Includes the variables that were significant in the multivariate study.
eMethods. Mutational Analysis
eTable 1. Characteristics of the Population According to the Pattern of Dissemination
eTable 2. Univariate Cox Proportional Hazards Regression Analysis for Lymphatic Metastasis and Hematogenous Metastasis
eFigure. Histogram of Months of Follow-up
Customize your JAMA Network experience by selecting one or more topics from the list below.
Calomarde-Rees L, García-Calatayud R, Requena Caballero C, et al. Risk Factors for Lymphatic and Hematogenous Dissemination in Patients With Stages I to II Cutaneous Melanoma. JAMA Dermatol. 2019;155(6):679–687. doi:10.1001/jamadermatol.2019.0069
What are the specific risk factors for recurrence of lymphatic and hematogenous dissemination among patients with stages I to II melanoma?
In this cohort study of 1177 patients with melanoma, age, site of primary tumor, tumor thickness, and vascular invasion were associated with lymphatic recurrence, whereas thickness, absence of regression, TERT promoter, and BRAF mutation were associated with hematogenous metastasis.
With validation, these findings suggest that follow-up and adjuvant treatment strategies should be adapted to individual clinical, histopathologic, and molecular characteristics.
The lymphatic and the hematogenous pathways have been proposed for disease progression in cutaneous melanoma, but association with recurrence has not been studied separately to date.
To identify the risk factors associated with lymphatic and hematogenous metastasis.
Design, Setting, and Participants
This retrospective cohort study included 1177 patients with malignant melanoma treated at Instituto Valenciano de Oncología, València, Spain. Data were retrieved from the melanoma database from January 1, 2000, through December 31, 2015, and analyzed from June 1 to 30, 2018.
Malignant melanoma at stages I to II.
Main Outcomes and Measures
Analyses of survival free of lymphatic and hematogenous metastasis were performed using Kaplan-Meier curves and Cox proportional hazards regression.
For the 1177 patients included in the study analysis (51.1% women; median age at diagnosis, 55 years [interquartile range, 42-68 years), median follow-up was 75 months (interquartile range, 33-121 months); 108 (9.2%) developed lymphatic metastasis, and 108 (9.2%) developed hematogenous metastasis. In the multivariate analysis, being older than 55 years (hazard ratio [HR], 1.9; 95% CI, 1.2-3.1), tumor in the head/neck (HR, 1.7; 95% CI, 1.0-2.9) and acral locations (HR, 2.4; 95% CI, 1.3-4.5), greater Breslow thickness (HR for >4.00 mm, 5.4; 95% CI, 2.4-12.4), and presence of vascular invasion (HR, 3.2; 95% CI, 0.9-10.6) were associated with lymphatic spreading. Hematogenous metastasis was associated with greater Breslow thickness (HR for >4.00 mm, 10.4; 95% CI, 3.6-29.7), the absence of regression (HR, 0.1; 95% CI, 0.0-1.0), TERT promoter mutations (HR, 2.9; 95% CI, 1.5-5.7), and BRAF mutations (HR, 1.9; 95% CI, 1.1-3.6).
Conclusions and Relevance
Risk factors associated with lymphatic and hematogenous metastasis differ. Follow-up and adjuvant treatment strategies may therefore need to be adapted to individual clinical, histopathologic, and molecular characteristics.
Melanoma is a highly metastatic disease. It accounts for 1.5% of all cancers in men and women, and incidence and mortality rates in developed countries have risen in recent decades.1 In Spain, incidence is 8.76 new cases per 100 000 inhabitants per year.2
Cutaneous melanoma can spread through the blood to distant organs (hematogenous metastasis) or through the lymphatic system to locoregional skin and lymph nodes (lymphatic metastasis). Three classic models of dissemination have been proposed. The first, described by Morton in 2012,3 is the sequential metastasis model, which holds that melanoma mainly spreads from the sentinel lymph node (SLN) to the regional nodes via the lymphatic system and then, after a variable latency period, to distant sites via the bloodstream. According to this model, SLN biopsy is essential in all patients with a diagnosis of melanoma.4
Investigators who do not believe that the systematic performance of SLN biopsy is justified, such Medalie and Ackerman,5 defend the simultaneous metastasis model, which holds that melanoma spreads through the blood and lymphatic system at the same time. According to this model, tumor cells in the SLN would be a marker for hematogenous and lymphatic metastasis.
The third and final model, the differential metastasis model supported by Clark,6 differs from the other models because it takes a holistic view of independent dissemination pathways.7 According to Clark,6 not all melanomas have the ability to metastasize, and those that do largely spread through blood and the lymphatic system, producing lymphatic and hematogenous metastases. The remaining tumors would spread to the regional lymph nodes or to distant sites through the bloodstream. This third model would explain why complete, lasting remission is observed in almost 30% of patients with melanoma, regardless of the lymph node dissection performed (prophylactic, immediately after a positive SLN biopsy finding, or delayed).8 These cases would correspond to patients with melanomas that are only capable of developing lymphatic metastasis. This third model would also explain why prophylactic and SLN-guided lymph node dissection do not improve final survival outcomes, because they would not be curative in patients with simultaneous nodal and distant metastasis.9-11 Finally, the third model would also explain why a negative SLN biopsy finding does not guarantee survival in patients with melanomas that develop exclusively hematogenous metastasis.
According to the model proposed by Clark,6 disease progression has distinct pathways. Which factors favor one pathway or another remain unknown. The overall picture can be further compounded by the phenomenon of cancer dormancy that delays metastasis development through mechanisms that include angiogenic dormancy, cellular dormancy, and immunosurveillance. The immune system can restrain disseminated cancer cells, promoting permanent dormancy where CD8+ T lymphocytes are important in maintaining immune equilibrium with metastatic dormant cells.12,13 Tumor-draining lymph nodes contain tumor effector as well as suppressor immune components.14 Our aim was to identify risk factors associated with lymphatic (locoregional metastasis) or hematogenous (distant metastasis) progression because these have not been studied separately to date in patients with localized melanoma.
We designed a longitudinal cohort study drawing on data from the melanoma database held by the Dermatology Department at the Instituto Valenciano de Oncología, València, Spain. This database contains prospectively recorded data on patients with melanoma diagnosed and treated at the Instituto Valenciano de Oncología since January 1, 2000. We included patients diagnosed to December 31, 2015. The database meets all legal requirements and was approved by the ethics committee of Instituto Valenciano de Oncología. Patients gave written informed consent for data collection.
The database contained 1784 patients at the time of the study, but we excluded those with melanoma in situ (n = 215), extracutaneous melanoma (n = 27), melanoma of unknown origin (n = 40), presence of metastasis at diagnosis (lymphatic or hematogenous) (n = 241), and more than 1 invasive melanoma (n = 84). The final sample thus included 1177 patients with stages I to II melanoma classified according to the seventh edition of the American Joint Committee on Cancer Staging System.15 Guidelines for staging in our center included SLN biopsy for all melanomas thicker than 0.75 mm and for those melanoma 0.75 mm or thinner if the tumor was ulcerated or had vascular invasion, microscopic satellite, or at least 1 mitosis present in the dermal component. Also, a minor proportion (29 of 159) of patients with stages IIB to IIC disease received adjuvant therapy with high-dose interferon. The main variable was the site of the first metastasis, and we distinguished between lymphatic dissemination (ie, with locoregional cutaneous or lymph node metastasis [stage III]) and hematogenous metastasis (ie, with involvement of distant organs [stage IV]).
The variables analyzed included age (categorized according to the median age of the study population), sex, tumor location (head/neck, upper extremities, trunk, lower extremities, or acral sites), Breslow thickness (≤1.00, 1.01-2.00, 2.01-4.00, or >4.00 mm), ulceration, tumor-infiltrating lymphocytes (none, brisk, or nonbrisk), tumor mitotic rate (0, 1-5, or >5 mitoses/mm2), regression, vascular invasion (defined as the unequivocal presence of tumor cells within the vascular lumen adhered to the endothelium), and mutational status for BRAF (OMIM 164757), NRAS (OMIM 164790), or TERT (OMIM 187270) promoter (description of the methods for mutational status is included in eMethods in the Supplement). The mutational analysis was routinely performed for all the patients with enough available and suitable material. Because most of our cases were referred for definitive management, we did not have sufficient material for genetic testing in many cases.
Data were analyzed from June 1 to 30, 2018. All melanomas with vascular invasion were tested for antibody D2-40 (podoplanin) and CD31 immunohistochemical staining. We used χ2 and Fisher exact tests in 2 × 2 tables when the expected frequency was less than 5 to compare defined groups. Survival analysis was performed using Kaplan-Meier curves and the log-rank test. For each analysis of development of lymphatic or hematogenous metastasis, patients who did not develop the defined event were censored. The size of the effect on survival due to each variable was first explored by univariate Cox proportional hazards regression analysis. The assumption of proportionality was evaluated graphically using a log-log graph. Then, stepwise forward multivariate Cox proportional hazards regression models were used to search for independent risk factors. All variables with P ≥ .10 were entered in each model. All tests were 2-sided, and P < .05 was considered statistically significant. All statistical tests were run in SPSS software (version 20.0; IBM Corp).
Of the 1177 patients included in the study, 601 (51.1%) were women and 576 (48.9%) were men. The median age of the group at diagnosis was 55 years (interquartile range, 42-68 years). During a median follow-up of 75 months (range, 0-209 months; interquartile range, 33-121 months) (eFigure in the Supplement), 1026 patients (87.2%) did not develop metastasis, 108 (9.2%) developed lymphatic metastasis, and 108 (9.2%) developed hematogenous metastasis. The characteristics of the population are shown in Table 1. Of those cases that developed metastasis, 65 (5.5%) had lymphatic and hematogenous metastasis. The characteristics of this group are shown in eTable 1 in the Supplement. Melanomas in this group presented the highest prevalence of BRAF and NRAS mutations and a relatively high prevalence of acral melanomas. The following variables were statistically significant in the univariate analysis of risk factors for lymphatic metastasis: being older than 55 years (hazard ratio [HR], 2.2; 95% CI, 1.5-3.3; log-rank P < .001), location in the head/neck (HR, 3.0; 95% CI, 1.9-4.6; log-rank P < .001) and acral sites (HR, 3.8; 95% CI, 2.2-6.5; log-rank P < .001), ulceration (HR, 2.8; 95% CI, 1.8-4.3; log-rank P < .001), a higher mitotic rate (HR for >5 mitoses/mm2, 4.5; 95% CI, 2.2-9.3; log-rank P < .001), vascular invasion (immunohistochemistry demonstrated the lymphatic nature of the vessels in all cases) (HR, 5.7; 95% CI, 2.3-13.9; log-rank P < .001), and the presence of BRAF mutations (HR, 1.9; 95% CI, 1.2-3.2; log-rank P = .004) (Table 1 and eTable 2 in the Supplement). The corresponding risk factors for hematogenous metastasis were being older than 55 years (HR, 1.8; 95% CI, 1.3-2.7; log-rank P = .002), a protective effect of female sex (HR, 0.5; 95% CI, 0.3-0.7; log-rank P < .001), location in the head/neck (HR, 3.1; 95% CI, 2.0-4.7; log-rank P < .001), greater Breslow thickness (HR for 1.01-2.00 mm, 4.2; 95% CI, 2.3-7.8; log-rank P < .001), ulceration (HR, 4.4; 95% CI, 2.9-6.6; log-rank P < .001), high mitotic rate (HR for >5 mitoses/mm2, 13.2; 95% CI, 5.9-29.3; log-rank P < .001), absence of regression (HR, 0.3; 95% CI, 0.1-0.8; log-rank P = .007), and the presence of BRAF (HR, 2.5; 95% CI, 1.6-4.1; log-rank P < .001) and TERT (HR, 2.8; 95% CI, 1.6-4.7; log-rank P < .001) mutations (Table 1 and eTable 2 in the Supplement). The distribution of variables according to 5-year lymphatic and hematogenous metastasis–free survival is shown in Table 2.
In the multivariate Cox proportional hazards regression analysis, the following variables retained their significance as independent risk factors for lymphatic metastasis: being older than 55 years (HR, 1.9; 95% CI, 1.2-3.1; P = .01), head/neck (HR, 1.7; 95% CI, 1.0-2.9; P = .04) and acral (HR, 2.4; 95% CI, 1.3-4.5; P = .005) locations, greater Breslow thickness (HR for >4.00 mm, 5.4; 95% CI, 2.4-12.4; P < .001), and presence of vascular invasion or satellite lesions (HR, 3.2; 95% CI, 0.9-10.6; P = .052) (Table 3 and Figure 1). The independent risk factors for hematogenous metastasis were greater Breslow thickness (HR for >4.00 mm, 10.4; 95% CI, 3.6-29.7; P < .001) and TERT (HR, 2.9; 95% CI, 1.5-5.7; P = .001) and BRAF (HR, 1.9; 95% CI, 1.1-3.6; P = .03) mutations; histologic regression exerted a protective effect (Table 3 and Figure 2). We further explored the prognostic value of the additive effect of the gene mutations (mutational status in the 4 mutually exclusive categories of wild type, MAPK pathway [BRAF or NRAS] mutation, TERT promoter mutation, and MAPK pathway and TERT promoter mutation), and multivariate analysis showed that, besides Breslow thickness and the absence of regression, worse prognosis occurred in tumors with TERT promoter mutation tumors (HR, 2.5; 95% CI, 0.9-7.1; P = .08) and a strongly significant worse prognosis for those patients with tumors harboring a combination of TERT promoter mutation and a mutation in either BRAF or NRAS (HR, 5.7; 95% CI, 2.2-14.5; P < .001) (model II for hematogenous metastasis in Table 3).
The factors that determine hematogenous and lymphatic metastasis in melanoma are currently unknown. Although several studies have analyzed time to recurrence and survival according to the characteristics of different types of melanoma, very few have identified factors associated with each of the metastatic pathways, making it difficult to determine the association with disease progression.8,16
Meier et al17 and Adler et al18 used data from the German Central Malignant Melanoma Registry to investigate clinical and histologic risk factors associated with different metastatic pathways in melanoma. The investigators demonstrated significant associations with sex and Breslow thickness and a particularly strong association with anatomical location. They also found different patterns of disease progression for melanomas on the trunk and upper limbs compared with those on the head and neck or lower limbs.17,18
In a Swedish study similar to ours, Cohn-Cedermark et al19 found that hematogenous metastasis was more strongly associated with head and neck melanomas than melanomas in other locations. Head and neck melanomas are aggressive; they are more likely to recur and are associated with considerably lower overall survival than melanomas at other sites. Why this recurrence occurs is unknown. Various theories have been proposed to explain the reason for a particular affinity of the metastatic disease to melanoma on the scalp, including the rich blood supply to the area.20-22 Melanomas of the scalp could behave similarly to lung or renal cell carcinoma, in which cells show a clear tendency to invade blood vessels and develop hematogenous metastasis.20,23-25
Patient sex, like anatomical location, is also of independent significance in the different pathways of melanoma progression.17,26 In a German population cohort study,26 female sex was associated with a lower risk of progression via lymphatic and hematogenous routes, and women showed better survival rates overall. Also in Germany, Mervic27 investigated differences in the time course and patterns of metastatic spread in 7338 men and women from the German Central Malignant Melanoma Registry. She found that, although men and women developed a similar number of lymphatic and hematogenous metastases, distant metastases developed later in women (40.5 vs 33.0 months).27
Tumor thickness has also been identified as an independent factor for the progression of melanoma through different routes.17,18 Our study shows that the prognostic factors associated with lymphatic and hematogenous metastasis differ, supporting the model of differential spread by Clark.6 According to our multivariate analysis, the factors associated with the greatest risk of lymphatic metastasis included being older than 55 years, location, greater Breslow thickness, and presence of vascular invasion.
As mentioned, vascular invasion was significantly associated with lymphatic metastasis only, suggesting that tumor emboli are preferentially located in lymph vessels, a fact that was confirmed in our series because the lymphatic nature of invaded vessels was routinely assessed by immunohistochemistry. Several studies have shown that lymphovascular invasion is by far the most common type of vascular invasion in melanoma.28-31 The clinical relevance of antibody D2-40 has also been demonstrated in numerous studies,32-36 and this marker should perhaps be routinely investigated as a histologic risk factor for lymphatic metastasis in melanoma.
The risk factors independently associated with hematogenous metastasis in our series were greater Breslow thickness and presence of TERT and BRAF mutations. The presence of histologic regression was a protective factor.
Histologic regression was observed mostly in patients who did not develop metastatic disease, and it was least common in those who developed distant metastasis. These observations indicate that the presence of tumor regression could perhaps help to distinguish between lymphatic and hematogenous metastasis and suggest that regression could be associated with good outcomes, contrasting with the traditional view that histologic regression indicates poor prognosis and can lead to underestimation of tumor.37,38 Our findings do, however, coincide with recent findings showing a protective role for histologic regression and suggest that it could be a histological marker of a good immunological host response against the tumor cells.39,40
TERT promoter and BRAF mutations were risk factors for hematogenous metastasis, supporting the idea that adjuvant targeted therapy could help stop the spread of disease through the blood in patients carrying these mutations. Furthermore, their effect was stronger when they were simultaneously present. This finding also suggests that the molecular tumor profile is a major determinant of hematogenous dissemination. The apparent protective effect of regression against distant metastasis also supports the use of adjuvant immunotherapy in patients with TERT promoter and BRAF mutations. According to our findings, we presume that patients with a high risk of hematogenous metastasis would not benefit from SLN biopsy, because a negative result would not guarantee complete disease remission owing to the independence of the hematogenous and lymphatic routes.7
Limitations of our study include the retrospective design based on data collected at a single institution. The high number of patients with missing data for mutational status is another limitation.
Risk factors for lymphatic and hematogenous metastasis differ. A greater understanding of the clinical, histopathologic, and molecular factors involved could help to identify patients with an increased risk of recurrence and guide the design of individualized follow-up programs and adjuvant targeted therapies.
Accepted for Publication: January 9, 2019.
Corresponding Author: Eduardo Nagore, MD, PhD, School of Medicine, Universidad Católica de Valencia San Vicente Mártir, c/Quevedo, 2, 46001 València, Spain (firstname.lastname@example.org).
Published Online: May 1, 2019. doi:10.1001/jamadermatol.2019.0069
Author Contributions: Drs Calomarde-Rees and García-Calatayud contributed equally to this work. Dr Nagore had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Calomarde-Rees, García-Calatayud, Manrique-Silva, Nagore.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Calomarde-Rees, García-Calatayud, Manrique-Silva, Kumar, Nagore.
Critical revision of the manuscript for important intellectual content: García-Calatayud, Requena Caballero, Traves, García-Casado, Soriano, Kumar, Nagore.
Statistical analysis: Calomarde-Rees, García-Calatayud, Nagore.
Obtained funding: Calomarde-Rees, Manrique-Silva.
Administrative, technical, or material support: García-Calatayud, Requena Caballero, Manrique-Silva, Traves, García-Casado.
Supervision: García-Calatayud, Requena Caballero, Kumar, Nagore.
Conflict of Interest Disclosures: None reported.
Create a personal account or sign in to: