predicting sentinel and residual lymph node basin disease after sentinel lymph node biopsy for...

Predicting Sentinel and Residual Lymph Node BasinDisease after Sentinel Lymph Node Biopsyfor Melanoma

Jeffrey D. Wagner, M.D.1

Michael S. Gordon, M.D.2

Tsu-Yi Chuang, M.D.3

John J. Coleman III, M.D.1

John T. Hayes, M.S.4

Sin-Ho Jung, Ph.D.4

Charlene Love, R.N.1

1 Department of Surgery/Plastic and Reconstruc-tive Surgery, Indiana University School of Medi-cine, Indiana University-Purdue University at Indi-anapolis, Indianapolis, Indiana.

2 Department of Medicine/Hematology/Oncology,Indiana University School of Medicine, Indiana Uni-versity-Purdue University at Indianapolis, Indianap-olis, Indiana.

3 Department of Dermatology, Indiana UniversitySchool of Medicine, Indiana University-Purdue Uni-versity at Indianapolis, Indianapolis, Indiana.

4 Department of Medicine/Biostatistics, IndianaUniversity School of Medicine, Indiana University-Purdue University at Indianapolis, Indianapolis, In-diana.

Presented at the 35th American Society of ClinicalOncology Meeting, Atlanta, Georgia, May 16, 1999.

Address for reprints: Jeffrey D. Wagner, M.D., RT471 Cancer Pavilion Bldg., 535 Barnhill Drive,Indianapolis, IN 46202.

Received November 8, 1999; revision receivedMarch 15, 2000; accepted March 15, 2000.

BACKGROUND. The selection of patients for sentinel lymph node biopsy (SNB) and

selective lymphadenectomy for histologically positive sentinel lymph nodes

(SLND) are areas of debate. The authors of the current study attempted to identify

predictors of metastases to the sentinel and residual nonsentinel lymph nodes in

patients with melanoma.

METHODS. The Indiana University Interdisciplinary Melanoma Program comput-

erized database was queried to identify all patients who underwent SNB for

clinically localized cutaneous melanoma. Demographic, surgical, and histopatho-

logic data were recorded. Univariate and multivariate logistic regression analyses

were performed to identify associations with SNB and nonsentinel lymph node

positivity. Classification tree and logistic procedures were performed to identify the

ideal tumor thickness cutpoint at which to perform SNB.

RESULTS. Two hundred seventy-five SNB procedures were performed to stage 348

regional lymph node basins for occult metastases from melanoma. Of the 275

melanomas, 54 (19.6%) had a positive SNB, as did 58 of 348 basins (16.7%).

Classification and logistic regression analysis identified a Breslow depth of 1.25

mm to be the most significant cutpoint for SNB positivity (odds ratio 8.8:1; P

5 0.0001). By multivariate analyses, a Breslow thickness cutpoint $ 1.25 mm (P

5 0.0002), ulceration (P 5 0.005), and high mitotic index (. 5 mitoses/high-power

field; P 5 0.04) were significant predictors of SNB results. SLND was performed in

53 SNB positive patients, 15 of whom (28.3%) had at least 1 additional positive

lymph node. SLND positivity was noted across a wide range of primary tumor

characteristics and was associated significantly with multiple positive SN, but not

with any other variable. SNB result correlated significantly with disease free and

overall survival.

CONCLUSIONS. Patients with a Breslow tumor thickness $ 1.25 mm, ulceration, and

high mitotic index are most likely to have positive SNB results. SLND is recom-

mended for all patients after positive SNB because it is difficult to identify patients

with residual lymph node disease. Cancer 2000;89:453– 62.

© 2000 American Cancer Society.

KEYWORDS: melanoma, lymph nodes, sentinel lymph node biopsy, metastases.

Cutaneous melanoma is rapidly increasing in incidence in theUnited States and elsewhere.1–3 By virtue of its metastatic poten-

tial, melanoma easily accounts for the vast majority of deaths fromcutaneous neoplasms. Surgical excision is the treatment of choice forboth primary lesions and clinically obvious regional lymph nodemetastases. Most patients with clinically localized disease can berendered clinically free of disease, but many with deep primary le-

453

© 2000 American Cancer Society

sions or regional lymphatic metastases will later die ofdistant metastatic disease.

Identification of the patient with occult mela-noma metastases is important for accurate staging,treatment planning, and prognosis. Factors predictiveof treatment failure in patients with melanoma aredescribed, including increasing tumor thickness, axialtumor location, ulceration, male gender, and age.4 – 6

However, prognostic factors only identify subsetswithin a population that are at increased risk for re-currence without identifying individuals who actuallywill recur. The demonstration of regional or distantmetastases provides the most important prognosticinformation, taking precedence over any primary tu-mor characteristic.7

The most common site of occult melanoma me-tastases and recurrence is the regional lymph nodes.The only accurate technique for staging nonpalpablelymph nodes is surgical excision and histologic anal-ysis. Elective lymphadenectomy based on melanomatumor characteristics is controversial because of mor-bidity associated with the procedure and absence ofan overall survival benefit, although subsets of pa-tients may benefit from this procedure.8 –11 It recentlyhas become possible to identify the subset of patientswith occult regional lymph node metastases withoutsubjecting them to complete formal lymphadenec-tomy. The development of sentinel lymph node bi-opsy (SNB) technology12–16 has been an importantadvance in the management of patients with mela-noma. In qualified hands, this procedure yields diag-nostic information comparable to complete lymphad-enectomy with considerably less morbidity. In mostcases, the sentinel lymph node(s) is the sole site ofmetastatic disease in the regional lymph node ba-sin.17,18 The histologic status of the sentinel lymphnode recently has been shown to be an importantprimary prognostic factor in melanoma.19,20

Although SNB is a relatively low morbidity proce-dure, the risk of potential complications and cost dic-tate that patients be appropriately selected for surgicalstaging of regional lymph nodes. Selection of patientsfor SNB is an area of debate. The decision to surgicallystage regional lymph nodes should be based on sev-eral factors, including the sensitivity of the diagnosticprocedure, the likelihood of finding occult metastaticdisease, the impact of such findings on treatment, andother clinical and psychosocial issues. For example,the incidence of occult regional lymph node metasta-ses in a population of thin melanomas is expected tobe low, but the knowledge of such disease in an indi-vidual would upstage the patient, worsen prognosis,and change treatment and surveillance algorithms.

Another area of controversy surrounds the need

for completion lymphadenectomy after a positive SNBresult. The incidence of residual occult disease re-maining in nonsentinel lymph nodes after a positiveSNB is variously reported to be 7–33%.12,13,16 –18 Thisrelatively low incidence, coupled with the uncertainoverall survival benefit of elective lymphadenectomy,has generated debate over the need to perform a ther-apeutic completion lymphadenectomy after positiveSNB. The purpose of this study is to identify patientand tumor factors that may be predictive of occultsentinel and nonsentinel lymph node metastases inpatients undergoing SNB for cutaneous melanoma lo-calized to the skin.

METHODSThe Indiana University Interdisciplinary MelanomaProgram computerized database was queried to retro-spectively identify patients who underwent SNB forclinical Stage I cutaneous melanoma at Indiana Uni-versity Medical Center and affiliated hospitals be-tween 11/1/1994 and 3/1/1999. During this time, SNBwas offered as a lymph node staging option to patientswith cutaneous melanoma who had no clinical evi-dence of lymph node or distant metastases. Writteninformed consent was obtained from each patient (orguardian) for the operative procedures. In general,SNB was offered to patients with primary melanomagreater than or equal to 1.0 mm Breslow thickness andthose with locally recurrent melanoma after prior ex-cision. Later, criteria were expanded to include pa-tients with primary melanomas less than 1.0 mm whowere thought to be at significant risk of occult lymphnode metastases and patients who requested the pro-cedure for prognostic information. All patients in thestudy population had melanoma localized to the skinwith nonpalpable regional lymph node basins and noevidence of distant metastatic disease at the time ofSNB (American Joint Committee on Cancer [AJCC]Stages I, II, and III ).21 Hospital and outpatient clinicalrecords were reviewed, and pertinent demographic,surgical, and histopathologic data were recorded.

Sentinel Lymph Node Biopsy TechniqueWe used the technique of SNB previously reported.17

Visual blue dye and radiolocalization were generallyused in a complimentary fashion to help ensure re-moval of all sentinel lymph nodes. Briefly, preopera-tive dynamic dermatolymphoscintigraphy was per-formed to identify the basin(s) at risk for lymph nodedisease. One to two millicuries of Tc99M sulfur colloidwas injected intradermally in 2– 4 divided doses at thetumor site 2– 4 hours before surgery. Continuous im-aging with a large field of view gamma camera wasperformed, depending on location of the melanoma,

454 CANCER July 15, 2000 / Volume 89 / Number 2

for up to 2.5 hours for axial tumor sites with poten-tially equivocal lymphatic drainage. After the induc-tion of anesthesia, 0.5–2.0 mL of isosulfan blue dye(Lymphazurin; Zenith Parenterals, Rosemont, IL) wasinjected intradermally at the site of the primary mel-anoma. All basins identified by lymphoscintigraphywere explored through limited incisions by using thehand-held gamma probe (C-Track; Care Wise MedicalProducts, Morgan Hill, CA). The primary indicator ofSN status was visual blue coloration. All blue lymphnodes were removed as SNs. After removal of bluelymph nodes, radioactivity ratios were the primarymeans of ensuring complete removal of all sentinellymph nodes. The ratio of the most radioactive bluesentinel lymph node compared with residual lymphnode basin radioactivity was determined. If necessary,additional nonblue radioactive lymph node(s) wereremoved as sentinel lymph nodes until the ratio of themost radioactive ex vivo SN to residual basin was$ 10:1. If no blue lymph nodes were found, the mostradioactive lymph node(s) were removed as sentinellymph nodes until the ex vivo SN to residual basinratio was $ 10:1.

Histologic Analysis of Surgical SpecimensSuspicious SNs were submitted for intraoperative fro-zen section analysis, at the discretion of the attendingsurgeon. Nonsuspicious and frozen section negativeSN were fixed in formalin and submitted for 1-mmstep sections. These sections were analyzed with he-matoxylin and eosin (H & E) stains. Lymph nodesnegative for metastases by this analysis had additionalsections stained with immunohistochemical markersto S-100 and/or HMB-45. Sentinel lymph nodes withevidence of metastases by light microscopy or immu-nohistochemical analysis were considered positive,and selective complete lymphadenectomy (SLND) wasperformed (or recommended) on the involved ba-sin(s). Nonsentinel lymph nodes and SLND specimenswere analyzed in routine fashion after formalin fixa-tion, with one to three sections from lymph nodesreserved for H & E staining.

Statistical AnalysisUnivariate analyses were performed to determine pos-sible relations between several independent variablesand the dependent variables of SNB and SLND histol-ogy. The following variables felt to be potentially pre-dictive of SN and SLND positivity were analyzed: gen-der, age, tumor location, growth phase, regression,lymphocytic tumor infiltrate, melanoma subtype, mi-totic index, ulceration, Clark level, Breslow thickness,AJCC tumor stage, number of basins biopsied, numberof SNs harvested, and presence of multiple positive

SN. Univariate analysis was performed using exactlogistic procedure. An exact multivariate regressionmodel then was devised for analysis of SNB result(dependent variable). A univariate p value of less than0.25 was required to enter the multivariate model.Backward and forward selection procedures with asignificance level of 0.10 for staying in the model wereused.

Two methods were used to categorize Breslowtumor thickness to identify the most efficient cutpointfor predicting SNB result. First, classification tree anal-ysis with SNB result as the outcome was performed.Second, a logistic procedure controlling for the signif-icant univariate covariables unrelated to tumor thick-ness (mitotic index, ulceration, growth phase, and re-gression) was performed to verify the classificationtree result.

RESULTSTwo hundred seventy-five consecutive SNB proce-dures were performed to stage 348 regional lymphnode basins in patients with CS-I cutaneous mela-noma. All patients had nonpalpable regional lymphnodes by clinical examination. The mean age of thestudy population was 51 years (range, 4 –92 years).Other population characteristics are shown in Table 1.For some variables, some results are unknown. MeanBreslow tumor thickness was 2.24 mm (range, 0.2–14.0mm). At least 1 SN was successfully localized andharvested in 346 basins (99.4%). SNs were not identi-fied in 2 basins (0.6%).

A total of 731 SNs were removed (mean, 2.1 SN perbasin; range, 1– 8). Fifty-four melanomas (19.6%) and58 basins (16.7%) had a positive SNB (4 patients eachhad 2 positive basins). The method of pathologic de-tection was by H & E light microscopy in 54 basins(93.1%) and by immunohistochemistry alone in 4 ba-sins (6.9%). SNB positivity correlated with AJCC tumorstage. The mean Breslow thickness of 3.01 mm for SNBpositive melanomas was significantly greater than 2.05mm for SNB negative melanomas (P 5 0.003, t test).

Univariate analysis of several patient and tumorcategoric variables by SNB result for 275 melanomas isillustrated in Table 2. Breslow thickness was analyzedas both continuous variable and as a categoric variableby using the derived cutpoint of 1.25 mm (see Fig. 1).Age was analyzed both as a continuous variable and asa categoric variable by using 50 years (median age forthe population) as a cutpoint. Clark level was analyzedas an ordinal variable and as a categoric variable.Growth phase was analyzed as categoric variable (ver-tical component present vs. absent). Mitotic index wasclassified as high ($ 6 mitoses per high-power field[hpf]), intermediate (3–5 mitoses/hpf), or low (# 2

Predicting Lymph Node Disease in Melanoma/Wagner et al. 455

mitoses/hpf) or indeterminate and analyzed as a cat-egoric variable. Lymphocytic tumor infiltrate wasgraded as brisk, moderate, minimal/absent, or inde-terminate and analyzed as a categoric variable. Loca-tion was graded as axial or extremity with shoulderand hip areas counted as axial. Ulceration and regres-sion were graded as present or absent. Ulceration wasconsidered present if noted histologically or if docu-mented by history or physical examination. Indeter-minate values were considered as separate categoriesand included for analyses, but unknown values wereexcluded. Univariate analysis revealed tumor thick-ness was significant when analyzed either as a contin-uous or as a categoric variable. SNB result also corre-lated with AJCC tumor stage (P 5 0.0001), ulceration(P 5 0.0002), mitotic index P 5 0.006), and Clark level(P 5 0.02).

TABLE 1Distribution of Clinical Variables by Sentinel Lymph Node BiopsyResult in 275 Melanoma Patients

Variable No. (%) SNB negative No. (%) SNB positive

Age (yrs)# 51 108 (81.8) 24 (18.2). 51 113 (79.0) 30 (21.0)

GenderFemale 93 (82.3) 20 (17.7)Male 128 (79.0) 34 (21.0)

Tumor stageT1 35 (97.2) 1 (2.8)T2 83 (89.2) 10 (10.8)T3 82 (73.9) 20 (26.1)T4 19 (57.6) 14 (42.4)Unknown 2 (100) 0 (0)

Breslow 1.25 1 (mm), 1.25 91 (95.8) 4 (4.2)$ 1.25 128 (71.9) 50 (28.1)

Tumor subtypeDesmoplastic 5 (100) 0 (0)Lentigo maligna 6 (100) 0 (0)Nodular 42 (77.8) 12 (22.2)Acral lentiginous 6 (66.7) 3 (33.3)Superficial spreading 130 (82.3) 28 (17.7)Indeterminate 32 (74.4) 11 (25.6)

Growth PhaseVertical 158 (77.4) 46 (22.6)Nonvertical 63 (88.7) 8 (11.3)

Mitotic indexHigh 19 (59.4) 13 (40.6)Intermediate 30 (75) 10 (25)Low 153 (85) 27 (15)Indeterminate 19 (82.6) 4 (17.4)

Lymphocytic infiltrateBrisk 29 (80.6) 7 (19.4)Moderate 48 (82.8) 10 (17.2)Minimal/absent 93 (76.9) 28 (23.1)Indeterminate 51 (85) 9 (15)

Clark levelI 1 (100) 0 (0)II 17 (100) 0 (0)III 51 (83.6) 10 (16.4)IV 138 (77.1) 41 (22.9)V 9 (75) 3 (25)Unknown 5 (100) 0 (0)

UlcerationPresent 88 (69.8) 38 (30.2)Absent 128 (88.9) 16 (11.1)Unknown 5 (100) 0 (0)

RegressionPresent 27 (90) 3 (10)Absent 184 (79.3) 48 (20.7)Unknown 10 (76.9) 3 (23.1)

LocationAxial 140 (78.2) 39 (21.8)Extremity 81 (84.4) 15 (15.6)

No. basins biopsied1 168 (81.6) 38 (18.4)2 51 (78.5) 14 (21.5)3 2 (50) 2 (50)

SNB: sentinel lymph node biopsy.

TABLE 2Sentinel Lymph Node Biopsy Result: Univariate Logistic Regression

VariableOddsratio

ExactP-value

Age (,vs.. 51 yrs) 1.19 0.67Gender (male) 1.23 0.61Breslow (continuous) 1.24 0.002Breslow (1.25 1 mm) 8.83 0.0001Tumor stage 2.65 0.0001Clark level 1.83 0.02Clark # II vs. $ III 6.86 0.03Clark # III vs. $ IV 2.06 0.07Mitotic index (high vs. others) 3.35 0.006Mitotic index (high or intermediate vs. others) 2.59 0.005Lymphocytic infiltrate 1.01 1.0Regression (yes) 0.43 0.24Ulceration (yes) 3.44 0.0002Vertical growth 2.29 0.05Location (axial) 1.50 0.29Nodular subtype 1.22 0.72

FIGURE 1. Plot of Z-statistics for all Breslow tumor thickness cutpoints

demonstrates that the most significant cutpoint (Z 52.57) is at 1.25 mm.


Classification tree analysis demonstrated Breslowtumor thickness of 1.23 mm to be the optimal decisioncutpoint for predicting SNB result. Verification of thisfinding by logistic procedure controlling for possiblesignificant nonthickness covariables confirmed a sim-ilar Breslow cutpoint of 1.25 mm to have the highestsignificant standardized parameter value (Fig. 1). Thiscutpoint yielded a sensitivity of 0.93, specificity of0.42, positive predictive value of 0.28, and negativepredictive value of 0.96. The association of the cate-goric 1.25-mm cutpoint with SNB result was highlysignificant (P 5 0.0001).

Multivariate logistic regression was performed todemonstrate the possible relation(s) between vari-ables with a univariate P value less than 0.25 and SNBresult. Variables included were ulceration, high mi-totic index, presence of vertical growth phase, regres-sion, Clark level, and tumor thickness. Tumor thick-ness was the most significant variable identified byunivariate analysis and was modeled in 3 ways: as acontinuous variable, as a categoric variable usingAJCC tumor stage cutpoints, and with the derivedcutpoint of 1.25 mm. Twenty-one patients had miss-ing data and were excluded from analysis leaving 254patients in the multivariate model. The results of theseanalyses are shown in Table 3. The 1.25-mm categoricBreslow thickness demonstrated the most significantassociation with SNB result (P 5 0.0002). Ulceration (P5 0.05) was marginally significant, and mitotic index(P 5 0.08) was not significant in the 1.25-mm Breslowmodel. Tumor thickness was not a significant inde-pendent predictor of SNB result when modeled as acontinuous variable or using AJCC tumor stage cut-

points. However, ulceration and mitotic index wereeach independent predictors in both of these models(Table 3).

Fifty-three patients with a positive SNB under-went SLND of 57 basins. Fifteen patients (28.3%) hadat least 1 positive lymph node in the SLND specimen.Tumor characteristics and procedure data for thesepatients are shown in Table 4. Residual tumor-con-taining lymph nodes were found in the SLND speci-men across a wide range of tumor depths (range,1.3–10.0 mm; mean depth, 2.86 mm). Ulceration wasnoted in 6 patients with positive SLND (40%). SLNDresult by AJCC tumor stage cutpoints is shown inTable 5. No patients with T1 tumors had a positiveSLND result, but the rate of SLND positivity was sim-ilar among T2-4 lesions. Univariate analysis of therelation between tumor variables and SLND resultafter positive SNB is shown in Table 6. The presence ofmultiple positive sentinel lymph nodes was associatedwith SLND result (P 5 0.002), but was noted in only 6(40%) of SLND positive patients. No other variableswere significant. Multivariate modeling was not per-formed because of the small sample size.

At a median clinical follow-up of 12 months(range, 1–56 months) 23 patients (8.4%) had devel-oped recurrence and 14 (5.1%) had died of dissemi-nated melanoma. Disease free and overall survivalwere significantly associated with SNB result (Figs. 2and 3).

DISCUSSIONUntil recently, SNB technology has been available pri-marily in large melanoma centers, where it has beenperformed mostly in the context of investigational orclinical protocols. With the diagnostic efficacy of SNBin experienced hands now established, more surgeonsare now performing SNB in clinical practice. It istimely to consider optimal selection criteria for thisprocedure for patients with clinically localized mela-noma.

The status of the sentinel lymph node is clinicallyimportant for at least three major reasons. First, theSN status has been shown to be an important prog-nostic factor, with a positive result predictive of highrisk for treatment failure.19,20 The predictive value of anegative SNB is not certain at this time but may sug-gest a more favorable prognosis than would be pre-dicted by tumor characteristics alone. Second, SNBidentifies with minimal morbidity those high risk pa-tients who may benefit from additional therapy suchas complete lymphadenectomy or adjuvant interfer-on,22 or be eligible for participation in clinical trials.Last, the psychologic benefit for the patient whoseSNB result does not reveal metastases seems signifi-

TABLE 3Sentinel Lymph Node Biopsy Result: Multivariate Logistic Regression

VariableaOddsratio P-value

Modeled with Breslow as continuous variableBreslow 1.11 0.14Mitotic index (high) 2.40 0.04Ulceration 2.65 0.005

Modeled with Breslow as categorical variable: 6 1.25 mmBreslow (1.25 1) 6.09 0.0002Mitotic index (high) 2.27 0.08Ulceration 2.07 0.05

Modeled with Breslow at AJCC tumor stage cutpointsT classification (1 is reference) 0.07

2 4.103 8.044 10.9

Mitotic index (high) 0.04Ulceration 0.03

AJCC: American Joint Committee on Cancer.


cant. The possible therapeutic benefit of the SNB pro-cedure itself and SNB-based therapeutic decisions isunknown but is currently under investigation in pro-spective clinical trials.

The sentinel lymph node positivity rate of 19.6%in this series is similar to other reports (12–36%).12–16

The positive correlation of sentinel lymph node posi-tivity with tumor thickness is expected and generallyhas been observed. Tumor thickness has historicallybeen accepted as the most important prognostic fac-tor in patients with clinically localized melanoma.With a positivity rate of only 2.8% for T1 lesions, SNBis a relatively low yield procedure in most thin mela-nomas. The SNB positivity rate for T4 melanomas ispredictably much higher (42.4% in this report). Studieshave shown lymph node status is an independentprognostic factor in patients with thick melanomas.With the availability of adjuvant interferon for highrisk melanoma, a rational argument can be made toconsider patients with thick melanomas, who ordi-

narily would not be offered classic elective lymphnode dissection, for lymph node staging with SNB.

This report demonstrates a significant positivecorrelation between SNB positivity and increasing tu-mor thickness. The lack of multivariate significancefor thickness as a continuous variable or by AJCCtumor stage category as a predictor of SNB result isunexplained but may be an artifact of inadequatesample size or possibly to the influence of multipleconfounding covariables in this model. Nonetheless,because of its reproducibility and widely acceptedprognostic importance, tumor thickness should be theprimary consideration when selecting patients for

TABLE 4Characteristics of Sentinel Lymph Node Positive Patients with Metastatic Melanoma in Selective Lymphadenectomy Lymph Nodes

Patient no.Depth(mm) Clark Mitotic index Ulceration No. SNs No. positive SNs

SLNDresult

1 4.00 IV Unclassified Yes 2 1 1/132 1.73 IV Intermediate No 1 1 1/123 2.70 IV High No 1 1 1/134 4.50 III Intermediate Yes 1 1 1/115 2.50 IV Low Yes 1 1 1/166 1.35 IV High Yes 2 1 1/107 4.70 V Low No 1 1 1/488 3.25 IV Low Yes 2 1 2/119 10.0 IV High Yes 2 1 1/910 4.00 IV High No 3 2 1/1811 3.50 IV Intermediate No 2 2 1/812 1.30 IV Low No 2 2 1/413 4.00 IV Low No 2 2 1/1414 2.90 III Intermediate No 2 2 2/2115 1.50 III Intermediate No 2 2 3/15

SN: sentinel lymph node; SLND: selective lymphadenectomy.

TABLE 5Selective Lymphadenectomy Result after Positive Sentinel LymphNode Biopsy by AJCC Tumor Stage

AJCC tumor stage No. (%) positive SNBs No. (%) positive SLNDs

T1 1 (2.8) 0 (0)T2 10 (10.8) 3 (30)T3 29 (26.1) 9 (32.1)a

T4 14 (42.4) 3 (21.1)

AJCC: American Joint Committee on Cancer; SNB: sentinel lymph node biopsy; SLND: selective

lymphadenectomy.a One patient refused SLND.

TABLE 6Selective Lymph Node Dissection Result after Positive SentinelLymph Node Biopsy: Univariate Logistic Regression

Variable Odds ratio P value

Age (. 51 yrs) 1.34 0.86Gender (male) 1.16 1.0Breslow (continuous) 1.12 0.44Breslow 1.25 1 mm 2.22 0.50Tumor stage 1.09 1.0Clark level 0.99 1.0Mitotic index (high vs. other) 1.17 1.0Lymphocytic infiltrate 0.42 0.77Regression (yes) 1.45 1.0Ulceration (yes) 1.51 0.85Vertical growth 0.61 0.81Location (axial) 0.63 0.70Nodular subtype 0.81 1.0Multiple (1) SN 14.64 0.002

SN: sentinel lymph node.


SNB. Clark levels generally are thought to be of sec-ondary prognostic importance when tumor thicknessis accounted for. Although Clark level was not anindependent predictor of SNB result, it was noted thatno patient with a Clark level I or II tumor had apositive SNB. These patients are at very low risk foroccult lymph node metastases, and they do not appearto benefit from SNB.

Classification tree analysis and logistic modelingidentified a Breslow cutpoint of approximately 1.2 mmto be a highly significant decision cutpoint for predic-tion of SNB result. With a sensitivity of 0.93, this de-cision cutpoint has a sensitivity similar to that re-ported for the SNB procedure itself.12 It is emphasizedthat this cutpoint is derived from retrospective dataand is influenced by our prior selection criteria forperforming SNB. The ideal selection criteria for SNBare not known, and therefore this and other cutpointsare somewhat arbitrary. The cutpoint of 1.2 mm maybe useful as an initial guideline to counsel patientsabout their likelihood of having a positive SNB. How-ever, caution is advised in application of this or anycutpoint as rigid selection criterion to perform SNB.Tumor thickness is a continuous variable associatedwith a gradually increasing level of risk for metastaticdisease. A minority of patients with melanomas lessthan 1.2 mm may have occult lymph node metastases,and some patients will be missed if SNB is not per-formed. It is probably prudent to use multiple selec-tion criteria in patients with thin melanomas.

Of the other traditional melanoma prognostic in-dicators, high mitotic index and ulceration were foundto be significant independent predictors of SNB result.This is not surprising because mitotic index and tumorulceration have been shown to be important prognos-tic factors for melanoma recurrence and progno-sis.5,7,23–28 Ulceration and mitotic index are probablybest considered secondary predictors of SNB resultbecause they demonstrate variable importance de-pending on how tumor thickness is modeled. How-ever, this finding suggests patients with ulcerated tu-mors and those with high mitotic index lesions are athigher risk for lymph node metastases and should beconsidered for SNB regardless of the depth of theirprimary tumors.

SLND yielded additional tumor-containing resid-ual lymph nodes in 28% of patients in this series. Thepositivity rate for SLND varies from 7% to 42% invarious series,12,13,16 –18,29 but is somewhat variable be-tween reports. In a recent report of 70 patients withmelanomas draining to the head and neck lymphnode areas, we noted positive SLND lymph nodes in42% of SNB positive patients.29 This and other reportsprobably underestimate the true SLND positivity ratebecause analysis of completion SLND specimens wasnot as rigorous as for SNB specimens. Standard patho-logic analysis identifies only approximately 70% ofmelanoma metastases in sentinel lymph nodes.18

FIGURE 3. Overall survival by sentinel lymph node biopsy result (median

follow-up, 12 months) is shown.

FIGURE 2. Disease free survival by sentinel lymph node biopsy result

(median follow-up, 12 months) is shown.


Therefore, the true incidence of residual tumor afterpositive SNB is unknown.

The relatively low incidence of positive lymphnodes found at SLND has led to speculation that SNBalone might be sufficient therapy for some patients,such as those without thick or ulcerated melanomas.18

It is known that in patients with positive SNB, themajority of tumor will be located in the SN(s). In arecent report on lymph node tumor volumes in pa-tients undergoing SNB/SLND, a weak positive corre-lation between tumor thickness and the volume oftumor in positive sentinel lymph nodes was demon-strated.17 This finding may have been due to severalpatients with relatively large tumor burdens. A similarrelation was not evident across AJCC tumor stages.The presence of multiple tumor deposits in sentinellymph nodes and the presence of multiple positivesentinel lymph nodes did not correlate with the pres-ence of residual lymph node disease at SLND. Thesefindings suggest that neither tumor thickness nor thepattern of lymph node metastases within sentinellymph nodes are exclusive predictors of the histologyof the residual basin after a positive SNB.

The current report provides further evidence thatprimary tumor characteristics are not useful predic-tors of the histology of the residual lymph node basinafter a positive SNB. A relation between primary mel-anoma tumor characteristics and the histology oflymph nodes remaining in the SNB positive basins isnot demonstrated. Although no SN positive patientwith primary tumor less than 1.3 mm thick had apositive SLND, the small number of patients (7) in thissubset prevents meaningful conclusions. Similarlythese data did not show ulceration to be predictive ofSLND histology after positive SNB (24% vs. 40% ulcer-ation in SLND negative vs. SLND positive patients,respectively). Collectively, these observations raiseskepticism about our current ability to identify pa-tients at nil risk for positive SLND. Detailed investiga-tions into the pattern(s) of lymph node metastasesand longer follow-up of much larger numbers of pa-tients are necessary before this approach can be rec-ommended.

A relation between primary melanoma tumor his-topathology and the histology of the residual lymphnode basin after positive SNB would not necessarily beexpected. Although it is possible that deep or ulcer-ated tumors are capable of producing earlier or morenumerous lymph node metastases that could result inspread to second-echelon lymph nodes, this is spec-ulative and possibly incorrect. Metastatic tumor vol-ume is more likely related to the length of time thattumor is present within the sentinel lymph node andthe rate of tumor growth. This function is probably

inconstant and determined by a complex interactionof ill-defined tumor and host factors. The finding ofresidual lymph node disease after SNB also could beaccounted for by other explanations. Progressive tu-mor growth within lymph nodes or lymphatic chan-nels may occlude efferent lymphatics, altering ormasking patterns of second-echelon metastases. Fur-thermore, tumor spillage from severed lymphatic ves-sels could result in extranodal tumor seeding andgrowth within the operative wound. The SNB proce-dure alone probably would not be an adequate onco-logic operation in these instances.

Residual tumor-containing lymph nodes may re-main in the residual basin after positive SNB in up to28% of patients and can be present across a widerange of primary tumor characteristics. A populationof patients highly likely to have tumor confined to theSNB specimen is not identified by major tumor prog-nostic factors. Data from recent prospective clinicaltrials suggest occult melanoma lymph node microme-tastases are clinically important and that early thera-peutic LND may favorably impact survival.10,11 Thus,SLND is potentially therapeutic for a relatively largesubset of patients. Because of the difficulty predictingthat patients have residual positive lymph nodes aftera positive SNB, SLND is the most appropriate therapy.The impact of SNB-directed therapy is currently un-known and is the focus of several ongoing clinicalstudies.

Several weaknesses of this study are apparent.First, this is a retrospective analysis, and results areunavoidably influenced by our previous selection cri-teria for performing the SNB procedure. Possible pa-tient selection bias cannot be eliminated. The rela-tively small number of thin melanomas does notpermit detailed analysis of predictive factors for thissubset. It is at least possible that a population of SNBpositive patients with thin tumors may have diseaseconfined to the sentinel lymph node(s). The his-topathologic data in this series were not prospectivelycollected for this analysis and therefore are vulnerableto subjectivity and inconsistency. These data do notinclude analysis of primary tumor microsatellitosis orangiolymphatic invasion, factors others have associ-ated with worse prognosis and regional lymph nodemetastases.23,24,30 Last, this series does not includepatients that were observed or treated with adjuvanttherapy without SLND after a positive SNB. The nat-ural history of patients treated in this manner is cur-rently unknown. A large, prospective randomizedstudy with a predefined study population stratified formajor known prognostic factors with protocol-drivencontrol of preoperative, operative, and pathology pro-cedures and prolonged follow-up would be required


to achieve statistically significant subset analyses andto eliminate internal biases.

Although the median follow-up duration in thisseries is too short to provide definitive information onpatterns of recurrence or survival, this report confirmsSNB yields clinical information that is useful for pre-dicting early treatment failure. It is noteworthy thatthe SNB positivity rates for the various AJCC tumorstages in this series are similar to the expected 5-yearmortality rates for the corresponding stage groups.21

The outcome measures of survival and recurrence at-tributable to SNB-directed care are of primary impor-tance and ultimately will determine the future of SNBas a clinically useful tool or a staging procedure ofmainly historical interest.

CONCLUSIONSSeveral primary melanoma tumor characteristics areassociated with SNB result. Tumor thickness is themost significant and useful predictor of SNB resultand should be the primary factor considered in patientselection for this diagnostic procedure. A Breslow tu-mor thickness of approximately 1.2 mm is proposed asa useful initial guideline to counsel patients regardingSNB. Patients with ulcerated tumors and tumors witha high mitotic index are at higher risk for occult lymphnode metastases and should be considered for SNBregardless of thickness. SNB alone should not be con-sidered an oncologically adequate procedure in mel-anoma because additional tumor-containing lymphnodes remain in the lymph node basin in up to 28% ofpatients who have a positive SNB. Primary tumorcharacteristics do not appear to be predictive of thehistology of the residual lymph node basin after apositive SNB. Completion lymphadenectomy is rec-ommended for this reason. SNB provides importantprognostic information, correlating with disease freeand overall survival.

REFERENCES1. Parker SL, Tona T, Bolden S, Wingo PA. Cancer statistics:

1996. CA Cancer J Clin 1996;46:5–27.2. Armstrong BK, Kricker A. Cutaneous melanoma: trends in

cancer incidence and mortality. Cancer Surv 1994;19:219 –40.

3. MacLennan R, Green AC, McLeod GR, Martin NG. Increas-ing incidence of cutaneous melanoma in Queensland, Aus-tralia. J Natl Cancer Inst 1992;84:1427–31.

4. Breslow A. Thickness, cross-sectional areas and depth ofinvasion in the prognosis of cutaneous melanoma. Ann Surg1970;172:1902– 8.

5. Balch CM, Murad TM, Soong SJ, Murad TM, Ingalis AL,Maddox WA. A multifactorial analysis of melanoma II: prog-nostic factors in patients with stage I (localized) melanoma.Surgery 1979;86:343–51.

6. Balch CM, Murad TM, Soong SJ, Ingalis AL, Richards PC,

Maddox WA. Tumor thickness as a guide to surgical man-agement of clinical stage I melanoma patients. Cancer 1979;43:883– 8.

7. Stadelmann WK, Rapaport DP, Soong SJ, Reintgen DS,Buzaid AC, Balch CM. Prognostic clinical and pathologicfeatures. In: Balch CM, Houghton AN, Sober AJ, Soong SJ,editors. Cutaneous melanoma. 3rd edition. St. Louis, MO:Quality Medical Publishing, Inc., 1998:11–35.

8. Sim FH, Taylor WF, Pritchard DJ, Soule EH. Lymphadenec-tomy in the management of stage I malignant melanoma: aprospective randomized study. Mayo Clin Proc 1986;61:697–705.

9. Veronesi U, Adamus J, Bandiera DC, Brennhovd IO, CaceresE, Cascinelli N, et al. Stage I melanoma of the limbs. Imme-diate versus delayed lymph node dissection. Tumori 1980;66:373–96.

10. Balch CM, Soong SJ, Bartolucci AA, Urist MM, KarakousisCP, Smith TJ, et al. Efficacy of an elective regional lymphnode dissection of 1 to 4 mm thick melanomas for pa-tients 60 years of age and younger. Ann Surg 1996;224:255– 63.

11. Cascinelli N, Morabito A, Santinami M, MacKie RM, Belli F.Immediate or delayed dissection of regional nodes in pa-tients with melanoma of the trunk: a randomised trial. WHOmelanoma programme. Lancet 1998;351:793– 6.

12. Morton DL, Wen DR, Wong JH, Economou S, Cagle LA,Storm FK, et al. Technical details of intraoperative lym-phatic mapping for early stage melanoma. Arch Surg 1992;127:392–9.

13. Reintgen D, Cruse CW, Wells K, Berman C, Fenske N, GlassF, et al. The orderly progression of melanoma nodal metas-tases. Ann Surg 1994;220:759 – 67.

14. Morton DL, Wen DR, Foshag LJ, Essner R, Cochran A. In-traoperative lymphatic mapping and selective cervicallymphadenectomy for early stage melanomas of the headand neck. J Clin Oncol 1993;11:1751– 6.

15. Albertini JJ, Cruse CW, Rapaport D, Wells K, Ross M, De-Conti R, et al. Intraoperative radiolymphoscintigraphy im-proves sentinel lymph node identification for patients withmelanoma. Ann Surg 1996;223:217–24.

16. Krag DL, Meijer SJ, Weaver DL, Loggie BW, Harlow SP,Tanabe KK, et al. Minimal-access surgery for staging ofmalignant melanoma. Arch Surg 1995;130:654 – 8.

17. Wagner JD, Davidson D, Coleman JJ III, Hutchins G, Schau-wecker D, Park HM, et al. Lymph node tumor volumes inpatients undergoing sentinel lymph node biopsy for cuta-neous melanoma. Ann Surg Oncol 1999;6:398 – 404.

18. Joseph E, Brobeil A, Glass F, Glass J, Messina J, DeConti R,et al. Results of complete lymph node dissection in 83patients with positive sentinel nodes. Ann Surg Oncol1998;5:119 –25.

19. Gershenwald JE, Thompson W, Mansfield PF, Lee JE, Co-lome MI, Tseng CH, et al. Multi-institutional melanomalymphatic mapping experience: the prognostic value of sen-tinel lymph node status in 612 stage I or II melanomapatients. J Clin Oncol 1999;17:976 – 83.

20. MacKie RM, Byrne D, Lyngam MK, Sajid M, MacKay AK.Prognostic evaluation of sentinel node biopsy. MelanomaRes 1997;7:S102.

21. Fleming ID, Cooper JS, Henson DE, Hutter RVP, Kennedy BJ,Murphy GP, et al. AJCC Cancer Staging Manual. 5th edition.Philadelphia: Lippincott-Raven, 1997:163–70.


22. Kirkwood JM, Strawderman MH, Ernstaff MS, Smith TJ,Bordern EC, Blum RH. Interferon alfa-2b adjuvant therapyof high risk resected cutaneous melanoma: the Eastern Co-operative Pncology Group Trial EST 1684. J Clin Oncol 1996;14:7–17.

23. Callery C, Cochran AJ, Roe DJ, Rees W, Nathanson SD,Benedetti JK, et al. Factors prognostic for survival in patientswith malignant melanoma spread in regional lymph nodes.Ann Surg 1981;196:69 –75.

24. Mraz-Gernhard S, Sagebiel RW, Kashani-Sabet M, Miller JRIII, Leong SP. Prediction of sentinel lymph node microme-tastasis by histological features in primary cutaneous ma-lignant melanoma. Arch Dermatol 1998;134:983–7.

25. Balch CM, Soong SJ, Shaw HM, Urist MM, McCarthy WH. Ananalysis of prognostic factors in 8500 patients with cutane-ous melanoma. In: Balch CM, Houghton AN, Milton GW,Sober AJ, Soong SJ, editors. Cutaneous melanoma. 2nd edi-tion. Philadelphia: JB Lippincott, 1992:165– 87.

26. Soong SJ, Weiss HL. Predicting outcome in patients with

localized melanoma. In: Balch CM, Houghton AN, SoberAJ, Soong SJ, editors. Cutaneous melanoma. 3rd edition.St. Louis, MO: Quality Medical Publishing, Inc., 1998:11–35.

27. Pontikes LA, Temple WJ, Cassar SL, Lafreniere R, HuchcroftSA, Jerry LM, et al. Influence of level and depth on recur-rence rate in thin melanomas. Am J Surg 1993;165:225– 8.

28. Haffner AC, Garbe C, Burg G, Buttner P, Orfanos CE, RassnerG. The prognosis of primary and metastasizing melanoma:an evaluation of the TNM classification in 2,495 patients.Br J Cancer 1992;66:856 – 61.

29. Wagner JD, Park HM, Coleman JJ III, Love C, Hayes JT.Cervical sentinel lymph node biopsy for melanomas of thehead and neck and upper thorax. Arch Otolaryngol HeadNeck Surg. 2000;126:313–21.

30. Leon P, Daly JM, Synnestvedt M, Schultz DJ, Elder DE, ClarkWH Jr. The prognostic implications of microscopic satellitesin patients with clinical stage I melanoma. Arch Surg 1991;26:1461– 8.


predicting sentinel and residual lymph node basin disease after sentinel lymph node biopsy for...

Documents