comparison of classification systems in melanoma sentinel lymph nodes—an analysis of 697 patients...

Comparison of Classification Systems inMelanoma Sentinel Lymph Nodes—AnAnalysis of 697 Patients From a SingleCenterAndre Meier, MD; Imke Satzger, MD; Bernward Volker, MD; Alexander Kapp, MD; and Ralf Gutzmer, MD

BACKGROUND: In melanoma, different classification systems have been proposed that predict overall survival (OS) and

recurrence-free survival (RFS) based on findings in the sentinel lymph node (SLN). The authors of this report compared

the RFS and OS of 697 melanoma patients as predicted by various classification systems. METHODS: The Rotterdam

system (based on the greatest dimension of the largest tumor cell deposit), the Augsburg S-classification (based on

tumor penetrative depth [TPD]), and the Hannover system (based on a combination of tumor load, TPD, and invasion

of the capsule) were studied in 697 consecutive melanoma patients who underwent SLN biopsy at the authors’ cen-

ter. RESULTS: In univariate analyses, the Rotterdam and Hannover systems (but not the S-classification) identified 1

group of SLN-positive patients that had OS and RFS similar to the OS and RFS of SLN-negative patients. The inter-

mediate groups from all classification systems did not differ significantly with regard to RFS and/or OS from the ad-

jacent groups. In multivariate analysis using a Cox model, the greatest dimension of the largest tumor cell deposit

(cutoff point, <0.1 mm vs �0.1 mm), the TPD (cutoff point, �2 mm vs >2 mm), and capsular involvement represented

independent parameters for RFS; and TPD and capsular involvement also were independent parameters for OS. On

the basis of these 3 parameters, a new scoring system for risk assessment in patients with melanoma is proposed

that can distinguish 3 separate groups of patients that differed significantly in OS and RFS. CONCLUSIONS: Different

parameters of independent prognostic significance were identified in SLNs from patients with melanoma. Combining

these parameters, the prognosis of patients with melanoma was predicted more precisely by the new scoring system

than by currently published classification systems. Cancer 2010;116:3178–88. VC 2010 American Cancer Society.

KEYWORDS: histology, melanoma, prognosis, sentinel lymph node.

Sentinel lymph node (SLN) biopsy has been an established procedure in melanoma since the early 1990s.1 The statusof the SLN is the most important predictor of survival, as indicated in many studies,2-8 and, thus, is included in the currentAmerican Joint Committee on Cancer (AJCC) staging system for melanoma.9 In the current AJCC classification, numberof positive SLNs is used as a parameter for SLN involvement; ie, 1 positive SLN is classified N1a, and 2 or 3 positive SLNsare classified as N2a.9

However, the histopathologic pattern of SLN involvement is heterogeneous, ranging from single, isolated melanomacells to the complete replacement of lymph node tissue by melanoma cells. To predict the prognosis of patients more pre-cisely, different parameters and risk-assessment systems of SLN involvement have been developed in recent years. TheseSLN parameters reflect mainly the amount and pattern of melanoma involvement, such as the depth and invasion of mela-noma cells within the lymph node parenchyma (tumor penetrative depth [TPD]),10 the expansion of melanoma cells intothe lymph node capsule (capsular involvement),6 and the greatest dimension (diameter) of the largest tumor cell deposit(tumor burden). van Akkooi et al published a risk-assessment system (called the Rotterdam system) based on the greatestdimension of the tumor cell deposit. Three different groups were defined with increasing deposit size and decreasing sur-vival rates.11,12 Patients who had metastases in an SLN that measured <0.1 mm had significantly better overall survival

DOI: 10.1002/cncr.25074, Received: July 22, 2009; Revised: September 22, 2009; Accepted: September 23, 2009, Published online April 19, 2010 in Wiley Inter-

Science (www.interscience.wiley.com)

Corresponding author: Imke Satzger, MD, Skin Cancer Centre Hannover, Department of Dermatology and Allergy, Hannover Medical School, Ricklinger Strasse

5, D-30449 Hannover, Germany; Fax: (011) 49-511-9246-422; [email protected]

The first 2 authors contributed equally to this work.

Department of Dermatology and Allergy, Skin Cancer Center Hannover, Hannover Medical School, Hannover, Germany.

3178 Cancer July 1, 2010

Original Article

(OS) at 5 years than patients who had micrometastasesthat measured from 0.1 mm to 1.0 mm or >1.0 mm.Starz et al developed the so-called new S-classificationbased on the TPD.10 Patients who had a TPD <0.3 mm(S-I) reportedly had a significantly better prognosis thanpatients who had a TPD between 0.3 mm and 1 mm (S-II) and patients who had a TPD >1 mm (S-III). Finally,we established the Hannover system to differentiatebetween patients who have a favorable prognosis andpatients who have a poor prognosis by combining 3 pa-rameters of tumor load (<30 cells vs �30 cells), TPD(�2 mm vs >2 mm), and capsular invasion (absent vspresent).6 Each parameter was assigned 1 point, and, themore points a patient had, the worse their prognosis.

Theoretically, risk-assessment systems need to fulfill2 criteria: 1) maximizing the variability of outcomesamong the different groups of patients and 2) minimizingthe variability of outcomes among 1 group of patients.13

To date, unique risk-assessment systems have been ana-lyzed in unique populations, but the discriminative powerof different risk-assessment systems for predicting progno-sis in patients with melanoma has not been compared.The objective of the current study was to compare theprognostic value of different histopathologic and clinicalparameters as well as risk-assessment systems by applyingthem to the same group of patients.

MATERIALS AND METHODSWe studied a population of 697 consecutive patients whohad primary cutaneous melanoma (Breslow tumor thick-ness�1 mm) diagnosed in our department between April2000 and December 2006. Primary melanomas wereexcised with a 2-cm safety margin. At the time of diagno-sis, patients were without clinical or radiologic evidence ofmetastasis (initial staging was performed by chest x-ray,abdominal ultrasound, assessment of peripheral lymphnodes, and measuring S-100 tumor marker levels in se-rum). All patients underwent SLN biopsy as describedpreviously.14 In 217 patients (31%), melanoma cells inthe SLN were detected by hematoxylin and eosin (H&E)staining or immunohistochemistry (see below) and wereconsidered micrometastases. Thus, micrometastases weredefined as melanoma cells in the SLN identified by histo-pathology or immunohistochemistry that could not bedetected by clinical or radiologic investigation.

One hundred seventy-five patients underwent com-pletion lymph node dissection (CLND) after a positiveSLN biopsy, and the CLND specimen was positive in 27

of 175 patients (15.4%). Eight patients refused toundergo CLND, 12 patients did not undergo CLND formedical reasons that prohibited a major procedure undergeneral anesthesia, and CLND was not recommended in22 patients because of low tumor burden. Follow-upinvestigations were performed according to the recom-mendations of the German Dermatologic Society15 andincluded a clinical examination every 3 months and achest x-ray and ultrasound studies of lymph nodes and theabdomen every 6 months. The mean follow-up was 49.5months, and the median follow-up was 47.8 months(range, 0.1-110.1 months). One hundred forty-threepatients (21%) had recurrent melanoma, and 95 patients(14%) died of causes related to melanoma.

Histopathologic and ImmunohistochemicalStaining: Workup of SLN

Lymph nodes were fixed in formaldehyde, embedded inparaffin, and processed as described previously14 accord-ing to the recommendation of Cochran et al.16 Depend-ing on the size of the lymph node, between 1 and 8sections were cut, because it has been demonstrated thatthe examination of multiple sections increases sensitivityfor detecting micrometastases.

Each section was processed cut-face down for rou-tine paraffin embedding using automated devices (Path-center and Histocenter 640CX51; Shandon, Astmoor,United Kingdom), and sections were cut for H&E stain-ing and immunohistochemistry following the recommen-dation of Cochran et al.16 For all patients, the sameimmunohistochemical procedures were used. Highly sen-sitive immunohistochemistry for HMB45 (cloneHMB45; 1:60 dilution; DakoCytomation, Hamburg,Germany) and Melan A (clone A103; dilution 1:300;DakoCytomation) was performed with an automatedimmunostainer (Horizon; DakoCytomation) using anindirect streptavidin-biotin method following the stand-ardized methylation-sensitive amplified polymorphismprotocol provided by the manufacturer. Pretreatment forantigen retrieval was performed by gentle cooking over-night at 70�C in Target Retrieval Solution (DakoCyto-mation) for Melan A and by proteinase K for HMB45.Negative controls were performed using isotype-matched,irrelevant antibodies of the same species. Newfuchsin wasused as chromogen, because its bright-red color allowseasy discrimination from brown melanin pigment in mel-anocytic cells and macrophages and from other endoge-nous pigments like siderin.

Sentinel LN and Prognosis in Melanoma/Meier et al

Cancer July 1, 2010 3179

Histopathologic Parameters Assessed in theSLN

Melanoma cells were detected in 1 SLN in 166 of217 patients (76.5%) and in more than 1 SLN in 51 of217 patients (23.5%). In the latter patients, the lymphnode with the highest tumor burden was used for furtheranalysis.

Parameters that we assessed in the SLNs can begrouped in 4 categories: First, tumor burden (ie, theamount of melanoma cells) was assessed in different ways:1) the size of the largest tumor cell deposit was estimatedsemiquantitatively by counting the number of cells indirect contact with one another (isolated cells, smalldeposits of <30 cells, large deposits of �30 cells); 2) thegreatest dimension of the largest tumor cell deposit wasmeasured with an ocular micrometer and grouped into4 categories (a greatest dimension<0.1 mm, 0.1-1.0 mm,>1.0 mm, and �0.2 mm) according to van Akkooi et aland Scheri et al11,17; or 3) the number of positive SLNs (1vs �1). Second, the peripheral expansion of tumor cellswas assessed (ie, the presence of tumor cells within thecapsule of the SLN denoted capsular involvement). Third,the central expansion of tumor cells was assessed (ie, thedepth of invasion was measured according to Starz et alwith the ocular micrometer as the maximal distancebetween the inner layer of the lymph node capsule or tra-becule and melanoma cells in the lymph node paren-chyma10). The term ‘‘tumor penetrative depth’’ (TPD)was used to designate this distance according to the pro-posal of Scolyer and coworkers.7

For statistical analyses, chi-square tests and Kaplan-Meier tests were performed with the software SPSS (ver-sion 13.0; SPSS Inc; Chicago, Ill). For multivariate analy-ses, a Cox model was used and included parameters thatwere significant in univariate analyses and were describedin 1 of the established risk-assessment systems (Rotterdamsystem, S-classification, Hannover system) or parametersthat were established in previous studies (TPD >2 mm[Scolyer et al7]; isolated, immunohistochemically positivecells [Satzger et al18]; a greatest dimension �0.2 mm ofthe largest tumor cell deposit [Scheri et al17]; and thenumber of positive SLNs [Gershenwald et al19]). The fol-lowing histopathologic parameters of the SLN were ana-lyzed in different combinations: capsular involvement,TPD (with cutoff points of 0.3 mm, 1 mm, and 2 mm),tumor load (with cutoff points of <30 cells vs �30 cells),the number of positive SLNs (1 vs �1), and the greatestdimension of the largest tumor cell deposit (with cutoffpoints of 0.1 mm, 0.2 mm. and 1 mm). P values and con-

fidence intervals were calculated. P values <.05 were con-sidered statistically significant.

RESULTS

Characterization of Patients

Six hundred ninety-seven consecutive patients with pri-mary melanoma (Breslow thickness �1 mm) werereviewed who underwent SLN biopsy in our departmentbetween April 2000 and December 2006. The SLN waspositive in 217 of 697 patients (31.1%).

In univariate analyses, we compared patients whohad micrometastases in SLNs with patients who had his-topathologically and immunohistochemically negativeSLNs. Patient characteristics with regard to clinic, pri-mary melanoma, and the number of positive SLNs arelisted in Table 1. The mean follow-up was 49.5 months(median, 47.8 months; range, 0.1-110.1 months).

There was a predominance men among the patientswho had positive SLNs (55.8%) compared with thepatients who had negative SLNs (49.8%). The mean agesof the 2 groups did not differ significantly. Primary mela-nomas with positive SLNs, as expected, were significantlythicker, had higher Clark levels, and were ulcerated moreoften. The primary disease sites were head and neck(12.7% for SLN-negative patients and 6.4% for patientswith micrometastases), trunk (33.8% for SLN-negativepatients and 41.5% for patients with micrometastases),and extremities (53.5% for SLN-negative patients and52.1% for patients with micrometastases).

In SLN-negative patients, 63 of 480 (13.1%) devel-oped recurrent disease, and 41 of 480 (8.5%) died frommelanoma-related causes; whereas, in the patients whohad micrometastases, 80 of 217 (36.9%) developed recur-rent disease, and 54 of 217 (24.9%) died of melanoma-related causes. In SLN-positive patients who did notundergo CLND, 10 of 42 (23.8%) developed recurrentdisease compared with 70 of 105 (40%) SLN-positivepatients who underwent CLND.

Univariate Analyses of Histopathologic SLNParameters and Prognosis

In univariate analyses, several SLN parameters werecorrelated with recurrence-free survival (RFS) and OS.Different cutoff points were applied, as suggested in theliterature and mentioned above (see Materials andMeth-ods). Increasing tumor burden was associated with apoorer prognosis, as assessed by a semiquantitative evalu-ation of the number of melanoma cells in the largest

Original Article


Table 1. Characterization of 697 Patients With Melanoma Patients Who Underwent Sentinel LymphNode Evaluation

No. of Patients (%)

Characteristic Negative SLN Positive SLN P

No. of patients 480 (68.9) 217 (31.1)

Age, y .128

Median 62 62

Mean 59 57

Range 19-90 18-91

Sex .008

Women 241 (50.2) 96 (44.2)

Men 239 (49.8) 121 (55.8)

Breslow thickness, mm <.001

Median 1.70 2.50

Mean 2.25 3.35

Range 1-18 1-17

1-2 321 (66.9) 84 (38.7) <.001

>2-4 109 (22.7) 72 (33.2)

>4 50 (10.4) 61 (28.1)

Clark level <.001

III 224 (46.7) 60 (27.6)

IV 210 (43.8) 113 (52.1)

V 46 (9.6) 44 (20.3)

Ulceration <.001

No 387 (80.6) 134 (61.8)

Yes 93 (19.4) 80 (36.9)

Regression .007

No 387 (80.6) 187 (86.2)

Yes 93 (19.4) 30 (13.8)

Location .003

Head 61 (12.7) 14 (6.4)

Trunk 162 (33.8) 90 (41.5)

Extremities 257 (53.5) 113 (52.1)

No. of SLNs .001

Median 1 3

Mean 1.6 1.9

Range 1-5 1-6

No. of positive SLNs NA

1 — 166 (76.5)

2 — 43 (19.8)

3 — 7 (3.2)

4 — 1 (0.5)

Recurrence <.001

No 417 (86.9) 137 (63.1)

Yes 63 (13.1) 80 (36.9)

Initial recurrence <.001

Locoregional cutaneous 20/63 (31.7) 29/80 (36.2)

Regional lymph nodes 17/63 (27) 13/80 (16.3)

Distant metastases 26/63 (41.3) 38/80 (47.5)

Total distant metastases <.001

No 436 (90.8) 154 (71)

Yes 44 (9.2) 63 (29)

Death <.001

No 409 (85.2) 152 (70)

Yes, melanoma 41 (8.5) 54 (24.9)

Yes, other 30 (6.3) 11 (5.1)

SLN, sentinel lymph node; NA, not available.

cluster (<30 cells vs �30 cells), in terms of both OS(P < .001) and RFS (P < .001). In SLN-positivepatients, a greatest dimension of the largest cluster <0.1mm versus 0.1 mm to 1 mm versus>1.0 mm (accordingto the Rotterdam system) was associated significantlywith survival rates in terms of both OS (P < .001) andRFS (P < .001) compared with survival rates in SLN-negative patients. The number of positive SLNs achievedsignificance for RFS (P¼ .002) but not for OS(P¼ .136). With regard to the peripheral expansion oftumor cells into the SLN, capsular involvement was cor-related with poorer RFS (P < .001) and OS (P < .001).The central expansion of tumor cells measured as TPDwas correlated significantly with OS and RFS with cutoffpoints of �2 mm versus >2 mm (P < .001 and P <

.001, respectively) and, as measured according to theAugsburg S-classification, with cutoff points of <0.3mm versus 0.3 mm to 1.0 mm versus >1.0 mm(P¼ .006 and P< .001, respectively). Finally the Hann-over system was able to discriminate between groups ofpatients (0 points vs 1 point vs 2 points vs 3 points) thathad significant differences in OS (P < .001) and RFS(P< .001).

Kaplan-Meier Analyses

Rotterdam system

In Kaplan-Meier analyses, the Rotterdam criteria fortumor burden were highly significant predictors of RFS(P < .001) and OS (P < .001) in SLN-positive patientscompared with SLN-negative patients (Fig. 1a,b). Theseparation of Kaplan-Meier curves for OS (Rotterdam cri-teria: 0.1-1 mm vs >1 mm) did not achieve significance(P¼ .103) (Table 2). Patients who had micrometastases<0.1 mm (greatest dimension of the largest tumor celldeposit) had favorable survival rates, similar to those forSLN-negative patients, in terms of both OS (P¼ .400)and RFS (P¼ .183).

S-classification

The central expansion of tumor cells measured asTPD was correlated significantly with OS (P¼ .006) andRFS (P < .001) with different cutoff points according tothe new S-classification (Fig. 1c,d). Patients who had aTPD <0.3 mm had a significantly poorer RFS thanpatients who had negative SLNs (P¼ .019), but the OSwas similar (P¼ .261). Kaplan-Meier curves for adjacentgroups of SLN-positive patients (TPD<0.3 mm vs 0.3-1mm and TPD 0.3-1 mm vs>1 mm, respectively) did notdiverge significantly in terms of OS or RFS (Table 2).

Hannover system

The Hannover classification with a combination ofdifferent parameters (tumor load, TPD, and capsularinvasion) could achieve significance in Kaplan-Meieranalyses for RFS (P < .001) or OS (P < .001) (Fig. 1e,f).Separate Kaplan-Meier curves (Hannover system: 1 pointvs 2 points and 2 points vs 3 points) did not diverge signif-icantly for OS or RFS (Table 2). Patients who had micro-metastases in an SLN but no additional positiveparameter (tumor load �30 cells, TPD >2 mm, capsularinvasion) had favorable survival rates, similar to the ratesfor SLN-negative patients, in terms of both OS(P¼ .520) and RFS (P¼ .119).

Multivariate Correlation of HistopathologicParameters of the SLN and Prognosis

To determine independent prognostic parameters of theSLN, we performed a backwards, stepwise Cox multivari-ate analysis that included all parameters and all cutoffpoints that were included in 1 of the published risk-assess-ment systems (Rotterdam system, S-classification, Hann-over system) or that have been established in previousstudies (isolated, immunohistochemically positive cells vscell clusters [Satzger et al18]; a greatest dimension of thelargest tumor cell deposit �0.2 mm vs >0.2 mm [Scheriet al17]; and the number of positive SLN [Gershenwald etal19]). These analyses revealed 3 parameters that had inde-pendent prognostic significance: ie, capsular involvement(present vs absent), TPD with a cutoff point at 2 mm (�2mm vs>2 mm), and the greatest dimension of the largesttumor cell deposit with cutoff point at 0.1 mm (<0.1 mmvs�0.1 mm) (Table 3).

Proposal for an ImprovedRisk-Assessment System

We developed a new risk-assessment system that differ-entiates significantly between all groups of patients butis as convenient as possible. This system was based onthe parameters that were associated significantly andindependently with prognosis (Table 3): ie, the greatestdimension of the largest tumor cell deposit (cutoffpoint, <0.1 mm/�0.1 mm), TPD (�2 mm/>2 mm),and capsular involvement (absent/present). It appearsto be sufficient to consider patients with deposits<0.1 mm as a single group. This group has a progno-sis similar to that of SLN-negative patients, and addi-tional parameters (TPD and capsular involvement) donot add prognostic information for these patients. Ifwe focus on the 85 patients who had micrometastases

Original Article


Figure 1. The Kaplan-Meier was used to analyze overall survival (OS) and recurrence-free survival (RFS) according to (a,b) theRotterdam system, (c,d) the Augsburg S-classification, (e,f) the Hannover system, and (g,h) the Hannover-II system. SLN indi-cates sentinel lymph node.



Table

2.Comparisonofth

eDiscriminativePowerofDifferentRisk-A

ssessmentSystems(theRotterd

am

System,th

eAugsburg

S-C

lassification,th

eHannoverSystem,

and

theHannover-IISystem)W

ithPValuesCalculatedfo

rSeparate

Gro

ups

P

OS

RFS

Scoring

System

Parameters/and

Cutoffs

No.of

Patients

SLN-N

egative

vsGroup1

Group1

vsGroup2

Group2

vsGroup3

Group3

vsGroup4

SLN-N

egative

vsGroup1

Group1

vsGroup2

Group2

vsGroup3

Group3

vsGroup4

Rotterdam

system

(basedongreatest

dim

ensionofdeposit)

Group1

<0.1

mm

85

.400

.016

.016

NA

.183

.001

.103

NA

Group2

0.1-1.0

mm

77

Group3

>1.0

mm

55

Augsburg

S-classification(basedontumor

penetrativedepth),mm

Group1

<0.3

mm

65

.261

.161

.073

NA

.019

.513

.003

NA

Group2

0.3-1.0

mm

67

Group3

>1.0

mm

85

Hannoversystem

(basedontumorpenetrative

depth,>30cells,capsularinvolvement)

Group1

0Points

77

.520

.252

.068

.001

.119

.077

.189

<.001

Group2

1Points

61

Group3

2Points

53

Group4

3Points

26

Hannover-IIsystem

(basedontumorpenetrative

depth,greatestdim

ensionofdeposit<0.1

mm,andcapsularinvolvement)

Group1

<0.1

mm

85

.400

.012

<.001

NA

.183

.001

<.001

NA

Group2

�0.1

mm

and0or

1points

105

Group3

�0.1

mm

and2

points

27

OSindicatesoverallsurvival;RFS,recurrence-freesurvival;SLN,sentinellymphnode;NA,notapplicable.

Original Article


in the SLN and a greatest dimension of the largest tu-mor cell deposit <0.1 mm, in 15 of those patients, anadditional parameter (TPD >2 mm or capsularinvolvement) was detected. We compared the RFS andOS rates of patients who had micrometastases meas-uring <0.1 mm with the survival rates of patients whohad micrometastases <0.1 mm and 1 additional pa-rameter from the Hannover classification (TPD or cap-sular involvement), and neither group differedsignificantly (P¼ .412 and P¼ .137, respectively).Therefore, patients who had a greatest dimension ofthe largest tumor cell deposit <0.1 mm were consid-ered as a single group (Hannover-II, Group 1).Patients who had micrometastases �0.1 mm and, atmost, 1 additional point were classified into Hannover-II, Group 2. Patients who had micrometastases �0.1mm and 2 additional points were classified into Hann-over-II, Group 3 (Table 4). In the Hannover-II classifi-cation, all separate groups (Group 1 vs Group 2 andGroup 2 vs Group 3) diverged significantly in terms of

OS and RFS (Fig. 1g,h; Table 2). Patients with a neg-ative SLN and patients with micrometastases <0.1 mm(OS, P¼ .542; RFS, P¼ .507) had similar survivalrates, and all other separated groups of patientsdiverged significantly.

DISCUSSIONWe compared histopathologic parameters and risk-assess-ment systems that predict prognosis in our group of 697melanoma patients with known clinical course. SLNswere obtained in our center and analyzed by the samestandardized histopathologic workup and evaluation.6

All risk-assessment systems were able to separategroups of patients with excellent and poor prognoses.However, the discriminative power of the various systemsdiffered considerably.

In the current study, we confirmed the prognosticrelevance of a greatest dimension of 0.1 mm for the largesttumor cell deposit, as described previously by van Akkooi

Table 3. The Prognostic Significance of Histopathologic Parameters in Sentinel Lymph Nodes Analyzed by Backwards, StepwiseCox Multivariate Analysis

OS RFS

Parameter/Cutoff P HR 95%CI P HR 95%CI

Greatest dimension of deposit, mm<0.1 vs ‡0.1 .051 0.441 0.194-1.003 .006 0.400 0.208-0.772

<0.2 vs ‡0.2 .911 1.059 0.389-2.883 .543 0.752 0.300-1.885

<1 vs ‡1 .995 1.003 0.427-2.351 .477 1.261 0.666-2.386

IPC: Yes vs no .907 0.906 0.175-4.703 .388 1.693 0.513-5.588

No. of positive lymph nodes: 1 vs �1 .631 1.170 0.615-2.226 .152 0.702 0.432-1.139

Tumor load: <30 cells vs �30 cells .937 0.952 0.279-3.243 .222 1.671 0.733-3.810

TPD, mm£0.3 vs >0.3 .260 0.602 0.248-1.457 .450 0.773 0.397-1.506

£1 vs >1 .416 1,364 0.645-2.883 .884 0.951 0.487-1.859

£2 vs >2 .034 0.498 0.261-0.949 .001 0.401 0.234-0.685

Capsular involvement: Yes vs no .002 0.360 0.188-0.691 .034 0.594 0.367-0.960

OS indicates overall survival; HR, hazard ratio; CI, confidence interval; IPC, isolated immunohistochemically positive cells; TPD, tumor penetrative depth.

Table 4. An Improved Hannover Scoring System (Hannover-II) for Patients With Melanoma That Relies on a HistopathologicEvaluation of a Sentinel Lymph Node Based on 3 Criteria: Greatest Dimension of the Largest Deposit, Tumor Penetrative Depth,and Involvement of the Capsule

Group Criterion 1 Criterion 2 Remark

1 Largest deposit <0.1 mm Prognosis similar to patients with negative SLN

2 Largest deposit �0.1 mm Maximum of 1 additional parameter Intermediate prognosis

3 Largest deposit �0.1 mm Both TPD >2 mm and involvement of capsule Poor prognosis

SLN indicates sentinel lymph node; TPD, tumor penetrative depth.



et al.11,12 Patients who had deposits<0.1 mm had a prog-nosis similar to that of SLN-negative patients. This isremarkable considering the differences between the studyby van Akkooi et al and our study with regard to primarytumor parameters and SLN involvement. Whereas vanAkkooi et al indicated a median Breslow thickness of4 mm and 40 of their 388 patients (10%) had deposits<0.1 mm in an SLN, in our patients, the median Breslowthickness was 2.5 mm, and a greater proportion ofpatients had micrometastases <0.1 mm (85 of 217patients; 39%). Thus, the tumors in our patients were inan earlier stage of development. Furthermore, the cutoffpoint of 0.1 mm is consistent with our previous findingthat SLN-positive patients who have a very low tumorburden (isolated, immunohistochemically positive cells inthe SLN) have a prognosis similar to that of SLN-negativepatients.18 The 0.1-mm cutoff point (according to theRotterdam criteria11,12) was superior to the 0.2-mm cut-off proposed by Scheri et al and Govindarajan et al,17,20

which also agrees with our previous finding that patientswho have isolated, immunohistochemically positive cellshave a better prognosis than patients who have cell clustersthat measure up to 0.2 mm in greatest dimention.21

Our Kaplan-Meier analyses of separated groupsrevealed significant differences between almost all groupsaccording to the Rotterdam system. Only the comparisonof patients in Rotterdam Group 2 (0.1-1 mm) versuspatients in Rotterdam Group 3 (>1 mm) did not achievestatistical significance for OS; ie, the comparative powerof the Rotterdam system was high but, in a comparisonbetween adjacent groups of patients with higher tumorburden, it was suboptimal, and the cutoff point at 1 mmseems to be dispensable, because it does not add inde-pendent prognostic information.

The new S-classification was developed by Starzet al and records the penetration depth of tumor cellsin the lymph node.10 Patients with a TPD <0.3 mm(S-I) and a TPD between 0.3 mm and 1 mm (S-II)reportedly had a favorable prognosis, similar to SLN-negative patients. Survival rates for patients who had aTPD >1 mm (S-III) were significantly worse, althougha statistical comparison of single groups was not per-formed.10 Our Kaplan-Meier analyses achieved similarresults: Patients who had a TPD >1 mm had a signifi-cantly worse OS than patients who had a lower TPD(<0.3 mm and 0.3-1 mm). However, by comparingsingle groups, we were able to demonstrate that thesurvival rates for patients in adjacent groups, such asGroup 2 versus Group 3 for OS and Group 1 versus

Group 2 for RFS, did not diverge not significantly.Altogether, the discriminative power of the Rotterdamsystem appeared to exceed the discriminative power ofthe S-classification. One possible reason for these find-ings may be suboptimal cutoff points of the S-classifi-cation. Scolyer et al proposed a higher TPD limit of>2 mm to characterize patients with melanoma whohad a poor prognosis.7 Consistent with Scolyer et al,in our previous study, we demonstrated that a cutoffpoint for TPD at >2 mm was superior to a TPD cut-off point at >1 mm.6 Again, in the multivariate analy-ses from the current study, a TPD >2 mm (but not aTPD >1 mm) was identified as an independent prog-nostic parameter. Therefore, a TPD >2 mm was inte-grated into the Hannover system, which combines theparameters TPD >2 mm, capsular involvement, andtumor burden >30 cells. The Hannover system is ableto select SLN-positive patients who have an excellentprognosis, similar to the prognosis of SLN-negativepatients. However, adjacent groups (such as Group 2vs Group 3 for OS or Group 3 vs Group 4 for RFS)did not diverge significantly in the Hannover system.These findings may be related to the higher number ofgroups (in the Hannover system, there are 4 differentgroups compared with 3 groups in the Rotterdam sys-tem and S-classification). To enhance the discrimina-tive power of the Hannover system and to consider thedata from our multivariate analyses, we developed anew Hannover system (Hannover-II), which combines3 parameters that had independent prognostic rele-vance in multivariate analyses (TPD >2 mm, capsularinvolvement, and a greatest dimension of the largesttumor cell deposit �0.1 mm). However, the determi-nation of all parameters is required only for patientswho have a maximal diameter deposit �0.1 mm. Theadditional parameters of TPD and capsular involve-ment had no impact on prognosis in patients who hadmicrometastases that measured <0.1 mm.

The Hannover-II system represents a unique risk-assessment system that has the ability to significantly sepa-rate all groups of patients, and particularly patients withhigher tumor burden. Determining the other 2 parame-ters (TPD and capsular invasion) entailed little additionalwork. TPD can be measured very quickly, and capsularinvolvement can be detected easily during routine histo-pathology of the SLN. The combination of different inde-pendent parameters is leading to a considerable gain inknowledge, especially with regard to patients with highertumor burden.

Original Article


Several studies have demonstrated an impact on sur-vival by the number of positive lymph nodes; therefore,that parameter is included in the current AJCC classifica-tion system.9,19 However, Jakub et al reported that thenumber of positive SLNs did not affect survival when themetastatic disease was confined to the SLNs and did notspread to the non-SLNs.22 In line with these findings, theparameter number of positive SLNs (1 vs >1) couldachieve significance only in RFS, and not in OS. in ourunivariate analyses.

In SLN-positive patients who did not undergoCLND, 10 of 42 patients (23.8%) developed recurrentdisease, compared with 70 of 105 (40%) SLN-positivepatients with CLND. The lower percentage of recurrencesin the group of patients without CLND can be explainedby the lower tumor burden of the respective SLNs. In 22of 42 patients who did not undergo CLND, we did notrecommended a CLND procedure because of the verylow tumor burden (isolated, immunohistochemically pos-itive cells).

The underlying study was calculated with a medianfollow-up of 47.8 months, similar to what was reportedby Starz et al (median, 45.5 months) and longer than thatreported by van Akkooi et al (median, 36 months).10,12

Nevertheless, the possibility cannot be excluded that, overthe long run, different findings can be achieved because oflead time bias. This phenomenon is based on the observa-tion that smaller metastatic deposits will take longer forrecurrence and the detection of possible metastatic dis-ease.23 Future analyses with long-term follow-up shouldaddress this problem.

In conclusion, in this report, we present 3 statisti-cally independent prognostic parameters that can bedetermined easily in a routine histopathologic examina-tion of SLNs. The combination of these parameters wasable to predict the prognosis for patients with melanomamost precisely. However, further studies in multi-institu-tional patient populations will be necessary to substantiateour findings.

CONFLICT OF INTEREST DISCLOSURESThe authors made no disclosures.

REFERENCES

1. Morton DL, Wen DR, Wong JH, et al. Technical details ofintraoperative lymphatic mapping for early stage melanoma.Arch Surg. 1992;127:392-399.

2. Carlson GW, Murray DR, Lyles RH, Staley CA, Hestley A,Cohen C. The amount of metastatic melanoma in a sentinel

lymph node: does it have prognostic significance? Ann SurgOncol. 2003;10:575-581.

3. Fink AM, Weihsengruber F, Spangl B, et al. S-classificationof sentinel lymph node biopsy predicts the results of com-plete regional lymph node dissection. Melanoma Res. 2005;15:267-271.

4. Lee JH, Essner R, Torisu-Itakura H, Wanek L, Wang H,Morton DL. Factors predictive of tumor-positive nonsenti-nel lymph nodes after tumor-positive sentinel lymph nodedissection for melanoma. J Clin Oncol. 2004;22:3677-3684.

5. Ranieri JM, Wagner JD, Azuaje R, et al. Prognostic impor-tance of lymph node tumor burden in melanoma patientsstaged by sentinel node biopsy. Ann Surg Oncol. 2002;9:975-981.

6. Satzger I, Volker B, Al Ghazal M, Meier A, Kapp A,Gutzmer R. Prognostic significance of histopathological pa-rameters in sentinel nodes of melanoma patients. Histopa-thology. 2007;50:764-772.

7. Scolyer RA, Li LX, McCarthy SW, et al. Micromorphomet-ric features of positive sentinel lymph nodes predict involve-ment of nonsentinel nodes in patients with melanoma. Am JClin Pathol. 2004;122:532-539.

8. Starz H, Balda BR, Kramer KU, Buchels H, Wang H. Amicromorphometry-based concept for routine classificationof sentinel lymph node metastases and its clinical rele-vance for patients with melanoma. Cancer. 2001;91:2110-2121.

9. Balch CM, Buzaid AC, Soong SJ, et al. Final version of theAmerican Joint Committee on Cancer staging system forcutaneous melanoma. J Clin Oncol. 2001;19:3635-3648.

10. Starz H, Siedlecki K, Balda BR. Sentinel lymphonodectomyand S-classification: a successful strategy for better predictionand improvement of outcome of melanoma. Ann Surg.Oncol. 2004;11:162S-168S.

11. van Akkooi A, de Wilt J, Verhoef C, et al. Clinical rele-vance of melanoma micrometastases (<0.1 mm) in sentinelnodes: are these nodes to be considered negative? AnnOncol. 2006;17:1578-1585.

12. van Akkooi AC, Nowecki ZI, Voit C, et al. Sentinel nodetumor burden according to the Rotterdam criteria is themost important prognostic factor for survival in melanomapatients: a multicenter study in 388 patients with positivesentinel nodes. Ann Surg. 2008;248:949-955.

13. Hermanek P, Sobin LH, Wittekind C. How to improve thepresent TNM staging system. Cancer. 1999;86:2189-2191.

14. Gutzmer R, Al Ghazal M, Geerlings H, Kapp A. Sentinelnode biopsy in melanoma delays recurrence but does notchange melanoma-related survival: a retrospective analysis of673 patients. Br J Dermatol. 2005;153:1137-1141.

15. Garbe C, Schadendorf D. Surveillance and follow-up exami-nations in cutaneous melanoma. Onkologie. 2003;26:241-246.

16. Cochran AJ, Balda BR, Starz H, et al. The AugsburgConsensus. Techniques of lymphatic mapping, sentinellymphadenectomy, and completion lymphadenectomy incutaneous malignancies. Cancer. 2000;89:236-241.

17. Scheri RP, Essner R, Turner RR, Ye X, Morton DL.Isolated tumor cells in the sentinel node affect long-termprognosis of patients with melanoma. Ann Surg Oncol.2007;14:2861-2866.

18. Satzger I, Volker B, Meier A, Schenck F, Kapp A, GutzmerR. Prognostic significance of isolated HMB45 or Melan A



positive cells in melanoma sentinel lymph nodes. Am J SurgPathol. 2007;31:1175-1180.

19. Gershenwald JE, Thompson W, Mansfield PF, et al.Multi-institutional melanoma lymphatic mapping experi-ence: the prognostic value of sentinel lymph node status in612 stage I or II melanoma patients. J Clin Oncol. 1999;17:976-983.

20. Govindarajan A, Ghazarian DM, McCready DR, LeongWL. Histological features of melanoma sentinel lymph nodemetastases associated with status of the completion lymph-adenectomy and rate of subsequent relapse. Ann Surg Oncol.2007;14:906-912.

21. Satzger I, Meier A, Gutzmer R. Correspondence to: theprognostic significance of isolated immunohistochemicallypositive cells in sentinel lymph nodes of melanoma patients.Am J Surg Pathol. 2008;32:1107-1108.

22. Jakub JW, Huebner M, Shivers S, et al. The number oflymph nodes involved with metastatic disease does not affectoutcome in melanoma patients as long as all disease is con-fined to the sentinel lymph node. Ann Surg Oncol. 2009;16:2245-2251.

23. Scolyer RA, Murali R, Satzger I, Thompson JF. The detec-tion and significance of melanoma micrometastases in senti-nel nodes. Surg Oncol. 2008;17:165-174.

Original Article


comparison of classification systems in melanoma sentinel lymph nodes—an analysis of 697 patients...

Documents