Lymphatic channels play an important role in diverse forms of human disease, including inflammation, wound healing, hypertension, obesity, and cancer (Alitalo et al., 2005; Tammela and Alitalo, 2010). Preclinical studies have shown that solid cancers produce lymphangiogenic growth factors such as those belonging to the vascular endothelial growth factor (VEGF) family. In addition to promoting new lymphatic vessel formation (lymphangiogenesis), such factors promote tumor cell adhesion to, and migration of tumor cells into, lymphatic vessels (Achen et al., 2001; Christiansen and Detmar, 2011; Ferrara et al., 2003; Padera et al., 2002; Rinderknecht and Detmar, 2008; Skobe et al., 2001a,b; Stacker et al., 2001; Swartz and Skobe, 2001). The VEGF-related gene family of angiogenic and lymphangiogenic growth factors comprises seven secreted glycoproteins referred to as VEGF-A, VEGF-B, VEGF-C, VEGF-D, VEGF-E, and placenta growth factor (PlGF) -1 and -2 (Hicklin and Ellis, 2005). VEGF-A (commonly referred to simply as VEGF) was the first factor identified (Senger et al., 1983) and is able to couple both the receptors VEGFR-1 and -2, thereby contributing to neoangiogenesis. However, VEGF-C and VEGF-D and their receptor VEGFR-3 have been shown to be the most relevant factors for lymphangiogenesis in human cancers (Hicklin and Ellis, 2005).
Activation of the VEGF/VGFR pathway, through its regulation of various cytokines and signaling cascades such as NOTCH, Neuropilins, Survivin, Cadherin, MAP kinase, nitric oxide (NO), and calmodulin, causes several events including endothelial progenitor cell recruitment, endothelial and cancer cell proliferation and survival, and maturation of dendritic cells (www.mmmp.org, 2008). Ultimately, the secretion of lymphangiogenetic factors and their interactions lead to the formation of new vessels and to the transmigration of tumor cells into vascular spaces which, in turn, results in cancer dissemination and metastasis (Hicklin and Ellis, 2005).
In cutaneous melanoma, the presence of tumor cells within lymphatic vessels (i.e., lymphovascular invasion, LVI) is occasionally identified in routine pathologic examination of hematoxylin and eosin-stained sections and its presence has been strongly correlated with poor patient outcome (Scoggins et al., 2006). Recently, the availability of antibodies with high sensitivity and specificity for lymphatic vessels (e.g., LYVE-1 and D2-40), melanoma (S-100, Mart-1), proliferating cells (Ki-67 or PCNA), and other lymphangiogenetic markers (e.g., VEGFR3, VEGF-C, and -D) has facilitated more accurate detection of LVI, accurate quantitation of lymphatic vessel density (LVD) and new lymphatic vessel formation within and surrounding the tumor (Ji, 2006; Xu et al., 2008). Up to 50% of patients with no H&E-detected LVI have melanoma cells within lymphatic vessels after evaluation with immunohistochemistry (IHC) (Petersson et al., 2009; Petitt et al., 2009; Rose et al., 2011). Increasing tumor thickness, ulceration, regression, and higher mitotic rate as well as nodular melanoma subtype are factors associated with the presence of melanoma cells within lymphatics (LVI) (Da Costa et al., 2011; Rose et al., 2011; Storr et al., 2012).
By using the above-mentioned markers, some studies have found that LVD, particularly when quantitated according to its location (i.e., intratumoral or peritumoral), as well as LVI, correlate with both regional lymph node metastasis [including sentinel node (SN) status] and survival (Dadras et al., 2003; Huber et al., 2010; Kilvaer et al., 2010; Liu et al., 2008; Massi et al., 2006; Petersson et al., 2009).
Some studies have suggested that LVD may even be a more powerful prognostic factor than other currently well-known clinicopathologic prognostic factors (e.g., Breslow thickness) (Dadras et al., 2005; Schietroma et al., 2003) However, other studies found no association between LVD and outcome (Sahni et al., 2005; Wobser et al., 2006). This heterogeneity of results might be due to differences in study design such the lymphatic biomarkers utilized and the method used for determining LVD.
The aim of the present study was to critically review the evidence for correlation between intratumor and/or peritumoral LVD, LVI, or expression of lymphatic markers by melanoma tumor cells themselves and regional lymph node status [including sentinel lymph node (SN) status] and outcome in melanoma patients. We therefore conducted a systematic review of the retrospective and prospective studies reported in the literature that have assessed the expression of lymphangiogenic markers, detected with IHC, in primary cutaneous melanoma. As part of this review, we provide an interpretation of the available evidence and make recommendations for conducting future additional studies.
A literature search using the keywords described in the 'Methods' section yielded 84 original articles. Of these, 38 studies/articles were published between 1990 and 2012, and tested the hypothesis that lymphatic marker assessment in the primary tumor by IHC might predict SN status and/or outcome in patients with cutaneous melanoma. Sixteen studies were considered ineligible because they did not fulfill the inclusion criteria: three studies did not analyze the association between lymphatic marker expression in the primary melanoma and lymph node status and prognosis (De Waal et al., 1997; Goydos and Gorski, 2003; Schietroma et al., 2003), one study assessed lymphangiogenic markers only in SN specimens (Vitoux et al., 2009), five studies correlated lymphatic marker assessment and response to therapy (Kurschat et al., 2007; Lin et al., 2005; Molavi et al., 2008; Tsunoda et al., 2007; Vihinen et al., 2007), one study considered the utility of lymphatic biomarkers for distinguishing benign from malignant melanocytic skin tumors on the basis of the lymphovascular pattern (Giorgadze et al., 2004), two studies compared tumor specimens from different stages of melanoma (Pritchard-Jones et al., 2007; Redondo et al., 2003), one study analyzed lymphangiogenic markers as predictors of metastasis without referring to lymph node metastasis (Clarijs et al., 2002), two studies investigated an angiogenic marker (i.e., VEGF-A) only (Chua et al., 2009; Depasquale and Thompson, 2008), and one study did not use a lymphatic-specific biomarker (Fallowfield and Cook, 1990). The remaining 22 studies investigated the ability of lymphatic biomarkers to predict SN metastasis and disease recurrence or death and were further analyzed (Boone et al., 2008; Brychtova et al., 2008; Dadras et al., 2005, 2003; Depasquale and Thompson, 2008; Doeden et al., 2009; Emmett et al., 2010; Fohn et al., 2011; Gallego et al., 2011; Liu et al., 2008; Massi et al., 2006; Niakosari et al., 2008; Petersson et al., 2009; Rose et al., 2011; Sahni et al., 2005; Shayan et al., 2012; Storr et al., 2012; Saume et al., 2003; Wobser et al., 2006; Xu et al., 2012, 2008; Yun et al., 2011). A full list of the studies which met the inclusion criteria is presented in Tables 1 and 2. Lymphatic markers used to identify peritumoral and intratumoral LVD, the evaluation of LVI and lymphatic markers utilized for assessment of expression in tumor cells and detection of LVI are also listed in Table 1. In Table 2, associations between lymphatic marker expression and both regional lymph node status and patient survival are highlighted.
Table 1. Summary of the assessment of lymphatic markers utilized in the 22 studies included in this systematic review categorized according to the lymphatic features assessed and the method employed for its detection: (i) lymphovascular density (LVD, intratumor, and/or peritumor), (ii) lymphovascular invasion (LVI) and (iii) the expression of lymphatic markers measured in either tumor cells or lymphatic vessels
|Dadras et al. (2003)||LYVE-1||LYVE-1||H&E, VEGF-C||VEGF-C||Present or absent||LYVE-1||No. of vessels, vessel size, vessel area/tumor area|
|Straume et al. (2003)||D2-40, LYVE-1||D2-40, LYVE-1||H&E||NA||NA||D2-40, LYVE-1||No. of vessels|
|Wobser et al. (2006)||VEFGR-3, MART1||VEFGR-3||NA||NA||NA||VEGFR-3, MART1||Percentage|
|Dadras et al. (2005)||LYVE-1||LYVE-1|| |
|VEGF-C, -D||Present or absent||D2-40, LYVE-1, VEGF-C, -D||No. of vessels, vessel size, vessel area/tumor area|
|Sahni et al. (2005)||S-100, LYVE-1||S-100, LYVE-1||H&E, S-100||NA||NA||LYVE-1||Hot spot-based score|
|Massi et al. (2006)||D2-40||D2-40||NA||VEGF-C||Score based on % of positive cells||D2-40||No. of vessels, vessel size, vessel/field area|
|Brychtova et al. (2008)||VEGF-A, -C, VEGFRF1,2,3||VEGF-A, -C, VEGFRF1,2,3||NA|| |
|Liu et al. (2008)||NA||LYVE-1||VEGF-C, -D, VEGFR-3||VEGF-C, -D||>10% of cells positive||LYVE-1||No. of vessels|
|VEGFR-3||>5% of cells positive|
|Niakosari et al. (2008)||D2-40||D2-40||H&E||NA||NA||D2-40||Present or absent|
|Boone et al. (2008)||VEGF-C, MelanA||NA||NA||VEGF-C||VEGF-C as percentage of MelanA||VEGF-C||VEGF-C: MelanA|
|Xu et al. (2008)||D2-40||D2-40||H&E, S-100||NA||NA||D2-40||Present or absent|
|Petersson et al. (2009)||D2-40||D2-40||H&E||NA||NA||D2-40||No. of vessels|
|Doeden et al. (2009)||NA||D2-40, LYVE-1||H&E||NA||NA||D2-40, LYVE-1||Present or absent|
|Petitt et al. (2009)||NA||D2-40, S-100||H&E, S-100||NA||NA||D2-40||Present or absent|
|Emmett et al. (2010)||NA||LYVE-1||H&E||NA||NA||LYVE-1||Vessels area|
|Gallego et al. (2011)||D2-40||D2-40||NA||VEGF-C||Staining intensity||D2-40||No. of vessels|
|Yun et al. (2011)||D2-40||D2-40||H&E, S-100||NA||NA||D2-40||No. of vessels in 1 mm2 in hot spots within the area of regression|
|Rose et al. (2011)||D2-40, CD34||D2-40, CD34||H&E, S-100 (selected cases)||NA||NA||D2-40||Present or absent|
|Fohn et al. (2011)||D2-40||D2-40||H&E||NA||NA||D2-40||Present or absent|
|Xu et al. (2012)||D2-40||D2-40||S-100||NA||NA||NA||Present or absent|
|Shayan et al. (2012)||D2-40||D2-40||NA||NA||NA||D2-40||Degree of patency of lymphatic vessel lumina|
|Storr et al. (2012)||D2-40, CD34||D2-40, CD34||H&E||NA||NA||D2-40||Number of vessel/area (low/high)|
Table 2. Results of studies assessing associations of lymphatic vessel density (LVD) and lymphatic vessel invasion (LVI) (determined with the use of immunohistochemistry for lymphatic biomarker) with lymph node status and patient outcome in melanoma patients
|Dadras et al. (2003)||LVD|| |
|Straume et al. (2003)||LVD||D2-40; LYVE-1||202||Not significant||Significant|
|Wobser et al. (2006)||LVD||VEGFR-3||26||Not done||Not done|
|Dadras et al. (2005)||LVI, LVD|| |
|Sahni et al. (2005)||LVI, LVD||LYVE-1||36||Not significant||Not done|
|Massi et al. (2006)||LVD|| |
|Brychtova et al. (2008)||LVD|| |
|Liu et al. (2008)||LVD|| |
|Niakosari et al. (2008)||LVI||D2-40||96||Significant||Not done|
|Boone et al. (2008)||LVD||VEGF-C||113||Significant||Not significant|
|Xu et al. (2008)||LVI||D2-40||106||Not done||Significantb|
|Petersson et al. (2009)||LVI||D2-40||74||Significant||Significant|
|Doeden et al. (2009)||LVI, LVD||D2-40; LYVE-1||94||Not significant||Not significant|
|Petitt et al. (2009)||LVI||D2-40||27||Not significant||Not done|
|Emmett et al. (2010)||Schield's indexc||LYVE-1||18||Significant||Not done|
|Gallego et al. (2011)||LVD|| |
|Yun et al. (2011)||LVI, LVDe||D2-40||321||Not done||Significant|
|Rose et al. (2011)||LVI||D2-40||246||Not done||Significant|
|Fohn et al. (2011)||LVI||D2-40||64||Significant||Not done|
|Xu et al. (2012)||LVI||D2-40||251||Not done||Significant|
|Shayan et al. (2012)||LVD||D2-40||22||Significant||Not done|
|Storr et al. (2012)||LVI, LVD||D2-40, CD34||202||Not significant||Not significant|
Study methods and definitions
In total, 2251 patients were evaluated in the 22 studies.
Fourteen studies investigated intratumoral and peritumoral LVD (Brychtova et al., 2008; Dadras et al., 2005, 2003; Gallego et al., 2011; Massi et al., 2006; Niakosari et al., 2008; Petersson et al., 2009; Sahni et al., 2005; Shayan et al., 2012; Storr et al., 2012; Straume et al., 2003; Wobser et al., 2006; Xu et al., 2008; Yun et al., 2011); one (Boone et al., 2008) and four studies (Doeden et al., 2009; Emmett et al., 2010; Liu et al., 2008; Petitt et al., 2009) assessed intratumoral or peritumoral LVD only, respectively. LVI was assessed in 16 studies (Dadras et al., 2005, 2003; Doeden et al., 2009; Emmett et al., 2010; Fohn et al., 2011; Liu et al., 2008; Niakosari et al., 2008; Petersson et al., 2009; Petitt et al., 2009; Rose et al., 2011; Sahni et al., 2005; Storr et al., 2012; Straume et al., 2003; Xu et al., 2012, 2008).
With regard to the method for biomarker detection, IHC was used in all studies and was supplemented by polymerase chain reaction (Liu et al., 2008) and in situ hybridization (Dadras et al., 2003) methods in two studies.
The intensity of the lymphatic biomarker expression was evaluated in a limited number of articles. The intensity of VEGF-C and VEGF–D signals was scored using a semi-quantitative method in several papers (Boone et al., 2008; Brychtova et al., 2008; Dadras et al., 2005; Gallego et al., 2011; Massi et al., 2006).
Several definitions of the peritumoral area were applied. Dadras and Massi considered this area to be the tissue surrounding the tumor up to a distance of 100 µm (Dadras et al., 2005, 2003) and 500 µm (Massi et al., 2006; Xu et al., 2008), respectively. The peritumoral area was defined as the tissue within one microscopic field of the tumor border utilizing a ×10 (Doeden et al., 2009; Shayan et al., 2012) or a ×40 (Emmett et al., 2010; Sahni et al., 2005) microscope objective in other studies.
In two papers, VEGF-C expression in tumor cells and in tumor-associated macrophages (TAM) was evaluated (Boone et al., 2008; Gallego et al., 2011).
Quality assessment of reported studies
Table 3 describes eligible studies according to their adherence to standard methods for assessment of lymphatic markers expression, which were set by the first international consensus on the methodology of lymphangiogenesis quantification in solid human tumors (Van Der Auwera et al., 2006). None of the eligible studies, including those conducted after this consensus conference, fully satisfied the proposed criteria. To date, no study has been published utilizing double immunostaining with D2-40 and Ki-67 in primary cutaneous melanoma, and D2-40 stains were applied in only 14 of 22 studies. The remaining eight studies used other lymphatic vessels markers (Table 1), such as LYVE-1, whose expression could be influenced by other pathological conditions such as inflammation, thus providing less accurate detection of lymphatic vessels as compared with D2-40. The utilization of the proliferation marker Ki67, to indicate new lymphatic vessel growth, was used in only one study. Manual hot spot detection was not performed or not described in seven studies. Some studies performed computer-assisted lymphatic vessel counting. The Chalkley count, which is a more objective measure of LVD quantitation than manual hot spot counting, was performed only in two series.
Table 3. Summary of the quality of the 22 eligible studies assessed with regard to the first international consensus guidelines on the methodology of lymphatic vessels quantification in solid human tumors (Van Der Auwera et al., 2006)
|Dadras et al. (2003)||Not performed||Not performed (computer-assisted analysis)||Not performed||Not performed||Not performed|
|Straume et al. (2003)||Ki-67 only||Performed||Not performed||Performed||Not performed|
|Wobser et al. (2006)||Not performedc||Not described||Not performed||Not performed||Performed|
|Dadras et al. (2005)||Not performed||Not performed (computer-assisted analysis)||Not performed||Not performed||Performed|
|Sahni et al. (2005)||Not performedc||Performed||Performed||Not performed||Performed|
|Massi et al. (2006)||D2-40 only||Performed||Not performed||Not performed||Performed|
|Brychtova et al. (2008)||Not performed||Not performed||Not performed||Not performed||Not performed|
|Liu et al. (2008)||Not performed||Performed||Not performed||Not performed||Performed|
|Niakosari et al. (2008)||D2-40 onlyc||Not described||Not performed||Not performed||Performed|
|Boone et al. (2008)||Not performedc||Not performed||Not performed||Not performed||Performed|
|Xu et al. (2008)||D2-40 onlyc||Performed||Not performed||Not performed||Performed|
|Petersson et al. (2009)||D2-40 only||Not performed||Not performed||Not performed||Not performed|
|Doeden et al. (2009)||D2-40 only||Performed in peri-tumor tissue only||Not performed||Not performed||Performed|
|Petitt et al. (2009)||D2-40 onlyc||Performed in peri-tumor tissue only||Not performed||Not performed||Not performed|
|Emmett et al. (2010)||Not performed||Performed||Not performed||Not performed||Performed|
|Gallego et al. (2011)||D2-40 onlyc||Not performed (computer-assisted analysis)||Not performed||Not performed||Not reported|
|Yun et al. (2011)||D2-40 onlyc||Performed in area with regression||Not performed||Not performed||Performed|
|Rose et al. (2011)||D2-40, CD34||Performed||Not performed||Not performed||Not performed|
|Fohn et al. (2011)||D2-40 only||Not performed||Not performed||Not performed||Performed|
|Xu et al. (2012)||D2-40 onlyc||Performed||Not performed||Not performed||Performed|
|Shayan et al. (2012)||D2-40 onlyc||Performed||Not performed||Not performed||Performed|
|Storr et al. (2012)||D2-40 onlyc||Performed||Performed||Performed||Performed|
Lymphatic biomarkers and the prediction of lymph node metastasis
Overall, 17 studies detected a correlation between LVI and/or LVD and/or primary tumor lymphatic marker expression (i.e., VEGF-C and -D, and VEGFR3) and regional lymph node status. The presence of LVI in the primary tumor was correlated with lymph node positivity in five studies (Dadras et al., 2005; Emmett et al., 2010; Fohn et al., 2011; Niakosari et al., 2008; Petersson et al., 2009). In contrast, three papers reported no correlation between the presence of LVI and lymph node status (Doeden et al., 2009; Petitt et al., 2009; Sahni et al., 2005). The degree of LVD correlated with regional lymph node status in five studies (Boone et al., 2008; Dadras et al., 2005, 2003; Massi et al., 2006; Shayan et al., 2012), while no such association was reported in five other studies (Doeden et al., 2009; Gallego et al., 2011; Sahni et al., 2005; Storr et al., 2012; Straume et al., 2003).
Dadras et al. (2005, 2003) in two separate series which analyzed LYVE-1 and VEGF-C, respectively, reported that LVD [quantified as relative lymphatic vascular area (%), lymphatic vessels/mm2, and vessel size] was an accurate predictor of SN status. Moreover, patients developing metastasis also had larger vessel diameters. However, no differences in the density of tumor-associated blood vessels were found (Dadras et al., 2003). Lymphatic vascular area occupying >2% of the whole tumor area had a sensitivity of 100% in predicting SN metastasis (Dadras et al., 2005). Of note, the presence of LVI detected by H&E and VEGF-C immunostaining was not a predictor for SN metastasis. The same investigators did not find an independent correlation between LVI (detected utilizing the monoclonal antibodies LYVE-1 and D2-40) and SN positivity (Doeden et al., 2009). When IHC-detected LVI was combined with intratumoral location of the lymphatic vessels, the greatest diagnostic capability for predicting SN metastasis was observed (PPV and NPV of 80 and 72%, respectively). In another study, which investigated the diagnostic value of primary tumor LVI detected with D2-40 for predicting SN metastasis in patients with melanoma <2.0 mm thick (Fohn et al., 2011), LVI had a PPV and a NPV of 86 and 88%, respectively, indicating that tumors which did not show LVI were less likely to harbor metastasis in the SN.
Other studies have found that LVD predicts SN status more accurately than the presence of LVI. Emmett et al. evaluated LVD and LVI with the following measures: the LVD per square millimeter and the Shield's index. The latter is calculated from the following features: the absence or presence of LVI, the LVD (number of vessels/square millimeter), and tumor thickness (Emmett et al., 2010b). They also found that the LVD, described with the Shield's index, was a better predictor of SN status than the presence of LVI. The number and area of intratumor lymphatic vessels was also found to be significantly higher in patients with SN metastasis in another study of 45 melanoma patients (15 cases and 30 controls) in which LVD was assessed by D2-40 (Massi et al., 2006). However, three other studies failed to identify any significant correlation between LVD [assessed by either counting the absolute number of lymphatic vessels using the Chalkey method (Sahni et al., 2005; Storr et al., 2012) or the number of vessels per square millimeter (Storr et al., 2012; Straume et al., 2003)] and lymph node status. The latter studies had limitations that may have influenced their results: small sample size (Sahni et al., 2005), lack of a standardized method of lymph node staging (Straume et al., 2003), and failure to consider intratumoral and peritumoral LVD separately (Storr et al., 2012).
Regarding this latter issue, whether LVD occurred in the peritumoral or the intratumoral area may indicate different roles in tumor progression. Shayan et al. (2012) reported a greater LVD in the peritumoral area of the primary melanoma in patients with lymph node metastasis than in those with negative lymph nodes. In contrast, intratumoral LVD was higher in non-metastasizing melanomas than in melanomas with lymph node involvement. These authors proposed that assessment of the intratumoral and peritumoral LVD may be a useful marker for primary melanomas at higher risk of metastasizing to lymph nodes.
Detection of lymphatic biomarkers in primary cutaneous melanoma by polymerase chain reaction showed results consistent with those reported by IHC evaluation (Liu et al., 2008). Because IHC allows visual identification of the cells expressing lymphatic markers, IHC is the preferred method for lymphatic marker assessment.
The results of studies assessing the utility of the expression of lymphatic markers VEGF-C and VEGF-D and their receptor VEGFR-3 in predicting SN status are inconsistent. A study of 130 melanoma patients assessed expression of VEGF-A and -C as well as the three receptors VEGFR1, 2, and 3, and did not find a relationship between the expression of any of the markers and SN status (positive SNB rate 33%) (Brychtova et al., 2008). Dadras et al. (2005) identified VEGF-C but not VEGF-D as a significant predictor of lymph node metastasis. In another study, IHC and PCR analysis for both VEGF-C and VEGF-D, as well as their receptor VEGFR-3, significantly predicted SN status (Liu et al., 2008). Boone et al., (2008; Schietroma et al., 2003) also reported a significant association between lymph node metastasis and VEGF-C expression in primary melanomas. Interestingly, Gallego et al. (2011) reported that peritumoral rather than intratumoral expression of VEGF-C is associated lymph node metastasis.
Lymphatic biomarkers and prognosis
The prognostic impact of lymphatic marker expression in primary melanomas has been analyzed in 10 studies (Boone et al., 2008; Dadras et al., 2003; Doeden et al., 2009; Elwood et al., 2009; Liu et al., 2008; Massi et al., 2006; Petersson et al., 2009; Straume et al., 2003; Xu et al., 2008). IHC-detected LVI was a predictor of poor prognosis in five studies (Petersson et al., 2009; Rose et al., 2011; Xu et al., 2012, 2008; Yun et al., 2011), while it was not associated with patient outcome in two reports (Doeden et al., 2009; Storr et al., 2012). Four papers showed a correlation between LVD and prognosis (Dadras et al., 2003; Massi et al., 2006; Straume et al., 2003; Yun et al., 2011), and four did not report any association (Boone et al., 2008; Doeden et al., 2009; Liu et al., 2008; Storr et al., 2012).
In a prospective study of 246 melanoma patients with a median follow-up of 6 yr, LVI detected with the lymphatic endothelial marker D2-40 and the panvascular marker CD34 was associated with worse disease-free survival and overall survival, independently from AJCC tumor stage (Rose et al., 2011). This prognostic value of LVI may lead to improved patient stratification in the staging system, as suggested by Xu et al. (2012), who found that the presence of LVI identified patients at higher risk of disease progression among AJCC stages IB (T1b-T2aN0) and IIA (T2b-T3aN0).
The occurrence of LVI within the tumor and in the peritumoral area may reflect different roles in melanoma progression. In a study of 60 patients in which LVI was determined with D2-40, Petersson et al. reported that intratumoral LVI was an independent predictor of worse survival along with Breslow thickness and the presence of ulceration. In contrast, peritumoral LVI by melanoma cells was not associated with patient outcome (Petersson et al., 2009).
Results regarding correlation between LVD and prognosis are much more controversial. In the pioneering study conducted by Dadras et al. (2003) in which they analyzed 37 melanoma patients with a median follow-up of 6.5 yr, peritumoral LVD was a predictor of time to lymph node metastasis and survival, and intratumoral LVD showed independent prognostic value for survival only. Massi et al. (2006) reported a worse prognosis for patients having tumors with greater peritumoral LVD, while intratumoral LVD approached significance. The presence of a higher risk of death due to melanoma for patients with higher intratumoral and peritumoral LVD in the primary tumor was also reported by Xu et al. (2008) in a study of 106 patients with follow-up of at least 10 yr in all patients.
Despite these promising results, other studies did not reported any association between LVD and survival or even an inverse correlation. In a study involving 94 patients, peritumoral LVD did not predict disease-free or overall survival (Doeden et al., 2009). In one large case series, Straume et al. (2003) found a better survival in patients with higher grades of intratumoral and peritumoral LVD in their primary melanoma.
The association between the expression of other lymphatic markers and prognosis is also controversial. In a study conducted on 46 melanoma patients, Liu et al. (2008) analyzed the expression of VEGF-C, VEGF-D, and their receptor VEGFR-3 and found that they were independent predictors of survival; traditional staging parameters (Breslow thickness and ulceration) were not retained in the final prognostic model. In the first study of Dadras et al. (2003), VEGF-C, as detected by in situ hybridization, was expressed in a minority of tumors and was not a prognostic factor for survival, regardless of whether it was assessed in melanoma cell cytoplasm, epidermal keratinocytes overlying the tumor or in peritumoral stromal cells. Boone et al. (2008) found that the expression of VEGF-C in melanoma cells as well as in tissue-associated macrophages was a significant prognostic factor in 113 melanoma patients, although when it was considered with other clinicopathologic variables, it was not an independent predictor of survival on multivariate analysis.
The differing results of the studies evaluating the association between lymphatic marker expression and prognosis may be explained by the interactions occurring between lymphatic vessels and the patient's immune system (Ji, 2006, 2012; Li et al., 2012). Although analysis of the influence of the immune system on the expression of lymphatic markers and vice versa is beyond the scope of this review, recent studies have highlighted the association between lymphatic vessels, tumor regression, and macrophages (Da Costa et al., 2011; Massi et al., 2009, 2007; Storr et al., 2012; Yun et al., 2011). A study of 321 patients with a median follow-up of 20 yr suggested that LVD and, to a larger degree, LVI have a significant prognostic impact especially when measured/present within areas of complete histopathologic regression (Yun et al., 2011). Such associations are consistent with the observation that the degree of intratumoral and peritumoral LVD, as well as the presence of LVI, is associated with the presence of TAM in the area surrounding primary melanomas (Storr et al., 2012). TAMs secrete VEGF factors as well as NO, and this may underlie the association between lymphatic marker expression and tumor regression (Boone et al., 2008; Massi et al., 2009, 2007; Ran and Montgomery, 2012). Moreover, patients with more advanced primary melanoma (i.e., those with a higher tumor thickness and mitotic count as well as the presence of ulceration) tend to have greater LVD and presence of LVI (Storr et al., 2012) but tumor-infiltrating lymphocytes (TIL) are less likely to be detected (Azimi et al., 2012).
The aim of this review was to systematically collect and analyze the available evidence on LVD, LVI, and primary melanoma lymphatic marker expression as predictors of SN status and patient prognosis. As detailed above and summarized in Tables 1 and 2, the 22 published studies that investigated this relationship suggest that a probable correlation exists, although some of the results are conflicting. This is likely to reflect heterogeneous study design, including differences in the lymphatic biomarkers assessed, and methods utilized to detect and quantify biomarker expression and from statistical analysis.
Methodologic considerations affecting the results of reported studies
The disagreements between and within studies may reflect differences in methodology and deficiencies in the quality of some studies.
SNB was not performed in all the studies. This is important because SNB improves the accuracy of melanoma metastasis detection (Doubrovsky et al., 2004). Therefore, the presence of lymph node metastasis may have been underestimated in studies not utilizing SNB, and this could have contributed to the lack of a statistical association between the LVD (intratumoral and peritumoral) and lymph node status.
The overall quality of the studies was generally poor if benchmarked against the REMARK guidelines (McShane et al., 2005), and the guidance provided by the first international consensus on the methodology of lymphangiogenesis quantification in solid human tumors (Van Der Auwera et al., 2006). In particular, a minority of the studies that were considered stated explicitly that they conformed to the REMARK checklist (Emmett et al., 2010), and no study was conducted according to the standard method proposed by the 2006 consensus conference. The design of the studies was generally based on retrospective series, and limitations due to patient selection criteria cannot be ruled out (Hoek, 2007). Dual immunostaining with D2-40 and Ki-67, which will allow accurate detection of proliferating lymphatic vessel cells (and therefore infer lymphangiogenesis), as recommended by the consensus on lymphangiogenesis quantification, has not been performed in any of the studies published to date. Lymphatic biomarker expression was measured using a variety of different indicators, as described in Table 1. Several descriptors, such as mean, median, percentage, and odds ratio were utilized to measure the outcomes. Multivariate analysis was performed in only a minority of the studies, thus ruling out the possibility of fully assessing the relationship between lymphatic biomarkers and other clinicopathologic predictors of outcome. Taken together, these observations suggest that the conduct of new studies designed according to existing guidelines is warranted in an effort to standardize the methods to yield data that can be appropriately compared between studies.
Recommendations for the conduct of further studies
From a methodologic viewpoint, future studies assessing LVD, LVI, and lymphangiogenesis as biomarkers in melanoma should be planned prospectively with an adequate number of patients to provide adequate statistical power. Marker selection for detecting LVD and LVI at IHC should be performed according to the methodology described by the consensus conference on lymphangiogenesis in solid tumors (Van Der Auwera et al., 2006), which suggested, based on the evidence then available, that D2-40 is the lymphatic vessel antibody with the highest accuracy. Moreover, investigating dual immunostaining for Ki-67 together with D2-40 will provide investigators with important data on lymphatic endothelial cell proliferation and thus on lymphangiogenesis. Studies utilizing double immunostaining with D2-40 and melanocytic markers, such as S-100 and MelanA, would improve the accuracy of the detection of LVI and be likely to provide important data on the significance of LVI in intratumoral and peritumoral lymphatic vessels. VEGFR3 and its ligands VEGF-C and -D have important limitations for investigating lymphatic vessels, as VEGR3 is expressed in lymphatic vessels as well as in fenestrated capillary of blood vessels. The ‘hot spot’ method should be utilized for quantitating LVD in tumor sections (Weidner et al., 1991). To limit potential bias, the Chalkley point graticule counting method (Fox et al., 1995), which offers a more objective measurement of LVD, should be performed routinely. While computer-assisted image analysis of LVD has the potential to more accurately and reproducibly assess LVD, there is at present no clear evidence of its reliability and accuracy compared with standard manual counting methods. It has also been recommended that two investigators separately assess IHC performed on primary melanoma specimens and that high inter-observer agreement should be achieved.
Future studies should assess both LVI and LVD detected in the intratumoral and peritumoral areas and investigate their association with SN status and patient survival. Moreover, they should account for influences of the immune system, by investigating the presence of TAM and tumor-infiltrating lymphocytes in the area surrounding the primary tumor.
Future prospects for lymphatic biomarkers influencing surgical and medical treatment of melanoma
In the future, routine assessment of lymphatic markers in primary cutaneous melanomas may provide more accurate risk stratification than is currently possible. This may improve patient selection for surgical and medical therapies in clinical trials. To date, predictive models for SN status based on the routinely assessed clinicopathologic features of the primary melanoma have been developed and validated in large datasets from independent institutions (Faries et al., 2010; Mocellin et al., 2009; Wong et al., 2005). For example, the prognostic nomogram developed at the Memorial Sloan–Kettering Cancer Center that considers simultaneously patient age, tumor site, primary tumor thickness, ulceration, and level of invasion obtained an accuracy of approximately 70% (Wong et al., 2005). In a recent report, predictive models obtained a negative predictive value of >90% by incorporating eight variables (age, sex, tumor thickness and ulceration, level of invasion, regression, melanoma subtype, and mitotic rate) (Mocellin et al., 2009). Biomarkers, such as those used to quantify lymphangiogenesis, could further improve the accuracy of these predictive models (Grimm et al., 2009) and provide a greater negative predictive value in defining the risk of SN metastasis (Pasquali et al., 2011). In other words, a predictive test characterized by a low percentage of false-negative cases (such as a test with high accuracy and high negative predictive value) might be used in clinical practice to safely avoid SNB in patients identified as being at very low risk of harboring SN metastases, thus minimizing the probability that a patient with melanoma cells in the SN would not receive a SNB.
Several anti-lymphangiogenic targeted therapies have been tested in pre-clinical models (Facchetti et al., 2007). In one experimental model, targeting NO synthetase resulted in decreased NO and a blockade of tumor lymphangiogenesis. The NO synthetasis inhibitor L-NMMA resulted in a decrease in VEGF-C and -D expression in the primary tumor and a consequent blockade of lymphatic vessel generation (Lahdenranta et al., 2009).
The VEGFR tyrosine kinase inhibitor, PTK/ZK, which targets the three VEGFRs, has been shown to reduce the incidence of lymph node metastasis and enhances the sensitivity to platinum-based chemotherapy in a metastatic melanoma model (Sini et al., 2008). Similarly, in a preclinical melanoma model based on cell lines derived from metastatic melanoma lymph nodes, inhibition of VEGFR-3 with a recombinant adeno-associated viral vector was shown to block primary tumor lymphangiogenesis and the development of lymph node metastases (Lin et al., 2005).
Similar agents have now been utilized in early phase clinical trials (Cook et al., 2010; De Jonge et al., 2011; Del Vecchio et al., 2010; Flaherty, 2007; Hersey et al., 2009; Krupitskaya and Wakelee, 2009).
This systematic review analyzed the currently available evidence assessing the role of IHC-detected LVI, LVD (intratumoral and peritumoral), and expression of lymphatic markers in tumor cells and lymphatics, as predictive factors for regional lymph nodes metastasis (including SN metastasis) and survival.
Although there has been a lack of uniformity in the methods utilized for assessing and quantifying the expression of lymphatic markers, the currently available evidence suggests that IHC-detected LVI is a predictor of SN metastasis and poorer survival. While the significance of LVD is still controversial, the presence of greater density of lymphatic vessels in the peritumoral area appears to be associated with melanoma spread to regional lymph node and distant sites.
Given these considerations, lymphatic biomarkers appear to hold great potential for stratifying patients according to their risk of harboring lymphatic metastases and of developing melanoma progression, but much further research within properly designed studies is necessary before it would be appropriate to recommend their assessment in routine pathologic diagnosis and reporting in a clinical setting.
A systematic review of articles published up to May 2012 that analyzed the role of lymphatic biomarkers in primary cutaneous melanoma in predicting regional lymph node status and prognosis was performed by searching MEDLINE, EMBASE, and the Cochrane database. The search strategy included the following keywords variously combined: ‘melanoma’, ‘lymph node’, ‘lymphangiogenesis’, ‘SN’, ‘survival’, ‘prognosis’, and ‘follow-up’. We sought not only original publications but also review articles because we considered the latter to represent an additional quality control measure to source original works that could have been overlooked. When appropriate, cited references from original and review articles were also used as an additional source for identification of relevant studies.
Other inclusion criteria were as follows: studies performed on human melanoma tissue, primary cutaneous melanoma lymphangiogenesis assessed by IHC, and assessment of lymph node status and patient prognosis. The following data were extracted from the published papers: lymphangiogenic biomarker assessed, sample size, methods utilized for biomarker detection, LVI assessed/not assessed, LVD (intratumoral and peritumoral) assessed/not assessed, method utilized for assessment of regional lymph node status (SNB performed/not performed), outcome measures, and the results of factors predictive of regional lymph node status and prognosis. Finally, the quality of each eligible study was assessed for adherence to Reporting Recommendations for Tumor Marker prognostic Studies (REMARK) criteria (McShane et al., 2005) and to the first international consensus on the methodology of lymphangiogenesis quantification in solid human tumors (Van Der Auwera et al., 2006). According to the latter guidelines, the following methodologic steps are suggested: (i) double immunostaining with the D2-40 and ki-67 monoclonal antibodies should be performed to assess the presence of lymphatic vessels and proliferating lymphatic endothelial cells (as a marker of ongoing lymphangiogenesis); (ii) manual hot spot selection at low magnification (e.g., ×10) in intratumoral and peritumoral tissue; (iii) Chalkley point graticule; (iv) counting the number of proliferating and non-proliferating lymphatic endothelial cells; and (v) independent assessment by two pathologists.
In concordance with the PRISMA guidelines for systematic reviews (Liberati et al., 2009), data were extracted by two investigators (S.P and A.VDP.) to ensure homogeneity of data collection and to rule out the effect of subjectivity in data gathering and entry. Disagreements were resolved by iteration, discussion, and consensus. To assess potential systematic biases, a third investigator (S.M.) conducted a concordance study by independently reviewing the eligible studies: complete concordance was attained.