Lymphatic Mapping and Sentinel Lymphadenectomy for Early-Stage Melanoma

Abstract

Objective:

Lymphatic mapping and sentinel lymphadenectomy (LM/SL) have been applied to virtually all solid neoplasms since our original description of LM/SL for melanoma. Our objectives were to determine the diagnostic and therapeutic utility of LM/SL, investigate carbon dye for mapping the microanatomy of lymphatic flow within the sentinel node (SN), and determine the prognostic accuracy of molecular assessment of the SN.

Methods:

Since 1985, 1599 patients with AJCC Stage I/II melanoma have been treated by LM/SL at our institution and 4590 have been treated by wide excision (WE) without nodal staging. We examined the incidence of clinical nodal recurrence after WE alone, the incidence of subclinical nodal metastases found by LM/SL, and the incidence of nodal recurrence in basins with histopathology-negative SNs.

Results:

In 1514 LM/SL patients with a primary of known Breslow thickness, the incidence of metastasis in nodes claimed to be sentinel was 7.3%, 19.7%, 33.2%, and 39.7% for primary lesions ≤1.0, 1.01–2.0, 2.01–4.0, and >4.0 mm, respectively. In 3652 WE-only patients, the corresponding rates of nodal recurrence were 12.0%, 32.0%, 34.4%, and 30.1%. Thus, LM/SL detected only 60% of expected nodal metastases from primary melanomas <2.01 mm. Forty of 1599 (3.1%) patients developed recurrence in basins with immunohistochemistry (IH)-negative SNs. To determine whether nonrandom intranodal distribution of tumor cells could explain missed SN metastases, we coinjected carbon particles and blue dye during LM/SL in 166 patients: 25 (16%) patients had nodal metastases, all of which were found only in nodal subsectors containing carbon particles. When paraffin-embedded SNs from a subset of 162 IH-negative patients were re-examined by quantitative multimarker reverse-transcriptase polymerase chain reaction (qRT) assay, 49 (30%) gave positive signals. These patients had a significantly higher risk of disease recurrence and death than did patients whose IH and qRT results were negative (p < 0.0001). Comparison of 287 prognostically matched pairs of patients who underwent immediate (after LM/SL) versus delayed (after observation) dissection of nodal metastases revealed 5-, 10-, and 15-year survival rates of 73%, 69%, and 69% versus 51%, 37%, and 32%, respectively (P ≤ 0.001).

Conclusions:

SN assessment based on intranodal compartmentalization of lymphatic flow (carbon dye mapping) should increase the accuracy of IH and, in combination with multimarker qRT assessment, will allow confident identification of most patients for whom surgery alone is curative. Our data suggest a significant therapeutic benefit for immediate dissection based on identification of a tumor-involved SN.

Sentinel node assessment based on the intranodal comparmentalization of lymphatic flow and on multimarker molecular as well as immunohistochemical techniques increases the accuracy of lymphatic mapping and allows confident identification of patients for whom surgery is likely to be curative without adjuvant therapy.

Elective lymph node dissection (ELND) of clinically normal regional lymph nodes in patients with cutaneous melanoma assumes that the primary tumor will colonize these nodes before spreading to distant sites. However, only 20% of patients with nonpalpable nodes actually have nodal metastases by immunohistopathology.^1–3 This suggests that only 1 in every 5 patients with early-stage melanoma may benefit from ELND. Results of the Intergroup Melanoma Trial have confirmed that ELND is beneficial only in certain subgroups of patients with clinical stage I melanoma,^4,5 and no prospective randomized trial has yet demonstrated a conclusive overall benefit for early removal of regional nodes in all patients with early stage melanoma.

We believe that the failure of these trials to resolve the century-old controversy (first described by Snow⁶) of elective (early) versus therapeutic (delayed) lymph node dissection (LND) reflects a problem in trial design. The question is not whether LND is beneficial to all patients but whether a particular patient is likely to have subclinical nodal metastases “incubating” in the regional nodes without systemic metastases. Our minimally invasive technique of intraoperative lymphatic mapping and sentinel lymphadenectomy (LM/SL) was developed to identify nodal involvement without performing complete nodal dissection.² As we recently validated in a phase III international trial, LM/SL with a vital blue dye and radiopharmaceutical identifies and removes the first draining node in the lymphatic pathway from the primary tumor site.³ Focused histologic evaluation based on immunohistochemical staining (IH) of multiple sections of this sentinel node (SN) is more likely to reveal nodal metastasis than is routine hematoxylin and eosin staining (H&E) of limited sections from 12–40 bivalved nodes in a conventional LND specimen.^7,8

Studies of LM/SL in melanoma,^3,9,10 breast cancer,¹¹ colon cancer,¹² and virtually all solid neoplasms¹³ that spread to lymph nodes have confirmed the SN concept of an orderly progression of metastatic cells from the primary site through the lymphatics to 1 or 2 regional SNs. However, if metastasis to systemic sites is always preceded by sequential nodal metastasis, these studies do not explain regional and distant recurrence of tumor when SNs are apparently tumor-free. We have considered 2 possible paths of metastasis from a primary melanoma (Fig. 1). According to the “incubator” hypothesis, the primary tumor targets SNs in the regional lymph basin, where the metastatic tumor cells may survive and slowly grow but remain latent before spreading to distant sites. The incubator hypothesis is compatible with a therapeutic role for LM/SL; early removal of SN metastases, before distant metastases have occurred, should abort systemic spread and increase survival. According to the “marker” hypothesis, however, a primary melanoma metastasizes simultaneously via lymphatic and hematogenous routes, so that the presence of SN metastases becomes a marker of the likelihood of systemic disease but not the only potential source for distant metastases. Although the marker hypothesis is also compatible with the SN concept, it suggests that early removal of tumor-positive SNs would not alter long-term survival because distant metastases have already occurred and are the primary determinant of death due to melanoma.

FIGURE 1. Two hypotheses regarding metastatic routes in melanoma. According to the incubator hypothesis (left), primary melanoma initially metastasizes via the lymphatics to the SN, which is immunosuppressed by factors released from the primary melanoma. Metastatic foci in the SN may grow but remain latent (incubate) before spreading to distant sites. Thus, finding tumor cells in the SN indicates that the primary melanoma has the ability to metastasize; removal of the tumor-involved SN before there is further spread should prevent distant metastasis. According to the marker hypothesis (right), a primary melanoma metastasizes simultaneously via lymphatic and hematogenous routes. Thus, finding tumor cells in the SN is merely a marker of a primary melanoma that can metastasize; removal of the tumor-involved SN is unlikely to influence the growth of distant metastases and would have no therapeutic effect. The absence of melanoma cells in the SN indicates a primary melanoma that is unlikely to spread to distant sites; this has important prognostic implications.

Many studies with relatively short median follow-up (<5 years) have reported same-basin recurrences and distant metastases after excision of histologically tumor-free SNs. How can we reconcile nodal and distant recurrence with the SN concept and the incubator hypothesis? The answer may lie in the techniques used to sample and analyze the SN. Our studies mapping lymph flow in the SN have revealed that these nodes receive afferent drainage in a compartmentalized fashion, with a particular area of the skin draining not only to a specific node but also to a specific area of that node. Thus micrometastases may be missed if the wrong part of a node is examined. Such “geographic misses” theoretically could be minimized by serial sectioning, but a 6-mm node would yield 1500 four-micron sections. Routine evaluation of more than 10 to 20 of these sections (less than 1.5% of the entire node) is not practical. Of course, the mapping agent will also be localized to a specific portion of the SN, but unfortunately neither the blue dye nor radiopharmaceutical agents are permanent markers; both dissipate or decay before the specimen reaches the pathologist. We have therefore introduced carbon dye mapping of the SN to label a node as sentinel and to identify for the pathologist the intranodal site of lymphatic drainage, which is the most likely target of tumor cells.¹⁴

Micrometastases will also be missed if IH fails to detect small numbers of tumor cells. In a study of snap-frozen sections of SNs from 72 patients with early-stage melanoma,¹⁵ our group demonstrated the prognostic significance of multimarker reverse-transcriptase (RT) polymerase chain reaction (PCR) to identify tumor markers in frozen sections. Recurrence was highly correlated (P = 0.005) with multimarker messenger RNA (mRNA) expression in frozen SN sections that were histopathologically negative for tumor cells even after intense restudy by IH of serial sections. This study showed that histopathology may underestimate the true number of micrometastases. However, although tumor mRNA markers in the SN may more accurately reflect micrometastases than histopathology alone, it is difficult to avoid RNA degradation and deal with the logistics of preparing frozen nodal specimens for routine clinical practice at different institutes worldwide. We therefore developed techniques to extract RNA from paraffin-embedded (PE) tissue and look for expression of multiple mRNA markers.

This article will address 5 questions concerning the surgical management of patients with early-stage melanoma. First, what is the diagnostic accuracy of LM/SL for detection of SN metastases? Second, does the use of carbon dye as a permanent mapping agent help eliminate geographic misses of tumor cells within the SN? Third, can molecular analysis based on PE sections of SNs improve the prognostic accuracy of SN assessment? Fourth, does LM/SL confer a long-term survival advantage over wide excision (WE) alone in patients with clinically normal regional nodes? Finally, can we eliminate the marker hypothesis in favor of the incubator hypothesis?

PATIENTS AND METHODS

John Wayne Cancer Institute (JWCI) Database

Patients who had undergone WE with or without LM/SL for clinically localized melanoma were identified by review of the JWCI database. This computerized database contains the records of all 11,000 patients seen by our staff since April 1971. Follow-up is complete to within 30 months of last follow-up or death for 94% of patients in the database. The database thus represents a 30-year prospective audit of the results of therapy by JWCI staff. Between 1971 and 2001, 6189 patients with AJCC stage I/II primary cutaneous melanoma were seen by JWCI staff (Fig. 2). Of these, 1599 patients underwent LM/SL followed by complete LND (CLND) if the SN was positive by H&E or IH. The remaining 4590 patients, who were seen between April 1, 1971, and December 31, 2001, received WE only; they were managed by nodal observation and underwent delayed therapeutic CLND only if the nodes became palpably enlarged. Complete data on Breslow thickness of the primary was available for 3652 of the 4590 patients.

FIGURE 2. Recurrence patterns of patients with AJCC stage I/II melanoma managed by WE + postoperative observation versus WE + LM/SL + selective CLND for tumor-involved SNs.

All patients underwent treatment by our group at the UCLA Division of Surgical Oncology (1971–1991) or at JWCI (1991–2001), or they were referred to us for postoperative treatment/follow-up during the same period. Only patients seen within 4 months of WE and still free of nodal or distant metastases were included in the study. Data collection was complete through December 31, 2002, providing a minimum of 1 year of follow-up. The postoperative absence of disease was confirmed by physical examination, laboratory tests, and chest x-ray. Patients with histopathologic or clinical evidence of nodal metastases underwent chest tomograms and brain and/or liver nuclear scans (prior to 1984), or they underwent computed tomography (CT) of the chest, abdomen and pelvis, and CT or magnetic resonance imaging (MRI) of the brain (1984 to present).

As a rule, ELND and LM/SL were offered as options to all patients with intermediate or thick melanomas; patients were told that the therapeutic effectiveness of these procedures was unproven. The decision to undergo ELND or LM/SL for staging purposes was the individual patient’s choice. Patients who underwent WE only were those who declined ELND or LM/SL. ELND was the only option available until we initiated LM/SL during 1985. Until 1991, LM/SL was routinely followed by immediate CLND; after that time, LM/SL was followed by CLND only in patients with tumor-involved SNs.

Technique of LM/SL

The mapping technique used at JWCI is based on our previously described technique.^2,3 Preoperative cutaneous lymphoscintigraphy is required in all cases.³ For those patients in whom carbon dye was also used as an additional mapping agent, the technique followed that described in our previous report.¹⁴

Histopathologic Examination of the SN

SN specimens were reviewed as PE sections according to previously described procedures using IH with S-100 and HMB-45.^3,8 Although frozen-section analysis was routinely used during our development of LM/SL, we have moved to permanent sections to minimize loss of diagnostic material during frozen-plus-paraffin tissue sectioning. Interpretation of permanent material is always more accurate, and this policy allows the preparation and evaluation of optimal IH sections.

Molecular Assessment of SNs

In a retrospective study, 215 AJCC stage I/II melanoma patients who had undergone LM/SL at JWCI prior to 1994 were randomly selected based on availability of paraffin blocks. All patients in this subgroup had at least 8 years of follow-up. Fifty-three (25%) patients had tumor-involved SNs by H&E and/or IHC, whereas 162 patients had IH-negative SNs. Sections of 308 PE SNs from the 215 patients were assessed by quantitative multimarker RT PCR qRT assay for mRNA of four melanoma-associated genes involved in melanogenesis: MART-1, tyrosinase, microphthalmia-associated transcription factor (MITF), and tyrosinase-related protein 2 (TRP-2). These four markers are highly expressed in melanomas.15A,15B The methodology is detailed in another publication¹⁶; in brief, mRNA marker levels were quantitated using a standard curve of respective marker cDNA in each assay. All assays were performed at least twice, and each had positive and negative controls for RNA extraction, RT and PCR analysis. All specimens had intact mRNA as assessed by GAPDH house-keeping gene.

Statistics

Estimated melanoma-specific survival rates were obtained by the nonparametric Kaplan-Meier method. The log-rank test was used to determine survival differences according to different levels of risk factors and immediate versus delayed therapeutic CLND. Multivariate analysis was performed by the Cox proportional hazards regression model. The relative importance of prognostic factors was based on the Wald test of the coefficient associated with the prognostic factors in the Cox model. A computerized matching of patients undergoing WE+LM/SL versus WE alone was conducted by a program that matched at a 1:1 ratio based upon covariates selected for matching. A P value less than 0.05 was considered significant. All statistical analyses were two-tailed and performed using SAS software (SAS Institute, Cary, NC). Overall survival was defined as the time a patient remained alive after WE. Survival times were considered censored for patients who were alive at the last follow up or who died without evidence of melanoma.

RESULTS

Table 1 and Figure 2 show the characteristics of the 1599 patients who underwent WE plus LM/SL. The median age of this group was 51 years (mean, 52 years; range, 5–91 years). The median Breslow thickness was 1.43 mm (mean, 2.02 mm; SD, 1.77). SN metastases were found in 322 (20.1%) of all 1599 patients.

TABLE 1. Characteristics of 1599 Patients Undergoing Lymphatic Mapping and Sentinel Lymphadenectomy

Accuracy of LM/SL for Detection of SN Metastases

Of 1514 patients who underwent LM/SL for primaries of known Breslow thickness, 313 (20.7%) had nodal metastases; the incidence of nodal metastases was 7.3%, 19.7%, 33.2%, and 39.7% for primary lesions ≤1.0, 1.01–2.0, 2.01–4.0, and >4.0 mm, respectively. In a comparison group of 3652 patients who underwent WE alone for primaries of known Breslow thickness, the corresponding rates of nodal recurrence were 12.0%, 32.0%, 34.4%, and 30.1%, respectively.

If the incidence of palpable nodal recurrence after WE is the true incidence of nodal metastases that should be identified during LM/SL, then the data in Table 2 indicate that LM/SL understages a significant proportion of thin (≤2.0 mm) primary melanomas. LM/SL with IH detected only 60% of expected nodal metastases from primary melanomas <2.01 mm (7.3%/12.0% for ≤1.0 mm and 19.7%/32.0% for lesions 1.01–2.0 mm) but appeared to accurately detect metastases from thicker lesions. This is probably not caused by failure to identify the true SN because the incidence of nodal metastases detected by LM/SL was higher for lesions ≥ 4.0 mm and similar for lesions of 2.01–4.0 mm (Table 2). Instead, because the size and frequency of micrometastases in the SN increase with Breslow thickness of the primary,¹⁷ understaging more likely reflects a failure to detect small micrometastases from lesions ≤ 2.0 mm. Small metastases are more likely to be missed by IH, and a nonrandom distribution of tumor cells in the SN would further increase the risk of “geographic misses” (sampling error) when metastases are small.

TABLE 2. Incidence of Positive Nodes Found at LM/SL Compared with the Incidence of Clinically Evident Nodal Metastases after WE Alone

It is not surprising that our data indicate a direct relationship between nodal involvement and Breslow thickness, and an inverse relationship between time to development of clinical nodal metastases and Breslow thickness (Table 2). Note that the mean, median, and maximum time to development of clinical nodal recurrence after WE alone was inversely related to the thickness of the primary melanoma, reaching up to 22 years for thin primaries (< 1.0 mm). This corroborates our previous report on the frequency of subclinical tumor foci that remain latent in the regional nodes for many years after removal of thin melanomas, before becoming clinically palpable metastases.¹⁸ In addition, when patients were grouped according to the Breslow thickness, the mean number of tumor-involved nodes in the therapeutic CLND specimen was higher for patients with clinically palpable metastases found after WE and observation, than for those with SN metastases found by LM/SL, except when Breslow thickness was 2.01–4.0 mm (Table 2). This suggests metastatic spread along the lymph node chain within the drainage basin during the interval between excision of the primary melanoma and delayed therapeutic lymphadenectomy for palpable lymph nodes.

Forty of 1277 patients (3.1%) developed recurrence in a nodal basin with IH-negative SNs (Fig. 2); 7 of these were preceded by in-transit metastases that could have been the secondary source of the nodal metastases, but 33 (2.6%) were clearly missed, probably due to failure of the surgeon to identify the true SN or failure of the pathologist to identify micrometastases in this node. Thus, if primary tumor thickness is used as a surrogate for expected incidence of nodal metastases, then LM/SL can accurately detect clinically relevant micrometastases from lesions >2.0 mm (Table 2). However, LM/SL underestimates the true incidence of these metastases in patients with primary lesions ≤ 2.0 mm, because in this group the long-term rate of regional nodal recurrence after WE alone exceeds the rate of IH-positive SNs.

Mapping the SN Microanatomy with Carbon Dye

The risk of false-negative results for LM/SL in patients with primary melanomas <1.0 and 1.01–2.00 mm is clearly indicated by the discrepancy between the rate of nodal recurrence after WE alone and the rate of IH-positive SNs found at LM/SL (Table 2). We have developed 2 approaches to reduce the incidence of false-negative results due to IH. The first is the use of carbon dye to better localize the site of micrometastases in the SN.^14,19 Although currently used vital dyes and radioactive tracers may target a specific site within the SN, they are not useful to the pathologist because they diffuse from this site before the node is received by pathology from surgery. Based on our preliminary preclinical and clinical studies,^14,19 we hypothesized that a carbon particle suspension would allow histopathologic confirmation by the pathologist of a node’s identity as a true SN and serve as a control for the accuracy of the surgical procedure in identifying the true SN. In addition, our recent investigations indicate that the pattern of carbon deposition within the SN might show the intranodal site of tumor cells (Fig. 3). Table 3 shows the results of LM/SL performed after coinjection of carbon dye with isosulfan blue dye into the site of the primary melanoma. Carbon was identified in the SNs of 159 of 166 (96%) patients; 25 of 166 (16%) patients had IH-positive SNs, and in all 25 cases we found metastases only at the intranodal site of the carbon. Thus, the carbon dye permits histologic confirmation of SN status by the pathologist. In addition, the carbon particles direct the pathologist to the SN most likely to contain tumor when more than 1 SN has been removed, and the location of carbon particles within this SN indicates the particular nodal area that drains the primary cutaneous site where micrometastatic tumor cells will be found if they are present (Fig. 3).

FIGURE 3. Compartmentalization of intranodal lymph flow as detected in the operating room by radiopharmaceutical, blue dye, and carbon dye (left) and confirmed in the pathology department by carbon dye (right).

TABLE 3. Tumor Status of SNs Identified by a Surgeon Using Blue Dye and Radiocolloid in Relation to the Presence or Absence of Carbon Particles Identified by Pathologist

Molecular Staging of the Regional Nodes Using Sections of Paraffin-Embedded SNs

Among the 215 patients whose SNs were studied by multimarker molecular assays, 162 patients had IH-negative SNs. Of the 162, 49 (30%) had SNs that expressed at least 1 of the 4 qRT markers. These patients had a significantly higher risk of disease recurrence and death than did patients with negative IH and qRT results (P ≤ 0.0001) (Fig. 4, left). When all SNs were negative by IH and qRT, the rate of long-term OS exceeded 95% (Fig. 4, right). Because the multimarker qRT approach can avoid potential false-positive results and improve sensitivity, its results have stronger prognostic significance than those of single-marker assays. Moreover, the use of multiple markers circumvents the problem of heterogeneous marker expression among different melanomas. Finally, when compared with gel-based analysis systems, quantitation of mRNA copy numbers by qRT allows more accurate evaluation.

FIGURE 4. Left, Kaplan–Meier estimates of disease-free survival (DFS) in patients with H&E/IH-positive SNs (N = 53) and IH-negative SNs (N = 162) according to SN expression of no mRNA markers or at least 1 mRNA marker (tyrosinase, MART-1, MITF, and/or TRP-2). Right, Kaplan–Meier estimates of OS according to multimarker qRT and histopathology status of SNs in 215 patients.

Does LM/SL Improve Melanoma-Specific Survival Compared With WE Alone?

Table 4 gives the matching factors used to compare survival associated with early CLND (nodal metastases identified by histopathology of SN after LM/SL) versus delayed CLND (nodal metastases identified by palpation during observation after WE). A computer program was designed to randomly match the 2 groups of patients by the following prognostic variables: pT stage, ulceration, age, sex, and the total number of tumor-involved nodes. Figure 5 shows improved survival from the time of WE for patients who underwent immediate CLND versus delayed CLND. The respective 5-, 10-, and 15-year survivals were 73%, 69%, and 69% versus 51%, 37%, and 32% (P ≤ 0.001). This clearly indicates that a significant proportion (69%) of patients with nodal metastases have disease limited to the nodal site and therefore are potentially curable by early resection. However, even at a later date when the nodes are palpably enlarged, therapeutic CLND may be curative since 32% of such patients are alive at 15 years. These observations are most compatible with the incubator hypothesis because they suggest a period of latent growth in the regional nodes before the development of systemic metastases; CLND at that time can abort the metastatic cascade and lead to prolonged survival or cure.

TABLE 4. Prognostic Factors Used to Match 287 Patients Undergoing Immediate CLND (Tumor-Involved SN) With 287 Patients Undergoing Delayed CLND (Palpable Lymph Nodes)

FIGURE 5. Kaplan–Meier estimates of OS after WE for immediate versus delayed therapeutic CLND in 287 pairs of patients matched by the criteria shown in Table 4.

DISCUSSION

According to the incubator hypothesis (Fig. 1), there is a therapeutic window of opportunity when the metastatic cascade can be interrupted by removing tumor-involved lymph nodes before their micrometastases spread to distant sites. However, if too much time has elapsed before the SN is removed, distant metastasis may already have occurred. Thus, the proportion of patients who might benefit from early versus delayed removal of tumor-involved regional lymph nodes can be calculated as follows:

100% - [(the proportion of patients without SN micrometastases) + (the proportion of patients with SN metastases and systemic metastases at the time of LM/SL) + (the proportion of patients whose metastases will never achieve the ability to extend beyond the regional basins to distant sites and therefore can be salvaged by observation and delayed therapeutic CLND)]

If the matched-pair data shown in Table 4 and Figure 5 are representative of the entire population of patients with primary melanoma, we can calculate the expected overall benefit of LM/SL as follows:

100% - (80% rate of no nodal metastases) + (31% rate of melanoma deaths 15 years after WE and LM/SL × 0.20) + (32% rate of 15-year survival after CLND for palpable nodes × 0.20) = 100% - 92.6% = 7.4%.

Interestingly, the overall survival benefit of immediate CLND identified by Balch et al⁵ in the Melanoma Intergroup trial was only 4%, which was not significant (P ≤ 0.25), but reached 6–7% (P = 0.03) for selected subsets with nonulcerated or thin (1.01–2.0 mm) primary melanomas.

Because the incidence of nodal metastases clearly depends on the thickness of the primary melanoma (Table 2), early removal of the SN should have little effect on melanoma-specific survival when the primary tumor is <1.0 mm. Only 12% of these patients have nodal micrometastases; in the remaining 88%, the primary is removed before the melanoma has had an opportunity to spread to the regional nodes. Therefore the possible enhanced survival benefit would be less than 4%.

The problem of the false-negative SN cannot be overemphasized but it can be reduced by histopathologic sampling techniques that avoid errors due to the nonrandom distribution of tumor cell micrometastases within the SN. The use of carbon dye as a mapping adjunct can confirm the identity of the SN and permanently mark the correct portion of this node most likely to contain tumor cells. The risk of false-negative findings can also be further minimized by multimarker molecular assessment of the SN for clinically relevant submicroscopic micrometastases.

This article provides compelling rationale for the routine use of carbon dye to help the pathologist “map” the SN and confirm that the correct portion of this node is being examined for micrometastases. It also suggests that examination of the SN by multimarker qRT should become the standard of care in all patients whose SNs are negative by routine IH. Patients whose SNs are negative on examination by molecular techniques have enhanced long-term survival if the true SN is examined. Because the absence of such molecular metastases predicts 90% long-term DFS and >95% OS, these patients need little or no follow-up except by their dermatologist to detect the development of second primary melanomas.²⁰ By contrast, patients whose nodes contain mRNA for 1 of the melanoma multimarkers or have IH-detected nodal metastases have a greatly increased risk of recurrence and should receive adjuvant therapy and more frequent follow-up. The quality of RT-PCR analysis and the interpretation of results are critical because false-positive results of SN analysis might lead to unnecessary adjuvant therapy. Still, the results of molecular staging of the SN strongly support the incubator hypothesis, since 90% of patients with qRT-negative SNs will remain free of recurrence within 80 months and 95% will be long-term survivors. Thus SN micrometastases almost always precede systemic metastases, as stated in the incubator hypothesis. Recurrence in the regional lymph node basin or at distant sites in patients with IH-negative SNs therefore is probably the result of failure to detect tumor cells in transit from the primary or clinically relevant submicroscopic SN metastases rather than failure to identify the true SN. Most importantly, this study demonstrated clinical utility for qRT assessment of PE SNs. This technique could be easily applied in multicenter trials, and its results should significantly improve the staging of early melanoma.

During the past few years, we have learned a great deal regarding the pathophysiology of metastasis from the primary melanoma to the regional SNs to distant sites. The primary melanoma produces immunosuppressive factors, as we had previously hypothesized many years ago.^21–24 These factors lead to dramatic and profound immunosuppression of the SN exposed to the tumor’s influence via the direct drainage lymphatic pathway from the primary melanoma; non-SNs in the regional basin remain unaffected or less affected. This immunodownregulation is evident in multiple ways but especially in the density and maturity of interdigitating dendritic cells and T cells in the SNs (Fig. 6). ²⁴ We believe this immunosuppression leads to a reduction in host immunity that allows the successful implantation and growth of micrometastases in the SN (Fig. 1). Once the volume of metastatic tumor cells in the SN exceeds a critical mass, the production of immunosuppressive factors leads to further immunosuppression of lymph nodes along the lymphatic chain, and then to micrometastatic involvement of additional lymph nodes, which increases the risk of systemic immunosuppression and hematogenous spread. This is dramatically demonstrated by the highly significant increased incidence in the number of tumor-involved nodes identified when CLND is performed to remove palpable rather than subclinical metastases (Table 2). After the development of systemic metastases, surgical removal of the regional nodes should not influence the growth of distant metastases (unless these are very small and composed of a few tumor cells that might be rejected by host immune factors) nor change the long-term melanoma-specific survival. Thus, the patients who are most likely to derive benefit from LM/SL are those with thinner primary melanomas whose SN metastases are still “incubating” in the SN prior to systemic spread. Our data and this concept are fully compatible with the recent results of Balch et al,^4,5 who found a benefit for ELND only in patients whose primary melanomas were 1–2 mm in thickness or nonulcerated.

FIGURE 6. Comparison of the density and area of dendritic cells, T cells and B cells in 21 matched pairs of SNs and nonsentinel nodes (NSNs) from 11 patients with AJCC stage I/II melanoma. Each SN was size-matched with a NSN from the same patient (data from Cochran et al²⁴).

In summary, the risk of metastasis to the SN and to non-SNs is directly proportional to the thickness of the primary melanoma: primary lesions <1 mm have a low incidence of SN metastases, whereas lesions ≥1.0 mm are more likely to be associated with metastases in the SN. Intermediate lesions of 1.01–2.00 mm are more likely to have metastases only in the SN, whereas thicker lesions ≥2.01 are more likely to have additional involved lymph nodes and metastases at distant sites. Although primary tumor size as shown in Table 2 is probably the most important factor in determining the risk of regional nodal metastases and systemic metastases, other factors such as the extent of genetic changes in the primary melanoma associated with activation of various growth-factor oncogenes and the patient’s immune defenses are clearly involved. These factors would explain why a clinically palpable (2–3 cm diameter) nodal metastasis containing up to 32 × 10⁹ melanoma cells can exist without any evidence of additional nodal or systemic involvement. We have found approximately 40% of patients with delayed nodal recurrence after WE alone will still have a single involved node, and 32% of patients with 1 or more palpable nodal metastases will survive at least 15 years (Fig. 5), because their metastases are confined to the regional nodes by immune defenses that inhibit the implantation and growth of blood-borne circulating tumor cells (Fig. 1) or genetic characteristics of the metastatic clone which make tumor cells unable to adhere to the vascular endothelium and grow at distant sites.

We believe the endogenous host immune response to the growth of a primary melanoma probably is the most important factor in determining the long-term survival after surgical therapy for melanoma. Recent data from our laboratory have shown that the endogenous immune response to TA90, a melanoma-associated antigen, is the strongest independent prognostic factor in patients undergoing complete resection of distant metastases or receiving postsurgical adjuvant therapy with Canvaxin vaccine, a therapeutic polyvalent cancer vaccine.^25,26 Further study of the relationship of specific host immunity to tumor antigens in relation to the outcome of surgical therapy is clearly indicated. Our attempts to enhance this host immunity by active specific immunity with Canvaxin vaccine have been promising.^27,28

Discussion

Dr. Charles M. Balch (Alexandria, Virginia): I want to congratulate Dr. Morton and his colleagues for yet another major contribution to our understanding about the role of lymphatics in melanoma and indeed a physiologic principle that applies to other forms of cancer as well.

It is important to give credit to Dr. Morton who has been contributing pioneering work in this area for over 3 decades. He first described cutaneous lymphoscintigraphy, or lymphatic mapping, in 1974. He was the first to describe the concept of lymphatic mapping in sentinel node excision, which was presented at this Society’s meeting in 1991. And now he has gone to the next level of staging sophistication with intranodal mapping using the carbon black technique.

All of these techniques have enabled a more precise staging of metastatic melanoma, with reproducible detection of nodal metastasis down to a level of 10 to the fourth or 10 to the sixth cells, and sometimes less. No other staging technique has such precision for detecting micrometastases in melanoma. Indeed, the American Joint Committee on Cancer has now formally recommended that all patients with T2, T3, and T4 melanomas undergo the lymphatic mapping and sentinel node technique as a prerequisite for entry into melanoma clinical trials, both surgical and adjuvant therapy trials.

In this presentation today, Dr. Morton has shown that intranodal mapping with the carbon black technique and the use of molecular markers continues to improve our ability to partition patients into different risk categories and then apply appropriate treatments calibrated to the biology of their disease. So, from a staging perspective, Dr. Morton’s pioneering work in lymphatic mapping and sentinel node excision is a seminal contribution that has now been reproduced worldwide and adopted as a standard staging tool for melanoma. Importantly, this staging tool is increasingly finding value in a wide variety of other cancers as well.

Another focus of Dr. Morton’s presentation today asks the age-old question about the treatment value of lymphadenectomy, something that has been debated before this Society for at least 4 decades or more. Many of you in the audience know that this issue actually began over a century ago, beginning with Halsted’s hypothesis that excision of lymph nodes had an important role in cancer patients and continuing in the opposite direction with the Fisher hypothesis that there is no therapeutic role whatsoever for lymphadenectomy. But it appears that we now have a sophisticated staging capability for reexamining the potential treatment value of lymphadenectomy and confining the indications for complete lymphadenectomy to patients with pathologically demonstrated nodal metastases. We now have to determine whether or not this operation improves survival rates. The hypothesis here, of course, is that the primary source of distant disease that ultimately kills the patient is more likely to come from the lymph node metastasis than from the primary tumor itself.

At this Association meeting in 1996 I had the privilege of presenting our randomized prospective trial demonstrating that in prospectively defined groups of T3 and T2 melanoma patients there was a survival benefit with elective lymph node dissection (Ann. Surg. 224:255, 1996; Ann. Surg. Oncl. 7:87, 2000). This trial used a statistical probability of nodal metastases, and so included patients who, in retrospect, did not need a lymphadenectomy because after the lymph nodes were examined pathologically, they did not have demonstrated metastasis. However, this was a ‘proof of concept‘ study and thankfully the sentinel node technique has completely supplanted the need for elective node dissection and spares patients an operation for whom, in retrospect, they did not have metastasis to begin with.

To his credit, Dr. Morton has gone on and performed a prospective randomized trial comprised of 2,000 patients randomly selected to sentinel node excision and lymphadenectomy only if the nodes were positive compared with patients who had nodal observation and no sentinel node excision whatsoever. We are still awaiting results of this trial.

My final comment regards the extent of pathologic examination of the sentinel node by the pathologist, something that is still variable among pathology departments. An important conclusion of this paper is that we are still understaging patients even with serial sections of the sentinel node and that molecular marker techniques with RTPCR are more likely to demonstrate micrometastasis that otherwise might be missed by random slices through a lymph node.

Jeff Gershenwald and colleagues from M. D. Anderson similarly showed in patients who had a false negative result after sentinel lymph node excision – that is, when they relapsed later on and the original sentinel node was re-examined, they actually contained microscopic nodal metastasis when they went back and did serial sections (J. Clin. Onol. 16:2253, 1998).

In this new era of molecular diagnostics and improved staging for micrometastases, we will need to take yet another look at how we apply various surgical treatment options and what type of adjuvant therapy we give for our melanoma patients.

I would like to ask Dr. Morton 2 questions. First, under what circumstances with these new molecular markers and the ability to detect 100, 1,000, or 10,000 cells, would you recommend a completion lymphadenectomy? And second, what do we ask our pathologists to do after we have excised a sentinel node with regard to the use of immunohistochemical stains and RTPCR with molecular markers?

I want to congratulate you again on yet another outstanding contribution to the field and for the privilege of addressing this audience. Thank you.

Dr. Donald L. Morton (Santa Monica, California): Thank you, Dr. Balch. I appreciate your comments.

There are no objective data to support or refute complete lymph node dissection for molecular-positive sentinel nodes. A definitive answer must await completion of our large (N 4200) multicenter trial that will randomize patients with molecular-positive sentinel nodes to completion lymphadenectomy or to observation. This trial will begin shortly at about 25 melanoma centers in the United States, Europe and Australia. During the next decade it should produce results that will answer your question.

With respect to the number of sections to be examined by the pathologist, the usual range is 12 to 20, which represents about 1% of the total nodal volume. This is adequate for melanomas 2 mm or greater in thickness, but misses 40% of micrometastases from thinner primary melanomas. It is clear that nodal micrometastases are missed because the wrong portion of the sentinel node is being examined or the number of tumor cells is below the sensitivity of immunohistochemistry. I predict that the use of carbon to pinpoint the intranodal site of metastases and the application of molecular staging to correct these problems will become standard of care within a few years. These techniques should distinguish, with a high degree of accuracy, patients who have high risk of recurrence and need adjuvant therapy plus close follow-up, from patients in whom surgical resection alone is curative.