience with 92 patients with DM, Busam et al17 showed that a distinction can be made between combined DM (in which a desmoplastic component is seen in a lesion that is primarily a nondesmoplastic "conventional" invasive melanoma) and pure DM (in which the lesion is completely desmoplastic). In their series, in patients with pure DM, nodal status is negative more often and survival is better in pure DM compared with combined DM. Previously, Anstey et al24 did not find these differences after making a similar distinction, but their study population consisted of only 25 patients. In our series we did not distinguish between combined and pure DM. If combined versus pure DM histology proves to be an important prognostic variable, this factor may account for some differences in results between the studies. We did review the primary lesions of three of four patients with DM with positive nodes and found that one of these patients had pure DM. This patient had a primary lesion with a thickness of 7 mm that was diagnosed before the advent of sentinel lymph node biopsies, and multiple positive nodes were found after therapeutic lymph node dissection. Therefore, although we have no patients with pure DM with positive sentinel node biopsies, we believe that this procedure may be considered for patients with particularly thick pure-DM lesions.
There was no difference in the number of patients with distant metastases between the groups. Numerous studies of general melanoma populations reported the lungs, liver, and bone as the most common sites of metastasis.25,26 The results of our study are in accordance with this result and show that DM does not differ significantly from other forms of melanoma with respect to preferential sites of distant metastasis.
Previous studies provide conflicting conclusions concerning the prognosis of patients with DM versus non-DM, although the former usually present with locally more advanced disease.5,12,23 A direct comparison with a case-matched control group had not been made thus far. Overall and cause-specific survival was significantly worse in our DM study population than for all patients in our melanoma database. However, after matching for sex, age, and tumor thickness, there was no survival difference between patients with DM and control patients. From this important observation, we conclude that prognosis of DM is similar to that of other forms of melanoma of the same thickness.
We found tumor thickness, ulceration, mitotic rate, and presentation with AJCC stage III to be significant predictors of survival of patients with DM in univariate analysis, which is similar to other forms of melanoma.16 We have an insufficient number of patients with DM with neurotropism to render a meaningful analysis of the impact of neurotropism on outcome.
The histological feature of desmoplasia in melanoma is associated with a characteristic phenotype that is different from other forms of melanoma in its initial presentation and in the occurrence of lymphatic spread. Local aggressiveness with thicker primary tumors and reduced frequency of lymphatic spread is observed more commonly with DM. Comparison to a contemporaneous case-matched control group reveals that overall, cause-specific, and disease-free survival are identical to thickness-matched controls with other forms of melanoma.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
D.P.L. was supported in part by a scholarship from the Dutch Cancer Society.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Quinn MJ, Crotty KA, Thompson JF, et al: Desmoplastic and desmoplastic neurotropic melanoma. Cancer 83:1128-1135, 1998
Conley J, Lattes R, Orr W: Desmoplastic malignant melanoma (a rare variant of spindle cell melanoma). Cancer 28:914-936, 1971
Payne WG, Kearney R, Wells K, et al: Desmoplastic melanoma. Am Surg 67:1004-1006, 2001
Jaroszewski DE, Pockaj BA, DiCaudo DJ, et al: The clinical behavior of desmoplastic melanoma. Am J Surg 182:590-595, 2001
Carlson JA, Dickersin GR, Sober AJ, et al: Desmoplastic neurotropic melanoma: A clinicopathologic analysis of 28 cases. Cancer 75:478-494, 1995
Skelton HG, Smith KJ, Laskin WB, et al: Desmoplastic malignant melanoma. J Am Acad Dermatol 32:717-725, 1995
Smithers BM, McLeod GR, Little JH: Desmoplastic, neural transforming and neurotropic melanoma: A review of 45 cases. Aust N Z J Surg 60:967-972, 1990
Jain S, Allen PW: Desmoplastic malignant melanoma and its variants. A study of 45 cases. Am J Surg Pathol 13:358-373, 1989
Egbert B, Kempson R, Sagebiel R: Desmoplastic malignant melanoma: A clinicohistopathologic study of 25 cases. Cancer 62:2033-2041, 1988
Su LD, Fullen DR, Lowe L, et al: Desmoplastic and neurotropic melanoma: Analysis of 33 patients with lymphatic mapping and sentinel lymph node biopsy. Cancer 100:598-604, 2004
Gyorki DE, Busam K, Panageas K, et al: Sentinel lymph node biopsy for patients with cutaneous desmoplastic melanoma. Ann Surg Oncol 10:403-407, 2003
Tsao H, Sober AJ, Barnhill RL: Desmoplastic neurotropic melanoma. Semin Cutan Med Surg 16:131-136, 1997
Balch CM, Urist MM, Karakousis CP, et al: Efficacy of 2-cm surgical margins for intermediate-thickness melanomas (1 to 4 mm): Results of a multi-institutional randomized surgical trial. Ann Surg 218:262-267, 1993
Reintgen D, Balch CM, Kirkwood J, et al: Recent advances in the care of the patient with malignant melanoma. Ann Surg 225:1-14, 1997
Kim CJ, Reintgen DS, Balch CM, et al: The new melanoma staging system. Cancer Control 9:9-15, 2002
Balch CM, Soong SJ, Gershenwald JE, et al: Prognostic factors analysis of 17,600 melanoma patients: Validation of the American Joint Committee on Cancer melanoma staging system. J Clin Oncol 19:3622-3634, 2001
Busam KJ, Mujumdar U, Hummer AJ, et al: Cutaneous desmoplastic melanoma: Reappraisal of morphologic heterogeneity and prognostic factors. Am J Surg Pathol 28:1518-1525, 2004
Posthner KE, Mosca PJ, Selim MA, et al: Histopathologic characteristics, recurrence patterns and survival of 129 patients with desmoplastic melanoma. Ann Surg Oncol 11:S78, 2004 (abstr 93)
Baer SC, Schultz D, Synnestvedt M, et al: Desmoplasia and neurotropism: Prognostic variables in patients with stage I melanoma. Cancer 76:2242-2247, 1995
Daugaard S: Current soft-tissue sarcoma classifications. Eur J Cancer 40:543-548, 2004
Ross JA, Severson RK, Davis S, et al: Trends in the incidence of soft tissue sarcomas in the United States from 1973 through 1987. Cancer 72:486-490, 1993
Chao C, Wong SL, Edwards MJ, et al: Sentinel lymph node biopsy for head and neck melanomas. Ann Surg Oncol 10:21-26, 2003
Smithers BM, McLeod GR, Little JH: Desmoplastic melanoma: Patterns of recurrence. World J Surg 16:186-190, 1992
Anstey A, McKee P, Jones EW: Desmoplastic malignant melanoma: A clinicopathological study of 25 cases. Br J Dermatol 129:359-371, 1994
Balch CM, Soong SJ, Murad TM, et al: A multifactorial analysis of melanoma. IV. Prognostic factors in 200 melanoma patients with distant metastases (stage III). J Clin Oncol 1:126-134, 1983
Cohn-Cedermark G, Mansson-Brahme E, Rutqvist LE, et al: Metastatic patterns, clinical outcome, and malignant phenotype in malignant cutaneous melanoma. Acta Oncol 38:549-557, 1999(Daan P. Livestro, Alona M)
the Faculty of Medicine, University Medical Center, Utrecht, the Netherlands
ABSTRACT
PURPOSE: Previous studies have established that patients with desmoplastic melanoma (DM) have thicker primary tumors. Consequently, comparisons with other forms of melanoma have been strongly biased by differences in Breslow stage. This is the first case-matched control study comparing DM with other forms of melanoma.
PATIENTS AND METHODS: From a database of 3,202 melanoma patients treated at one institution, 89 patients with DM and 178 case-matched control patients (2:1) were identified by matching for tumor thickness, age, sex, and year of diagnosis. Clinical, pathologic, and outcome information was obtained from chart review.
RESULTS: Controls were matched successfully to patients for tumor thickness, age, sex, and year of diagnosis. Presentation with American Joint Committee on Cancer stage III or IV disease is less common in patients with DM compared to case-matched control patients (5% v 21%; P < .001). Re-excisions to obtain clear surgical margins are required more often in patients with DM compared to case-matched control patients (21% v 6%; P < .001). Risk of positive sentinel nodes is lower in patients with DM compared to case-matched control patients (8% v 34%; P = .013). Despite these differences, survival rates of patients with DM are the same as case-matched control patients.
CONCLUSION: Use of case-matched control patients matched for tumor thickness avoids biases introduced by the advanced Breslow stage of DMs. DMs are more locally aggressive than thickness-matched controls, and positive sentinel nodes are limited to patients with thick primary tumors. Importantly, patients with DM have survival rates similar to patients with other melanomas of similar thickness.
INTRODUCTION
Desmoplastic melanoma (DM) is a histological subtype of melanoma that has been reported to account for approximately 1% of all cases of melanoma1 and accounts for 2.8% of all cases in our database of patients with cutaneous melanoma seen at the Massachusetts General Hospital (MGH). DM was described first by Conley et al2 in 1971 as being characterized by a dermal spindle cell population of malignant melanocytes in a background of abundant collagen. DM may occur as a separate histological subtype or as a feature of a pre-existing pigmented lesion, most commonly lentigo maligna melanoma.
The largest series of patients with DM published thus far came from the Sydney Melanoma Unit.1 Published in 1998, this study of 280 patients found that DM was more common in males, occurred at a relatively late age, and was usually in the head and neck area. The median tumor thickness at presentation was 2.5 mm, which was much higher than the median thickness of 1.0 mm in the general Sydney Melanoma Unit population of patients with melanoma.
In general, series published by numerous other institutions support these findings.3-9 It has also been reported that the occurrence of lymph node involvement in patients with DM, both at presentation and as regional lymph node recurrence, is lower than that generally seen in the melanoma population as a whole.1,4,10,11 For this reason, some authors have suggested that sentinel lymph node biopsy may not be indicated for this group of patients.11 Although patients with DM usually present with thicker tumors, survival is thought to be better than for other forms of melanoma with the same thickness.6,12
However, a case-matched control study has never been performed. Thus, comparison with other forms of melanoma with respect to lymph node involvement and outcome has been influenced heavily by a significant difference in tumor thickness. In an effort to draw conclusions solely on the basis of the difference in histology, we compared the population of patients with DM seen at our institution with a group of controls matched for thickness, age, and sex. Because views on melanoma management have changed over the study period,13,14 we also matched for year of diagnosis to compare patients treated with similar standards of care.
PATIENTS AND METHODS
We identified 89 patients with DM from our database that contains information on 3,202 patients who presented to the MGH with melanoma between 1970 and 2002. Patients were classified as having DM if the pathology report, signed by a staff dermatopathologist, indicated there was at least a partial desmoplastic component. Because slides were often not available in our archives, the lesions were not rereviewed systematically. However, primary lesions of patients with DM with positive lymph nodes were rereviewed by one of our staff dermatopathologists (T.J.F.).
Subsequently, the two closest matches in the database were identified for each DM patient with respect to sex, tumor thickness, year of diagnosis, and age. If more than two matches for a DM patient were found in the initial query, two controls were selected by using computerized randomization. If the initial query did not yield two or more matches, the age range was broadened to a maximum of 5 years' difference. If this did not yield enough matches, the range for year of diagnosis was broadened to a maximum of 2 years' difference. For tumor thickness, although an exact match was not always possible, all matches were within the same American Joint Committee on Cancer (AJCC) staging thickness subgroup. All cases were sex-matched. Thus, a study population was formed and consisted of 89 patients with DM and 178 controls.
Information on patient characteristics, pathology, recurrences, and cause of death was subsequently collected from medical records, the social security death index, and the Massachusetts Department of Public Health Registry of Vital Records and Statistics. In case of conflicting sources, medical-record review was determined to be the most accurate. Patients were staged according to the criteria of the latest AJCC staging system for melanoma.15
The date of first recurrence or of last follow-up was used as the end point for determination of disease-free survival.
Statistical methods used included Student's t test for comparing age, the Mann-Whitney test for comparing tumor thickness, and Pearson's 2 test for comparing categoric variables such as sex, stage, and mitotic rate between patients with DM and controls. Survival curves were produced on the basis of the Kaplan-Meier method, and patients with DM were compared with controls by using a Cox proportional-hazards model. If cause of death was unknown, patients were censored in cause-specific survival analysis. The effects of clinical and pathologic characteristics on cause-specific survival were assessed with a Cox multivariate stepwise-regression model. P < .05 was considered statistically significant. SAS statistical software (SAS Institute, Cary, NC) was used for statistical analyses.
Surgical management of primary melanoma in general changed during the study period, as did surgical management of the lymph nodes. The extent of surgical margins was generally reduced during the study period to the current goal of resection of at least a 1-cm margin of normal skin on the face and at least a 2-cm margin in other anatomic locations, restricted only by functional or cosmetic limitations. The anticipated need for a skin graft or flap for adequate closure generally did not reduce the planned surgical margin. Elective lymph node dissection was used to stage lymph nodes before 1994, after which time sentinel node mapping was used to stage lymph nodes.
This retrospective study was carried out under a waiver of patient consent granted by the institutional review board of MGH.
RESULTS
Finding Case-Matched Controls
To identify criteria for matching a control group, the DM group was first compared to the general database of 3,202 melanoma patients seen at MGH between 1970 and 2002. Sex and age were examined in addition to aspects of the primary lesion that, according to the current staging system, affect outcome, such as Breslow thickness and ulceration.
The median Breslow thickness was 2.60 mm in the DM group and 0.96 mm in the overall database (P < .001). Thus, tumor thickness was selected as a matching criterion. The male-to-female ratio was 1.7:1 in the DM group and 1.1:1 in the overall database (P = .067). Because the difference approached statistical significance and it has been shown that males with melanoma have a poorer prognosis than females, 16 sex was used to select case-matched controls. The mean age was 62.3 years in the DM group and 52.7 years in the overall database (P < .001). Therefore, age was also chosen as one of the matching criteria. There was no significant difference in ulceration between the DM group and the overall database. The size of the general database would only allow us to match for one additional criterion, and we therefore preferred to match for year of diagnosis to ensure that control patients had the same diagnostic and therapeutic opportunities as patients with DM. This was necessary because the views on melanoma management evolved over the study period.13,14
Subsequently, 178 controls were identified for 89 patients with DM. A 2:1 matching ratio was chosen to obtain greater statistical power. After the matching procedure, there were no significant differences between the two groups with respect to sex, age, or tumor thickness (Table 1).
General Patient Characteristics
General patient and histology characteristics for the DM and case-matched control groups are listed in Table 1.
The most common site of the primary melanoma in the DM group was head and neck, whereas the trunk was the most common site in the control group. There was no difference in the incidence of nonpigmented melanomas between the groups.
Although data on neurotropism were incomplete, this histological feature was observed much more commonly in patients with DM. Clark's level of invasion was significantly higher and mitotic rate was significantly lower in the DM group.
Patients with DM presented with an earlier stage of disease than did control patients, although it should be noted that patients were matched for tumor thickness, which thus was not factored into this stage comparison. Five percent of the patients with DM versus 21% of control patients presented with stage III or IV disease (P < .001).
Sixteen patients (17.9%) with DM had more than one primary melanoma, whereas only 12 controls (6.7%) had more than one primary melanoma (P = .004). This was an unexpected finding, and we therefore examined patients with DM with multiple primary melanomas more closely. Second primaries were only scored as such if they were invasive; thus, none of the second primaries were in situ melanomas. Thirteen of these 16 patients had another primary melanoma before diagnosis of DM. None had more than one primary DM; six patients had lentigo maligna melanoma as a second primary, and five had superficial spreading melanoma. Family history was positive in five of 16 patients with DM with multiple primary melanomas and negative in nine of 16 patients with DM.
Patients with DM also had significantly more nonmelanoma skin cancers, which additional analysis showed to be caused by a higher prevalence of basal cell carcinoma. No difference was observed in the prevalence of squamous cell carcinoma between the two groups. Two patients with DM had a leiomyosarcoma, one in the stomach before diagnosis of the melanoma and one in the spinal cord after the diagnosis of melanoma. There was no significant difference in prevalence of other cancers, dysplastic nevi, immune suppression, or a positive family history.
Treatment of Primary Lesion
Standard treatment of the primary lesion was wide local excision, usually after an initial biopsy. We found that patients with DM required significantly more re-excisions than control patients to obtain clear margins. More than 20% of the patients with DM had microscopically positive margins after wide local excision, which required at least one re-excision, compared to only 6.4% of the controls (Table 2).
Lymph Node Evaluation
There was no difference between the two groups with respect to the number of sentinel lymph node biopsies or elective lymph node dissections. Table 3 shows the number of sentinel lymph node biopsies and lymph node dissections without lymphatic mapping for both groups, as well as the number of positive results. Although the proportions of DM and control patients who underwent sentinel node biopsy were similar, patients with DM had a significantly lower rate of positive sentinel lymph nodes compared with control patients (8.0% v 33.8%, respectively; P = .013). After having a positive sentinel lymph node, one of two patients with DM and four of 20 control patients had at least one additional positive node at completion of lymph node dissection (P = NS).
No positive lymph nodes were identified in patients who underwent elective lymph node dissection, either in the DM group or the control group. Therapeutic lymph node dissection was performed more often in the control group, but the difference did not reach statistical significance (P = .054). The maximum number of positive nodes in an individual patient was two in the DM group and six in the control group. Overall, 5% of the patients with DM had positive lymph nodes at the time of presentation, compared with 20% of the control patients (P = .005).
After closer analyses of the four patients with DM who had positive nodes, we found that the thicknesses of their primary lesions were 2.9, 3.0, 3.8, and 7.0 mm, respectively. All four primary lesions had evidence of neurotropism, and three of the four were ulcerated. The primary lesions of two of these patients showed a mitotic rate between one and four mitoses per 10 high-power fields; the other two had more than four mitoses per 10 high-power fields.
Busam et al17 recently described a distinction between "combined DM" (in which a desmoplastic component is seen in a lesion that is primarily a nondesmoplastic "conventional" invasive melanoma) and pure DM (in which the lesion is completely desmoplastic). We re-examined the primary lesions of three of the four patients with positive lymph nodes according to their criteria. One of two patients who underwent therapeutic lymphadenectomy for positive nodes had a primary melanoma that could be classified as pure DM. Neither of the two lymphatic mapping patients with positive sentinel nodes had pure DM. The slides of the primary lesion were irretrievable in one patient who underwent therapeutic lymphadenectomy for positive nodes.
Recurrences
Of all patients with DM, 24.7% had at least one recurrence, in comparison with 29.8% of the controls, a difference that is not statistically significant (Table 4). There was no significant difference in the number of local, regional, or distant recurrences, either as first or as any recurrence, between patients with DM and controls. The lungs were the most common site of distant metastasis in both groups, and there was no significant difference in the occurrence of metastases to lungs, bone, liver, CNS, or gastrointestinal tract. The sites of distant metastases are listed in Table 5.
Survival
Compared with the entire melanoma database, overall survival was significantly worse in the DM group (P < .001). Having a case-matched control group, however, allowed for more accurate comparison of survival between DM and other forms of melanoma. No difference in overall, cause-specific, or disease-free survival was observed between DM and the control group (Figs 1, 2, and 3, respectively). Thirty-three patients with DM (16 related, two of unknown cause) and 56 controls (35 related, five of unknown cause) died during the study period. Mean follow-up time was 5.5 years in the DM group and 6.0 years in the control group (median, 4.3 and 4.8 years, respectively). Five-year overall survival was 72.6% in the DM group and 76.9% in the control group, and 5-year disease-free survival was 65.7% and 67.4%, respectively.
In univariate analysis, Breslow thickness, ulceration, mitotic rate, and presentation with AJCC stage III (as opposed to stage I or II) were of significant negative influence on cause-specific survival of patients with DM. Sex, age, site of primary melanoma, Clark's level, neurotropism, and the necessity of multiple re-excisions had no statistically significant effect on cause-specific survival. In multivariate analysis, tumor thickness and presentation with AJCC stage III remained as independent significant predictors of poor outcome (Table 6).
DISCUSSION
This is the first study on DM to include a case-matched control group for direct comparison with other forms of melanoma. A comparison of patients with DM with a control group matched for sex, age, tumor thickness, and year of diagnosis provides interesting insights into the clinical and biologic behavior of this rare form of melanoma.
In general, the presented data are consistent with several previously described characteristics of the initial presentation of DM. Our study supports previous observations that DM is locally invasive. The average thickness of the primary lesions of patients with DM was greater than that of the other patients in our database of 3,288 patients with melanoma, which is in accordance with results from other studies on DM.1,18 In addition, Clark's level of invasion was higher in the DM group than in the case-matched control group. Obtaining clear surgical margins in patients with DM was significantly more difficult than in control patients. Looking more closely at the characteristics of patients with DM who required multiple re-excisions, we noted that 12 of these 19 patients had their primary lesion in the head and neck area, where anatomic and cosmetic considerations often limit surgical options. Furthermore, 10 of these 19 primary tumors showed neurotropism, which has been shown to be associated with local aggressiveness of DM.1,19 Regional or distant metastases at time of presentation, however, were much less frequent in the DM group compared to case-matched controls.
To our knowledge, this is the first study on DM to look at other diseases occurring in these patients. It is noteworthy that patients with DM have significantly more other primary melanomas, as well as basal cell carcinomas, than do control patients. The prevalence of second primary melanomas in patients with DM is surprisingly high, and we could not find a clear explanation for this observation. None of these patients had multiple DMs, and they did not have a positive family history more often than other patients. A possible explanation could be increased sun exposure in these patients, but we could not verify this retrospectively.
We also found that two patients with DM had leiomyosarcoma. Thus, the prevalence of leiomyosarcoma in our group of patients with DM is substantially greater than would be expected for the general population. The incidence of soft-tissue sarcomas is approximately two to four cases per 100,000, with leiomyosarcoma only constituting approximately 10% to 25% of them.20,21 Although this is an interesting observation, our study was not initially designed to address the relationship between leiomyosarcoma and DM. The fact remains that we observed only two cases of leiomyosarcoma.
Our results also confirm the lower incidence of positive sentinel node biopsies among patients with DM reported in literature. The 8% rate of positive sentinel node biopsies in our series is in accordance with a recently published study by Su et al,10 who reported a rate of 12%. The mean tumor thickness in their study was higher than in our series, which may explain their slightly higher fraction of patients with positive sentinel nodes.
Gyorki et al11 found no positive sentinel lymph nodes in their study of 24 patients with DM undergoing sentinel lymph node biopsy and concluded that sentinel lymph node biopsy might not be indicated in this subgroup of patients with melanoma. In our series, 8% of the patients who underwent mapping had positive nodes, and presentation with AJCC stage III was an independent predictor of cause-specific survival in multivariate analysis. It is our impression, therefore, that although the likelihood of a positive lymph node at time of diagnosis is lower in patients with DM, the prognostic value of such a finding is of no less significance. Considering the nature of the primary lesions of the four patients with DM in our series that had positive nodes, we suspect that sentinel node biopsy might be particularly useful in subgroups with advanced primary lesions (ie, tumors that are thicker than 2.0 mm, ulcerated, or have neurotropism).
We should emphasize that desmoplastic primary lesions are located significantly more often on the head and neck, because sentinel node biopsy has been reported to be more difficult in this area than in other parts of the body. A recent large study on this topic reported a lower number of positive sentinel nodes and higher false-negative rates for sentinel lymph nodes biopsies in the head and neck area than in other body parts.22
In contrast with suggestions from previous studies,1,23 we observed no significant difference in recurrence patterns between patients with DM and case-matched controls. Previous studies reported high local recurrence rates in patients with DM, although the most recent studies show somewhat lower local recurrence rates than earlier series. In this study we found a local recurrence rate of 7.9% as first recurrence and 11.2% as any recurrence. Although these rates are higher than for melanoma in general, they are not significantly higher than the local recurrence rates in the thickness-matched control group. However, initial local control, in terms of clear surgical margins, was significantly harder to obtain in the DM group; patients with DM required significantly more re-excisions.
The difference in regional recurrence between the two groups also did not reach statistical significance, which is contrary to results from prior studies.1,11 However, although not statistically significant, the local recurrence rate is twice as high in the DM group compared with the control group, and the regional recurrence rate was approximately twice as low. Despite this large absolute difference and the fairly large study population of 267 patients, the frequency of these events was too low for this difference to be statistically significant. Because this is the first study to include a control group for statistical comparison, it was not possible to combine these data with those from previous publications to perform a meta-analysis.
In a review of their experience with 92 patients with DM, Busam et al17 showed that a distinction can be made between combined DM (in which a desmoplastic component is seen in a lesion that is primarily a nondesmoplastic "conventional" invasive melanoma) and pure DM (in which the lesion is completely desmoplastic). In their series, in patients with pure DM, nodal status is negative more often and survival is better in pure DM compared with combined DM. Previously, Anstey et al24 did not find these differences after making a similar distinction, but their study population consisted of only 25 patients. In our series we did not distinguish between combined and pure DM. If combined versus pure DM histology proves to be an important prognostic variable, this factor may account for some differences in results between the studies. We did review the primary lesions of three of four patients with DM with positive nodes and found that one of these patients had pure DM. This patient had a primary lesion with a thickness of 7 mm that was diagnosed before the advent of sentinel lymph node biopsies, and multiple positive nodes were found after therapeutic lymph node dissection. Therefore, although we have no patients with pure DM with positive sentinel node biopsies, we believe that this procedure may be considered for patients with particularly thick pure-DM lesions.
There was no difference in the number of patients with distant metastases between the groups. Numerous studies of general melanoma populations reported the lungs, liver, and bone as the most common sites of metastasis.25,26 The results of our study are in accordance with this result and show that DM does not differ significantly from other forms of melanoma with respect to preferential sites of distant metastasis.
Previous studies provide conflicting conclusions concerning the prognosis of patients with DM versus non-DM, although the former usually present with locally more advanced disease.5,12,23 A direct comparison with a case-matched control group had not been made thus far. Overall and cause-specific survival was significantly worse in our DM study population than for all patients in our melanoma database. However, after matching for sex, age, and tumor thickness, there was no survival difference between patients with DM and control patients. From this important observation, we conclude that prognosis of DM is similar to that of other forms of melanoma of the same thickness.
We found tumor thickness, ulceration, mitotic rate, and presentation with AJCC stage III to be significant predictors of survival of patients with DM in univariate analysis, which is similar to other forms of melanoma.16 We have an insufficient number of patients with DM with neurotropism to render a meaningful analysis of the impact of neurotropism on outcome.
The histological feature of desmoplasia in melanoma is associated with a characteristic phenotype that is different from other forms of melanoma in its initial presentation and in the occurrence of lymphatic spread. Local aggressiveness with thicker primary tumors and reduced frequency of lymphatic spread is observed more commonly with DM. Comparison to a contemporaneous case-matched control group reveals that overall, cause-specific, and disease-free survival are identical to thickness-matched controls with other forms of melanoma.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
D.P.L. was supported in part by a scholarship from the Dutch Cancer Society.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Quinn MJ, Crotty KA, Thompson JF, et al: Desmoplastic and desmoplastic neurotropic melanoma. Cancer 83:1128-1135, 1998
Conley J, Lattes R, Orr W: Desmoplastic malignant melanoma (a rare variant of spindle cell melanoma). Cancer 28:914-936, 1971
Payne WG, Kearney R, Wells K, et al: Desmoplastic melanoma. Am Surg 67:1004-1006, 2001
Jaroszewski DE, Pockaj BA, DiCaudo DJ, et al: The clinical behavior of desmoplastic melanoma. Am J Surg 182:590-595, 2001
Carlson JA, Dickersin GR, Sober AJ, et al: Desmoplastic neurotropic melanoma: A clinicopathologic analysis of 28 cases. Cancer 75:478-494, 1995
Skelton HG, Smith KJ, Laskin WB, et al: Desmoplastic malignant melanoma. J Am Acad Dermatol 32:717-725, 1995
Smithers BM, McLeod GR, Little JH: Desmoplastic, neural transforming and neurotropic melanoma: A review of 45 cases. Aust N Z J Surg 60:967-972, 1990
Jain S, Allen PW: Desmoplastic malignant melanoma and its variants. A study of 45 cases. Am J Surg Pathol 13:358-373, 1989
Egbert B, Kempson R, Sagebiel R: Desmoplastic malignant melanoma: A clinicohistopathologic study of 25 cases. Cancer 62:2033-2041, 1988
Su LD, Fullen DR, Lowe L, et al: Desmoplastic and neurotropic melanoma: Analysis of 33 patients with lymphatic mapping and sentinel lymph node biopsy. Cancer 100:598-604, 2004
Gyorki DE, Busam K, Panageas K, et al: Sentinel lymph node biopsy for patients with cutaneous desmoplastic melanoma. Ann Surg Oncol 10:403-407, 2003
Tsao H, Sober AJ, Barnhill RL: Desmoplastic neurotropic melanoma. Semin Cutan Med Surg 16:131-136, 1997
Balch CM, Urist MM, Karakousis CP, et al: Efficacy of 2-cm surgical margins for intermediate-thickness melanomas (1 to 4 mm): Results of a multi-institutional randomized surgical trial. Ann Surg 218:262-267, 1993
Reintgen D, Balch CM, Kirkwood J, et al: Recent advances in the care of the patient with malignant melanoma. Ann Surg 225:1-14, 1997
Kim CJ, Reintgen DS, Balch CM, et al: The new melanoma staging system. Cancer Control 9:9-15, 2002
Balch CM, Soong SJ, Gershenwald JE, et al: Prognostic factors analysis of 17,600 melanoma patients: Validation of the American Joint Committee on Cancer melanoma staging system. J Clin Oncol 19:3622-3634, 2001
Busam KJ, Mujumdar U, Hummer AJ, et al: Cutaneous desmoplastic melanoma: Reappraisal of morphologic heterogeneity and prognostic factors. Am J Surg Pathol 28:1518-1525, 2004
Posthner KE, Mosca PJ, Selim MA, et al: Histopathologic characteristics, recurrence patterns and survival of 129 patients with desmoplastic melanoma. Ann Surg Oncol 11:S78, 2004 (abstr 93)
Baer SC, Schultz D, Synnestvedt M, et al: Desmoplasia and neurotropism: Prognostic variables in patients with stage I melanoma. Cancer 76:2242-2247, 1995
Daugaard S: Current soft-tissue sarcoma classifications. Eur J Cancer 40:543-548, 2004
Ross JA, Severson RK, Davis S, et al: Trends in the incidence of soft tissue sarcomas in the United States from 1973 through 1987. Cancer 72:486-490, 1993
Chao C, Wong SL, Edwards MJ, et al: Sentinel lymph node biopsy for head and neck melanomas. Ann Surg Oncol 10:21-26, 2003
Smithers BM, McLeod GR, Little JH: Desmoplastic melanoma: Patterns of recurrence. World J Surg 16:186-190, 1992
Anstey A, McKee P, Jones EW: Desmoplastic malignant melanoma: A clinicopathological study of 25 cases. Br J Dermatol 129:359-371, 1994
Balch CM, Soong SJ, Murad TM, et al: A multifactorial analysis of melanoma. IV. Prognostic factors in 200 melanoma patients with distant metastases (stage III). J Clin Oncol 1:126-134, 1983
Cohn-Cedermark G, Mansson-Brahme E, Rutqvist LE, et al: Metastatic patterns, clinical outcome, and malignant phenotype in malignant cutaneous melanoma. Acta Oncol 38:549-557, 1999(Daan P. Livestro, Alona M)