To examine the role of vascular invasion as a prognostic factor in melanoma.
Retrospective survival analysis.
Academic medical center.
A total of 526 patients with primary cutaneous melanoma from the University of California, San Francisco, Melanoma Center database with 2 years of follow-up or documented relapse.
Main Outcome Measures
(1) Presence of vascular involvement defined as vascular invasion with tumor cells within blood or lymphatic vessels; or uncertain vascular invasion, with melanoma cells immediately adjacent to the endothelium. (2) Percentage with metastasis or death and relapse-free and overall survival.
The presence of either type of vascular involvement significantly increased the risk of relapse and death and reduced the survival associated with melanoma. The impact of vascular involvement on these outcomes was similar to that of ulceration. In a multivariate analysis, vascular involvement was the second most important factor (after tumor thickness) in the primary tumor in predicting survival.
Vascular involvement is an important independent predictor of metastasis and survival in melanoma. The phenomenon of uncertain vascular invasion describes an earlier step than definite vascular invasion in tumor progression.
MALIGNANT melanoma exhibits a capricious clinical behavior. Thus, while tumor thickness is the most important prognostic factor determining survival, many patients with thick melanomas are free of metastasis while a small subset of patients with thin tumors die of their disease. Therefore, information on additional factors is important to refine the prognostic information gained by knowledge of the tumor thickness.
Numerous clinical and histologic prognostic factors have been identified in melanoma.1,2 Clinical factors include sex, anatomical location of the primary tumor, and at times patient age. Histologic prognostic factors, in addition to Breslow tumor thickness and Clark level of invasion, include ulceration, mitotic index, regression, microsatellitosis, host lymphocytic response, and vascular invasion.
The interactions between the tumor cell and the tumor vasculature are increasingly recognized for their effect on the biological behavior of the primary tumor.3,4 These interactions include vascular invasion, defined as presence of tumor cells within the vessel lumens. Vascular invasion has been suggested to play an important role in the progression of a number of solid tumors.5,6 However, the role of this factor in the prognostic assessment of malignant melanoma has remained controversial. For instance, several studies from large databases of patients with melanoma show conflicting results with respect to the importance of vascular invasion.7-12 Studies analyzing the role of vascular invasion in melanoma have been hampered by small patient numbers, inadequate follow-up, and/or the use of different techniques for optimal recognition of these tumor cell–tumor vasculature interactions.13,14 As a result, the presence of vascular invasion is not routinely documented in melanoma pathology reports. For this study, using a database of 526 patients, we evaluated the prognostic significance of vascular involvement in the metastasis and survival associated with malignant melanoma.
From the archives of the University of California, San Francisco, Melanoma Center, we selected 526 patients with primary cutaneous melanoma using the following eligibility criteria: (1) at least 2 years of follow-up or first relapse unequivocally documented; and (2) sufficient clinical and histologic information to be evaluable. Different types of data missing from the archives for some patients explain slight variations in calculation results. The clinical parameters included were age, sex, location of the primary tumor, relapse-free survival (RFS), clinical stage, site of first metastasis, and overall survival (OS). Metastasis was defined to include local dermal or subcutaneous recurrence, satellitosis, and regional nodal and distant metastasis.
Only routinely prepared slides with hematoxylin-eosin staining were used. The approach to the histologic analysis routinely performed has been previously described.15 The histologic parameters were recorded prospectively and included the following measures: tumor thickness, Clark level, mitotic index (defined as number of mitoses per square millimeter in the vertical growth phase), and presence of ulceration and/or microsatellites. Vascular invasion was defined as tumor cells fixed to the walls and within the lumens of lymphatic or blood vessels. Only those cases with unequivocal findings were recorded as demonstrating vascular invasion.
Statistical analysis of the cross-tabulated frequency data was performed via (1) the Fisher exact test in the case of all 2 × 2 tables; and (2) the Pearson χ2 test if either the number of rows in the table or the number of columns or both exceeded 2. Statistical analysis of all differences between means (eg, mean Breslow tumor thickness and mean elapsed time intervals until relapse or death) was performed via (1) the 2-sample (independent) t test in the case of all 2-sample comparisons; and (2) a 1-way analysis of variance (ANOVA, fixed effects model) if more than 2 means were compared. In the case of mean elapsed time intervals, censored observations were excluded. Where indicated, survival curves were generated using the Kaplan-Meier method,16 and differences between Kaplan-Meier curves were assessed using both log-rank and generalized Wilcoxon tests. Multivariate comparisons of the relative contributions of various high-risk prognostic factors were assessed via Cox regression analysis, assuming proportional hazards. Moreover, tests for possible interactions were performed in the multivariate analyses. All tests of statistical significance reported are 2-sided.
Table 1 gives the characteristics of the 526 patients included in the data set. We analyzed all patients' tumors at the time of initial diagnosis (prospectively) for evidence of vascular invasion and other high-risk histologic features. Two patterns of vascular involvement were noted in 78 patients (16%). In the first pattern, termed definite vascular invasion (45 cases [9%]), tumor cells were found invading the endothelium and vascular lumen (Figure 1A). In the second pattern, termed uncertain vascular invasion (33 cases [7%]), melanoma cells were observed immediately adjacent to the endothelium without intervening stroma, but not within the walls or lumens (Figure 1B).
There were no significant differences between the 2 groups of patients with vascular involvement and the 1 without with respect to age, sex, or location of the primary tumor (Table 1). However, there were important differences between the 2 groups with respect to tumor thickness and ulceration, both previously shown to correlate with vascular invasion.13,14
We examined the potential significance of vascular involvement on the metastatic phenotype and survival of patients with melanoma in this data set. Metastasis was documented in 42% of patients without vascular involvement compared with 82% of patients with vascular involvement (including both definite and uncertain cases, P<.001). Deaths were recorded in 19% of cases without vascular involvement compared with 44% of cases with either type of involvement (P<.001). There was no difference between the 2 patterns of vascular involvement with respect to the eventuality of metastasis or death. Moreover, 32% of patients without vascular invasion had regional nodal involvement compared with 59% of patients with either pattern of vascular involvement (P = .006). Distant metastasis was observed in 22% of patients without vascular involvement compared with 74% of patients with uncertain invasion and 65% of patients with definite invasion (P<.001). Finally, the effect of vascular involvement was examined on the RFS and OS of the patients in the data set. In patients without vascular invasion, the mean RFS was 26.1 months compared with 20.6 months in patients with uncertain vascular invasion, and 10.2 months in patients with definite vascular invasion (P<.001). The mean OS in cases without vascular involvement (54.2 months) was significantly longer than in cases with vascular involvement (32.6 months) (P = .001). Kaplan-Meier analysis of RFS and OS revealed significant reductions in survival with the presence of vascular involvement (Figure 2).
To determine the independent prognostic significance of vascular involvement in melanoma, 2 multivariate analyses were performed: RFS and OS were regressed separately against several well-known histologic variables, including tumor thickness, ulceration, Clark level, mitotic index, and microsatellitosis (Table 2). Tumor thickness was the only factor significant at P<.05 in the RFS and OS analysis. Clark level was significant (P = .01) in the RFS analysis and marginally significant (P = .06) in the OS analysis. Ulceration was marginally significant (P = .05) in the OS analysis, and nonsignificant (P = .09) in the RFS analysis. Then vascular involvement was added as a sixth prognostic variable to each analysis (Table 3). Tumor thickness remained the most potent prognostic factor in both analyses. Interestingly, vascular involvement emerged as the second most potent factor after tumor thickness in both analyses (Table 3). Given the recognized importance of ulceration and its inclusion in the proposed staging of melanoma,17,18 we compared the effect of ulceration and vascular involvement, taken separately, on several outcome parameters in this data set. As given in Table 4, these 2 prognostic factors have a virtually identical effect on the risk of relapse and death and the survival associated with melanoma.
To our knowledge, our work here constitutes the largest study of the role of vascular invasion as a prognostic factor in melanoma (526 patients with melanoma, of whom 78 had vascular involvement). The results of our analysis clearly indicate the independent role of vascular involvement in the prognostic assessment of patients with melanoma. The presence of vascular involvement was associated with an increased risk of melanoma relapse and death. Interestingly, involvement of the tumor vasculature, which is believed to represent largely lymphatic vessels, was predictive of distant as well as regional nodal metastasis. Moreover, vascular involvement was associated with significantly reduced RFS and OS. In fact, vascular involvement was the most important prognostic factor in the primary tumor after tumor thickness in determining survival.
Prior studies of definite vascular invasion have suggested the high-risk nature of this prognostic factor. Followfield and Cook13 analyzed 66 primary melanomas, 29 of which demonstrated vascular invasion as determined by lectin Ulex europaeus I staining. Vascular invasion was an uncommon finding in melanomas thinner than 1.5 mm, but increased with increasing tumor thickness. However, no correlation with outcome was reported. Subsequently, a population-based study of 476 patients with melanoma from the Swedish Cancer Registry revealed a 27.3% 5-year survival in cases with vascular invasion compared with 76.1% 5-year survival in cases without vascular invasion.19 Vascular invasion was associated with a 2.11-fold increased hazard of death and was an independent predictor of survival.19
More recently, Straume and Akslen14 examined 102 patients with nodular melanoma, including 15 with vascular invasion as determined by staining with factor VIII antibody. Vascular invasion was correlated with tumor thickness, ulceration, and initial patient stage. In univariate analysis, vascular invasion was a significant predictor of overall survival. In multivariate analysis, vascular invasion was of independent prognostic significance in predicting survival, but its effect was lost after adjusting for other factors.
Of particular interest was the prognostic significance of the phenomenon of uncertain vascular invasion. In this cohort, the tumor cells were not unequivocally found within the tumor vasculature; rather, they abutted the vessels without intervening stroma (Figure 1B) and therefore could not be classified as demonstrating definite invasion. However, the risk of metastasis and death from such tumors was as high as those exhibiting definite vascular invasion, suggesting the poor prognosis associated with this finding. There was no difference between the 2 patterns of vascular involvement with respect to overall relapse or death rate, and rate of regional, nodal, or distant metastasis. However, the mean RFS of patients with uncertain vascular invasion was 20 months, compared with 10 months for those with definite invasion. This suggests that the group with uncertain vascular invasion occupies an important earlier step in tumor progression.
Other investigators have described similar interactions between melanoma cells and vascular endothelial cells. The group with uncertain vascular invasion is similar morphologically to patients with the angiotumoral complex described by some investigators20 and to those with angiotropism described by others.21,22 More recently, the intriguing concept of vasculogenic mimicry described for ocular melanomas23 has suggested unanticipated interactions between melanoma cells and the tumor vasculature. Finally, recent studies have suggested the involvement of the β2 chain of laminin in the microvasculature associated with melanoma.24 Further studies will be required to define the molecular nature of these interactions given their prognostic significance in human melanoma.
In our analysis, vascular involvement was an independent prognostic factor determining relapse and survival in patients with melanoma. Our data suggest that the presence of either type of vascular involvement imparts a high-risk feature to the tumor similar to that associated with ulceration. In fact, in a multivariate analysis, vascular involvement had a greater effect than ulceration on both RFS and OS (Table 3).
More recently, the histologic status of the regional nodal basin, as determined by sentinel lymph node (SLN) mapping, has emerged as a powerful indicator of melanoma survival.25 An analysis by our group26 has shown that vascular invasion, along with ulceration, mitotic index, and microsatellitosis, are histologic features predictive of a positive SLN finding.26 A recent multivariate analysis performed by our group using a data set of patients undergoing SLN mapping revealed that vascular invasion remained a significant predictor of relapse and death when SLN status was included in the model (Leong SPL, Nguyen LH, Miller JR, et al; manuscript submitted, 2001). Importantly, in patients with a positive SLN, vascular invasion was one of the few factors in the primary tumor that was still significant in predicting outcome. This analysis, which was conducted with a largely different set of patients, validates the independent prognostic effect of vascular invasion even in the context of SLN mapping.
A salient feature of our present study was the histopathologic recording of prognostic features using routine hematoxylin-eosin stains. Studies using immunocytochemical stains may be difficult to compare with the present study. Ulex europaeus is thought to stain postcapillary venules and lymphatics,27 whereas factor VIII is regarded as a panendothelial marker that stains blood vessels more strongly than lymphatics.14 Involvement of the vascular spaces recorded here was graded solely on the basis of routinely stained slides. The advantage is that every histopathologist can evaluate this factor; the disadvantage is the lack of information regarding the nature of the vessel. However, solid evidence to indicate whether separation of the vascular and lymphatic circulation is predictive of the site of metastasis in melanoma is lacking. Interestingly, the percentage of cases with vascular involvement in our study (16%) was similar to that reported by Straume and Akslen14 (15%) in which factor VIII antibody staining was used to record vascular invasion.
In summary, our results underscore the importance of vascular involvement as a histologic prognostic factor in malignant melanoma. We propose that recording of vascular involvement (which requires no additional cost or techniques) in reports of primary malignant melanoma can play an important role in refining the prognosis of patients with melanoma as well as in determining their subsequent treatment.
Accepted for publication May 2, 2001.
Dr Kashani-Sabet was supported by the Leaders Society Clinical Career Development Award of the Dermatology Foundation, Evanston, Ill. Dr Ferreira was supported by a fellowship sponsored by the Conselho Nacional de Desenvolvimento Cientifico e Tecnologico–Conselho Nacional de Pesquisas, Brasilia–São Paulo, Brazil.
We thank Celia Hamilton for preparation of the manuscript.
Corresponding author and reprints: Richard W. Sagebiel, MD, University of California, San Francisco, Cancer Center, 1600 Divisadero St, Fourth Floor, San Francisco, CA 94115 (e-mail: firstname.lastname@example.org).
RW Prognostic factors in malignant melanoma. Surg Oncol Clin N Am.
1997;6599- 623Google Scholar
MEed Factors predictive of metastasis and outcome. Cutaneous Oncology: Pathophysiology, Diagnosis and Management
Malden, Mass Blackwell Science Inc1998;278- 291Google Scholar
J What is the evidence that tumors are angiogenesis dependent? J Natl Cancer Inst.
1990;824- 6Google ScholarCrossref
RL The biology of melanoma micrometastases. Recent Results Cancer Res.
2001;1583- 13Google Scholar
S Prognostic factors in gastric cancer: the value of vascular invasion, mitotic rate and lymphoplasmacytic infiltration. Br J Cancer.
1996;74766- 772Google ScholarCrossref
M Pathological prognostic factors in breast cancer, III: vascular invasion—relationship with recurrence and survival in a large study with long-term follow-up. Histopathology.
1994;2441- 47Google ScholarCrossref
D Model for predicting survival in stage I melanoma based on tumor progression. J Natl Cancer Inst.
1989;811893- 1904Google ScholarCrossref
G Therapeutic and clinical-pathological factors in the survival of 1,469 patients with primary cutaneous malignant melanoma in clinical stage I: a multivariate regression analysis. Virchows Arch (Pathol Anat).
1985;408249- 258Google ScholarCrossref
et al. Stage I cutaneous malignant melanoma: risk factors of loco-regional recurrence after wide local excision and clinical perspectives. Eur J Surg Oncol.
1992;18442- 448Google Scholar
GWed An analysis of prognostic factors in 8500 patients with cutaneous melanoma. Cutaneous Melanoma
2nd ed Philadelphia, Pa JB Lippincott1992;165- 198Google Scholar
HF Skin melanoma: cured fraction and survival time as functions of thickness, site, histologic type, age and sex. Cancer.
1993;721219- 1223Google ScholarCrossref
DC Prognostic models for sub-groups of melanoma patients from the Scottish Melanoma Group database 1979-86, and their subsequent validation. Br J Cancer.
1995;71173- 176Google ScholarCrossref
MG Vascular invasion in malignant melanomas: an independent prognostic variable? Am J Surg Pathol.
1989;13217- 220Google ScholarCrossref
LA Independent prognostic importance of vascular invasion in nodular melanomas. Cancer.
1996;781211- 1219Google ScholarCrossref
RW The pathology of melanoma as a basis for prognostic models: the UCSF experience. Pigment Cell Res.
1994;7101- 103Google ScholarCrossref
P Nonparametric estimations from incomplete observations. J Am Stat Assoc.
1958;53457- 481Google ScholarCrossref
et al. Critical analysis of the current American Joint Committee on Cancer staging system for cutaneous melanoma and proposal of a new staging system. J Clin Oncol.
1997;151039- 1051Google Scholar
et al. A new American Joint Committee on Cancer staging system for cutaneous melanoma. Cancer.
2000;881484- 1491Google ScholarCrossref
HO Clinical and histopathologic predictors of survival in patients with malignant melanoma: a population-based study in Sweden. J Natl Cancer Inst.
1994;86761- 769Google ScholarCrossref
JP Is the angio-tumoral complex in human malignant melanoma a marker of non-hematogenous migration? [abstract]. Melanoma Res.
V Angiotropic malignant melanoma: report of two cases. J Cutan Pathol.
1992;19325- 329Google ScholarCrossref
C Angiotropic melanoma: report of seven additional cases [abstract]. J Cutan Pathol.
et al. Vascular channel formation by human melanoma cells in vivo and in vitro: vasculogenic mimicry. Am J Pathol.
1999;155739- 752Google ScholarCrossref
RL Tumor microvessels in melanoma express the beta-2 chain of laminin: implications for melanoma metastasis. J Cutan Pathol.
1999;26222- 226Google ScholarCrossref
et al. Multi-institutional melanoma lymphatic mapping experience: the prognostic value of sentinel lymph node status in 612 stage I or II melanoma patients. J Clin Oncol.
1999;17976- 983Google Scholar
SPL Prediction of sentinel lymph node micrometastasis by histologic features in primary cutaneous malignant melanoma. Arch Dermatol.
1998;134983- 987Google ScholarCrossref
et al. Tumor vascularity is not a prognostic factor for skin melanoma. Am J Pathol.
1995;1471049- 1056Google Scholar