new england

The
journal of medicine
established in 1812 June 8, 2017 vol. 376 no. 23

Completion Dissection or Observation for Sentinel-Node

Metastasis in Melanoma
M.B. Faries, J.F. Thompson, A.J. Cochran, R.H. Andtbacka, N. Mozzillo, J.S. Zager, T. Jahkola, T.L. Bowles, A. Testori,
P.D. Beitsch, H.J. Hoekstra, M. Moncrieff, C. Ingvar, M.W.J.M. Wouters, M.S. Sabel, E.A. Levine, D. Agnese,
M. Henderson, R. Dummer, C.R. Rossi, R.I. Neves, S.D. Trocha, F. Wright, D.R. Byrd, M. Matter, E. Hsueh,
A. MacKenzie-Ross, D.B. Johnson, P. Terheyden, A.C. Berger, T.L. Huston, J.D. Wayne, B.M. Smithers, H.B. Neuman,
S. Schneebaum, J.E. Gershenwald, C.E. Ariyan, D.C. Desai, L. Jacobs, K.M. McMasters, A. Gesierich, P. Hersey,
S.D. Bines, J.M. Kane, R.J. Barth, G. McKinnon, J.M. Farma, E. Schultz, S. Vidal-Sicart, R.A. Hoefer, J.M. Lewis,
R. Scheri, M.C. Kelley, O.E. Nieweg, R.D. Noyes, D.S.B. Hoon, H.-J. Wang, D.A. Elashoff, and R.M. Elashoff

a bs t r ac t

BACKGROUND
Sentinel-lymph-node biopsy is associated with increased melanoma-specific survival (i.e., The authors’ full names, academic de-
survival until death from melanoma) among patients with node-positive intermediate- grees, and affiliations are listed in the
Appendix. Address reprint requests to Dr.
thickness melanomas (1.2 to 3.5 mm). The value of completion lymph-node dissection for Faries at 11818 Wilshire Blvd., Suite 200,
patients with sentinel-node metastases is not clear. Los Angeles, CA 90025, or at mfaries@
theangelesclinic .org.
METHODS
N Engl J Med 2017;376:2211-22.
In an international trial, we randomly assigned patients with sentinel-node metastases detected DOI: 10.1056/NEJMoa1613210
by means of standard pathological assessment or a multimarker molecular assay to immediate Copyright © 2017 Massachusetts Medical Society.
completion lymph-node dissection (dissection group) or nodal observation with ultrasonogra-
phy (observation group). The primary end point was melanoma-specific survival. Secondary end
points included disease-free survival and the cumulative rate of nonsentinel-node metastasis.
RESULTS
Immediate completion lymph-node dissection was not associated with increased melanoma-
specific survival among 1934 patients with data that could be evaluated in an intention-to-
treat analysis or among 1755 patients in the per-protocol analysis. In the per-protocol analy-
sis, the mean (±SE) 3-year rate of melanoma-specific survival was similar in the dissection
group and the observation group (86±1.3% and 86±1.2%, respectively; P = 0.42 by the log-
rank test) at a median follow-up of 43 months. The rate of disease-free survival was slightly
higher in the dissection group than in the observation group (68±1.7% and 63±1.7%, respec-
tively; P = 0.05 by the log-rank test) at 3 years, based on an increased rate of disease control
in the regional nodes at 3 years (92±1.0% vs. 77±1.5%; P<0.001 by the log-rank test); these
results must be interpreted with caution. Nonsentinel-node metastases, identified in 11.5%
of the patients in the dissection group, were a strong, independent prognostic factor for
recurrence (hazard ratio, 1.78; P = 0.005). Lymphedema was observed in 24.1% of the pa-
tients in the dissection group and in 6.3% of those in the observation group.
CONCLUSIONS
Immediate completion lymph-node dissection increased the rate of regional disease control
and provided prognostic information but did not increase melanoma-specific survival
among patients with melanoma and sentinel-node metastases. (Funded by the National
Cancer Institute and others; MSLT-II ClinicalTrials.gov number, NCT00297895.)

n engl j med 376;23 nejm.org June 8, 2017 2211

 The n e w e ng l a n d j o u r na l of m e dic i n e

S
entinel-lymph-node biopsy is a stan- Me thods
dard procedure in the care of appropriately
selected patients with melanoma. The first Trial Design and Oversight
Multicenter Selective Lymphadenectomy Trial MSLT-II, an international, multicenter, random-
(MSLT-I) confirmed the value of early nodal ized, phase 3 trial to evaluate the usefulness of
A Quick Take
is available at evaluation and treatment.1-3 This prospective, completion lymph-node dissection in patients
NEJM.org international, randomized trial showed that the with melanoma and sentinel-node metastases,
pathologic status of the sentinel node or nodes consisted of a screening phase in which patients
was the most important prognostic factor and were enrolled before sentinel-node biopsy and a
that patients who underwent sentinel-node biopsy randomization phase in which completion lymph-
had fewer recurrences of melanoma than pa- node dissection was compared with observation
tients who underwent wide excision and nodal and nodal ultrasonography (Fig. 1). The trial was
observation. Among patients with intermediate- conducted at 63 centers.
thickness melanomas (defined as 1.2 to 3.5 mm) MSLT-II was designed by the MSLT-II execu-
and nodal metastases, early surgical treatment, tive committee with input from the pathology
guided by sentinel-node biopsy, was associated and ultrasonography oversight committees (see
with increased melanoma-specific survival (sur- the Supplementary Appendix, available with the
vival until death from melanoma). These results full text of this article at NEJM.org). Data were
provide support for the recommendation by sev- collected prospectively on paper and later on
eral professional organizations that staging by Web-based case-report forms. The authors vouch
means of sentinel-node biopsy should be per- for the accuracy and completeness of the data
formed when appropriate.4-7 and analyses reported and for the fidelity of the
Currently, immediate completion lymph-node trial to the protocol, available at NEJM.org.
dissection (removal of the remaining regional Nodal metastasis was determined by means
lymph nodes after sentinel-node excision) is usu- of standard pathological assessment (including
ally recommended for patients with sentinel-node immunohistochemical tests performed according
metastases. However, prospective evidence of the to institutional protocols) or by means of a pre-
efficacy of completion lymph-node dissection is viously described quantitative reverse-transcrip-
lacking, and the procedure carries a risk of ad- tase–polymerase-chain-reaction (RT-PCR) assay
verse events.8 Results of retrospective evaluations during the screening phase.15 Patients had to
of the usefulness of completion lymph-node dis- have undergone randomization and completion
section are inconclusive.9-11 Available data from lymph-node dissection within 140 days after di-
one prospective study do not suggest a benefit agnostic biopsy.
from immediate dissection, but this study is not The randomization phase involved enrollment
sufficiently powered to rule out a clinically sig- of patients who had undergone screening and
nificant benefit.12 In addition, in most patients, had pathologically or molecularly positive sentinel-
nodal disease is limited to the sentinel lymph node metastases and patients who had not under-
node or nodes and is removed by means of bi- gone screening in whom sentinel-node metasta-
opsy. Conversely, patients with even microscopic ses were detected by means of pathological
involvement of nonsentinel nodes have an over- assessment. In the randomization phase, fewer
all poorer prognosis and outcomes that are patients with RT-PCR–positive findings than
similar to those in patients with clinically appar- anticipated were enrolled. In 2012, the data and
ent nodal disease.13,14 safety monitoring board determined that such
In the second Multicenter Selective Lymphad- patients should no longer undergo randomization,
enectomy Trial (MSLT-II), we evaluated the use- since attainment of sufficient power to evaluate
fulness of completion lymph-node dissection in
patients with melanoma and sentinel lymph-
Figure 1 (facing page). Trial Design, Enrollment,
node metastases as compared with observation
and Outcomes.
with frequent nodal ultrasonography and dissec-
RT-PCR denotes reverse-transcriptase polymerase
tion only in patients in whom clinically detected chain reaction.
nodal recurrence had developed.

2212 n engl j med 376;23 nejm.org June 8, 2017

 Completion Dissection or Observation in Melanoma

3531 Patients were screened and enrolled

251 Were ineligible

91 Had no data or did not undergo
sentinel-node biopsy

608 Were sentinel-node positive 2581 Were sentinel-node negative

835 Were not qualified for RT-PCR

673 Were not at RT-PCR center or joined
study after RT-PCR had ended
162 Had inadequate paraffin sample
231 Declined
randomization
1339 Were RT-PCR negative
407 Were RT-PCR positive

181 Declined
randomization

377 Underwent randomization 226 Underwent randomization

1431 Were directly enrolled in

1939 Underwent randomization
randomization phase
1934 Had data that could be evaluated
95 Were ineligible
1336 Underwent randomization

971 Were assigned to completion 968 Were assigned to nodal

lymph-node dissection observation
824 Underwent dissection 931 Underwent observation
140 Declined dissection 29 Declined observation
3 Did not undergo dissec- 7 Did not undergo obser-
tion for unknown reason vation for unknown
4 Were ineligible reason
1 Was ineligible

626 Were included in follow-up 664 Were included in follow-up

6 col
23 Completed follow-up 27 Completed405follow-up
pt
200 Died 197 Died
83 Withdrew 34 Withdrew
30 Were lost to follow-up 42 Were lost to follow-up
3 Had other disease 2 Had other disease
2 Had protocol violation 1 Had protocol violation

967 Were included in intention-to-treat analysis 967 Were included in intention-to-treat analysis
824 Were included in per-protocol analysis 931 Were included in per-protocol analysis

n engl j med 376;23 nejm.org June 8, 2017 2213

 The n e w e ng l a n d j o u r na l of m e dic i n e

a therapeutic effect in that group was not feasi- larity. Follow-up of the dissection group involved
ble. The data and safety monitoring board recom- the same schedule, but without protocol-man-
mended continued follow-up of these patients to dated nodal ultrasonography.
assess outcomes.
At the third interim analysis, the data and Statistical Analysis
safety monitoring board determined that detec- For the primary end point, melanoma-specific
tion of a significant survival difference between survival, we used the log-rank test to compare
the trial groups was unlikely and recommended the rates among patients in the dissection group
that the current primary end-point data be re- and the observation group in the intention-to-
leased. Intention-to-treat and per-protocol analy- treat population. Secondary end points included
ses of the outcome variables showed similar overall survival, disease-free survival, survival
results. Results of per-protocol analyses are without recurrence of regional nodal metastases,
reported in this article, since they are likely to distant metastasis–free survival, and the extent
be the most clinically pertinent. The intention-to- of nodal involvement. Time zero was the time of
treat data for the primary end point (melanoma- randomization. Melanoma-specific survival was
specific survival) are provided in Figure S1 in the determined at the time of melanoma-related
Supplementary Appendix. death. Disease-free survival was the time to any
recurrence. Survival without nodal recurrence
Patients was the time to recurrence within the draining
Eligible patients who provided written informed nodal basin. For survival, comparisons between
consent were randomly assigned to undergo com- the two groups were performed by means of the
pletion lymph-node dissection or nodal observa- log-rank test for univariable testing and Cox re-
tion. These patients were 18 to 75 years of age gression for adjusted comparisons. Nodal recur-
and had clinically localized cutaneous melanoma, rence occurred in a draining regional basin, local
an Eastern Cooperative Oncology Group perfor- and in-transit recurrence occurred between the
mance status of 0 or 1 (on a 5-point scale, with primary site and the regional basin, and distant
0 indicating an absence of disability and higher recurrence occurred beyond the regional basin.
numbers indicating greater disability), a non– We estimated that with a total sample of 1925
melanoma-related life expectancy of 10 years or patients, the trial would have a power of 83% to
more, and a tumor-positive sentinel node. The detect a between-group difference of 5 percent-
trial was opened before the universal application age points in melanoma-specific survival. All tests
of registration. The trial opened in December were two-tailed. Power was reassessed by the
2004 and was registered at Clinicaltrials.gov on data and safety monitoring board before closure
February 27, 2006. At the time of registration, of enrollment to ensure that an adequate sample
119 patients had been enrolled in the trial. size had been obtained. The cumulative rate of
Randomization was performed in a 1:1 ratio nonsentinel-node metastases was determined by
with the use of a permuted-block design, which clinical follow-up in the observation group and
was stratified according to Breslow thickness, according to total in-basin nodal recurrence or
ulceration, method of metastasis detection (stan- nonsentinel-node metastasis on immediate com-
dard pathological assessment or RT-PCR assay), pletion lymph-node dissection in the dissection
and enrollment at an MSLT-I center. Patients who group.
were assigned to the observation group were Data were summarized with means and stan-
monitored by means of clinical examination every dard deviations, medians and ranges, or both in
4 months during the first 2 years, every 6 months the intention-to-treat and per-protocol analyses.
during years 3 through 5, and then annually. Chi-square and Wilcoxon rank-sum tests were
Nodal ultrasonographic assessment of the senti- used to compare results for patients in the dis-
nel-node basin occurred at each visit for the first section group who actually underwent comple-
5 years; findings were considered to be abnor- tion lymph-node dissection (these patients were
mal on the basis of a length:depth ratio of less included in the per-protocol and intention-to-
than 2, a hypoechoic center, an absence of hilar treat analyses) with results for those who did not
vessels, or focal nodularity with increased vascu- undergo the assigned completion lymph-node dis-

2214 n engl j med 376;23 nejm.org June 8, 2017

 Completion Dissection or Observation in Melanoma

section (these patients were included only in the specific survival between the dissection group
intention-to-treat analysis). Survival curves were and the observation group in the per-protocol
computed with the use of the Kaplan–Meier analysis (86±1.3% and 86±1.2%, respectively;
method and stratified according to group alone P = 0.42 by the log-rank test) (Fig. 2A) or the inten-
or according to group and method of metastasis tion-to-treat analyses (Fig. S1 in the Supplementary
detection (pathological assessment vs. RT-PCR). Appendix). In addition, there was no significant
Cox proportional-hazards regression models were between-group difference in melanoma-specific
constructed separately for the two groups; these survival after adjustment for other prognostic
models included demographic factors, trial strati- factors (hazard ratio for death, 1.08; 95% confi-
fication factors, and nonsentinel-node metasta- dence interval [CI], 0.88 to 1.34; P = 0.42).
sis at the time of completion lymph-node dissec- An analysis with available follow-up data sug-
tion. Subgroup analyses included subgroups that gested that an RT-PCR–positive sentinel node did
were defined according to the patients’ sex and not have an effect on survival that was as nega-
age, the Breslow thickness, the location and tive as the effect anticipated in the statistical
number of positive nodes, and the presence or design of the trial, so the results from the two
absence of ulceration. Cox proportional-hazards groups are also reported separately here and in
regression was used to estimate the subgroup- the remaining analyses (Fig. 2B). The results of
specific hazard ratios. an analysis of both groups together are shown in
Figure S2 in the Supplementary Appendix. A sub-
group analysis, including an analysis based on
R e sult s
sentinel-node tumor burden, did not reveal any
Patients subgroups that derived a significant melanoma-
From December 2004 through March 2014, a specific survival benefit from completion lymph-
total of 3531 patients were enrolled in the node dissection (Fig. S3 in the Supplementary
screening phase and 1939 patients underwent Appendix).
randomization (Fig. 1). Demographic and patho- At 3 years of follow-up, the rate of disease-
logic features of the dissection and observation free survival was slightly higher in the dissection
groups were similar (Table 1, and Table S1 in the group than in the observation group (68±1.7%
Supplementary Appendix). A greater proportion and 63±1.7%, respectively; P = 0.05 by the log-rank
of patients assigned to completion lymph-node test) (Fig. 3A, and Fig. S2A in the Supplementary
dissection than to observation declined their as- Appendix), although the results of secondary
signed treatment. However, the per-protocol co- outcome analyses must be viewed cautiously given
horts were similar with respect to prognostic the lack of significance for the primary end
factors (Table 1). point. This difference in disease-free survival
In the dissection group, 143 patients were ex- appears to result from a reduction in the rate of
cluded from the per-protocol analysis. Of those nodal recurrence after completion lymph-node
patients, 140 declined the assigned treatment. dissection (Fig. 3B, and Fig. S2B in the Supple-
Patients who were excluded from the per-proto- mentary Appendix). This corresponds to an in-
col analysis were more likely than patients who crease in the rate of disease control in the regional
were not excluded to have never smoked, to have nodes at 3 years (92±1.0% in the dissection
nonulcerated primary tumors, and to have an group vs. 77±1.5% in the observation group,
RT-PCR–positive sentinel node (Table S2 in the P<0.001 by the log-rank test). After adjustment,
Supplementary Appendix). The sentinel-node tu- the rate of nodal recurrence among patients
mor burden in both groups was low; the median with sentinel-node metastases detected by means
diameter of the largest tumor deposit was 0.61 of pathological assessment was 69% lower in the
mm in the dissection group and 0.67 mm in the dissection group than in the observation group
observation group. (hazard ratio, 0.31; 95% CI, 0.24 to 0.41; P<0.001).
No significant between-group difference in dis-
Survival Rates tant metastasis–free survival was detected (ad-
At 3 years of follow-up, there was no significant justed hazard ratio, 1.10; 95% CI, 0.92 to 1.31;
difference in the mean (±SE) rate of melanoma- P = 0.31) (Fig. 3C, and Fig. S2C in the Supplemen-

n engl j med 376;23 nejm.org June 8, 2017 2215

 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 1. Baseline Characteristics of the Patients in the Per-Protocol Analysis.*

Dissection Observation
Characteristic (N = 824) (N = 931)
Sex — no. (%)
Female 346 (42.0) 382 (41.0)
Male 478 (58.0) 549 (59.0)
Age — yr
Mean 52.5±12.9 53.2±13.6
Median (range) 53.7 (18–76) 54.9 (19–76)
Smoking status — no./total no. (%)†
Never 463/803 (57.7) 522/907 (57.6)
Former 193/803 (24.0) 227/907 (25.0)
Current 147/803 (18.3) 158/907 (17.4)
Breslow thickness
Mean — mm 2.76±2.34 2.70±2.11
Median (range) — mm 2.10 (0.34–28.0) 2.10 (0.35–30.0)
<1.50 mm — no. (%) 237 (28.8) 257 (27.6)
1.50–3.50 mm — no. (%) 404 (49.0) 462 (49.6)
>3.50 mm — no. (%) 183 (22.2) 212 (22.8)
Primary site — no. (%)
Arm or leg 327 (39.7) 382 (41.0)
Head or neck 113 (13.7) 128 (13.7)
Trunk 384 (46.6) 421 (45.2)
Ulceration — no. (%)
Absent 508 (61.7) 578 (62.1)
Present 316 (38.3) 353 (37.9)
No. of positive sentinel lymph nodes — no. of
patients (%)
0, RT-PCR–positive 80 (9.7) 111 (11.9)
1 596 (72.3) 643 (69.1)
2 121 (14.7) 162 (17.4)
3 18 (2.2) 10 (1.1)
>3 9 (1.1) 5 (0.5)
Diameter of sentinel-lymph-node metastasis — mm‡
Mean 1.07 1.11
Median 0.61 0.67
Interquartile range 0.27–1.32 0.23–1.38
Size of sentinel-lymph-node metastasis — no. of
patients/total no. (%)
<0.1 mm 45/566 (8.0) 65/623 (10.4)
0.1–1.0 mm 333/566 (58.8) 343/623 (55.1)
>1.0 mm 188/566 (33.2) 215/623 (34.5)
Received adjuvant therapy — no./total no. (%)§ 66/814 (8.1) 60/922 (6.5)

* Plus–minus values are means ±SD. There were no significant between-group differences in the characteristics listed here.
Percentages may not total 100 because of rounding. RT-PCR denotes reverse-transcriptase polymerase chain reaction.
† Data were missing for 21 patients in the dissection group and 24 patients in the observation group.
‡ The sentinel-node metastasis burden, the longest diameter of the largest tumor deposit, was not available for 178 patients
in the dissection group and 197 patients in the observation group.
§ Data were not available for 10 patients in the dissection group and 9 patients in the observation group.

2216 n engl j med 376;23 nejm.org June 8, 2017

 Completion Dissection or Observation in Melanoma

tary Appendix). Types of initial recurrence are

A
listed in Table S4 in the Supplementary Appendix. 1.0

Probability of Melanoma-Specific
Censored

Prognostic Factors 0.8 Observation

Potential prognostic factors affecting melanoma-
0.6 Dissection

Survival
specific survival were examined. Since the patho-
logic status of nonsentinel nodes was unknown 0.4
in the observation group, the two groups of the
trial were considered separately. In the entire 0.2
trial (including patients with positive RT-PCR P=0.55
results), Breslow thickness and the number of 0.0
0 1 2 3 4 5 6 7 8 9 10
sentinel nodes that were positive on pathological
Years after Randomization
assessment (0 vs. >0) were significant prognos-
tic factors in both groups, and male sex was a No. at Risk
Dissection 824 759 654 510 389 275 191 128 83 39 13
significant prognostic factor in the observation Observation 931 856 734 564 425 304 217 151 95 55 13
group (Table S3 in the Supplementary Appen-
dix). However, since the RT-PCR group appeared B
1.0 Dissection, RT-PCR–positive

Probability of Melanoma-Specific
P=0.35
to be prognostically distinct, an analysis was
also performed that included only patients with 0.8

sentinel nodes that were positive on pathological 0.6 Observation, RT-PCR–positive

Survival
assessment (Table 2). In the observation group,
P=0.47
male sex was no longer a significant prognostic 0.4 Observation, pathologically detected
factor, and it remained nonsignificant in the dis- 0.2 Dissection, pathologically detected
section group. Breslow thickness was a signifi- Censored
cant prognostic factor in both groups, and the 0.0
0 1 2 3 4 5 6 7 8 9 10
pathologic status of nonsentinel nodes was a
Years after Randomization
significant prognostic factor in the dissection
group (hazard ratio for death, 1.78; P = 0.005). No. at Risk
Subgroup 1 744 682 581 441 326 214 144 92 53 21 6
The number of involved sentinel nodes was not Subgroup 2 820 751 639 482 348 241 163 109 64 34 8
a significant prognostic factor. Subgroup 3 80 77 73 69 63 61 47 36 30 18 7
Subgroup 4 111 105 95 82 77 63 54 42 31 21 5
Among patients who underwent immediate
completion lymph-node dissection, nonsentinel-
Figure 2. Melanoma-Specific Survival, According to Trial Group and Meth-
node metastases were detected on pathological od of Detection of Metastasis.
assessment in 11.5%, and over time, with nodal Panel A shows melanoma-specific survival according to trial group (com-
recurrences in that group, the percentage of pletion lymph-node dissection or observation) in the per-protocol analysis.
patients in whom nonsentinel-node metastases Panel B shows melanoma-specific survival according to the method of de-
were detected increased to an actuarial rate of tection of sentinel-node metastasis (RT-PCR or pathological assessment).
Subgroup 1 comprised patients in the dissection group with pathologically
17.9% at 3 years and 19.9% at 5 years (Fig. 3D).
detected metastases; subgroup 2, patients in the observation group with
In the observation group, the percentage of pathologically detected metastases; subgroup 3, those in the dissection
patients in whom ultrasonographic or physical group with RT-PCR–detected metastases; and subgroup 4, those in the ob-
examination revealed involved nonsentinel nodes servation group with RT-PCR–detected metastases. P values were calculat-
increased to 22.9% at 3 years and 26.1% at 5 years, ed with the use of log-rank tests.
exceeding the rate in the dissection group at both
time points (P = 0.02 and P = 0.005, respectively).
had had lymphedema (P<0.001). Among the pa-
Adverse Events tients who had lymphedema, this condition was
Adverse events were more common among pa- mild in 64%, moderate in 33%, and severe in 3%.
tients after completion lymph-node dissection
than among patients in the observation group. Discussion
At the most recent follow-up on April 30, 2016,
a total of 24.1% of the patients in the dissection The management of regional lymph nodes has
group and 6.3% of those in the observation group long been controversial in the treatment of many

n engl j med 376;23 nejm.org June 8, 2017 2217

 The n e w e ng l a n d j o u r na l of m e dic i n e

Dissection, RT-PCR–positive Observation, RT-PCR–positive Dissection, pathologically detected Observation, pathologically detected

A B
1.0 Censored 1.0 P=0.81

Probability of Nodal Recurrence–free

Probability of Disease-free Survival

P=0.66 0.8
0.8 P<0.001

0.6 0.6

Survival
P=0.02

0.4 0.4

0.2 0.2

0.0 0.0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

Years after Randomization Years after Randomization

No. at Risk No. at Risk

Subgroup 1 744 606 465 343 260 174 114 77 41 15 5 Subgroup 1 744 662 552 412 308 201 138 89 50 21 6
Subgroup 2 820 629 477 346 249 171 124 87 50 28 6 Subgroup 2 820 659 525 386 282 194 137 96 56 31 7
Subgroup 3 80 73 69 63 59 56 42 32 26 17 6 Subgroup 3 80 75 72 67 62 59 47 36 30 18 7
Subgroup 4 111 103 93 81 73 59 49 36 27 18 4 Subgroup 4 111 105 94 82 77 62 53 41 30 20 4

C D
1.0 Censored 0.5
Probability of Distant Disease–free

Cumulative Incidence of Nonsentinel-Node

P=0.66
0.8
0.4

0.6
Survival

P=0.27
0.3 Observation
Metastasis

0.4
Dissection
0.2
0.2

0.0 0.1
0 1 2 3 4 5 6 7 8 9 10
Years after Randomization
No. at Risk 0.0
Subgroup 1 744 649 520 384 286 192 128 86 45 17 5 0 2 4 6 8 10
Subgroup 2 820 708 580 431 312 217 154 106 62 34 8 Years after Randomization
Subgroup 3 80 76 72 66 61 58 44 33 27 17 6
Subgroup 4 111 104 94 81 73 59 50 38 28 19 5

Figure 3. Disease-free Survival, Survival without Nodal Recurrence, and Distant Metastasis–free Survival, According to Trial Group,
and the Cumulative Rate of Nonsentinel-Node Metastasis.
Panel A shows disease-free survival, Panel B shows survival without nodal recurrence, and Panel C shows distant metastasis–free survival
according to trial group (completion lymph-node dissection or observation). Subgroup 1 comprised patients in the dissection group
with pathologically detected metastases; subgroup 2, patients in the observation group with pathologically detected metastases; sub-
group 3, those in the dissection group with RT-PCR–detected metastases; and subgroup 4, those in the observation group with RT-PCR–
detected metastases. Panel D shows the cumulative rate of nonsentinel-node metastasis among patients in the dissection group who
had positive findings on pathological assessment or nodal recurrence and among patients in the observation group who had nodal re-
currence. P values were calculated with the use of log-rank tests.

solid tumors, particularly melanoma.16 The patients with intermediate-thickness melanoma.3

MSLT-I confirmed the staging value of sentinel- The findings of that trial provided support for
node biopsy and showed a therapeutic advantage the use of sentinel-node biopsy, which is now
of early treatment of nodal metastases among recommended in the guidelines of most national

2218 n engl j med 376;23 nejm.org June 8, 2017

 Completion Dissection or Observation in Melanoma

Table 2. Hazard Ratios for Melanoma-Related Death, According to Multivariable Prognostic Factors.*

Prognostic Factor Dissection Observation

Hazard Ratio Hazard Ratio

(95% CI) P Value (95% CI) P Value
Sex: male vs. female 1.13 (0.80–1.59) 0.50 1.41 (0.98–2.05) 0.07
Age, per 1-yr increase 1.00 (0.99–1.01) 0.93 1.01 (0.99–1.02) 0.15
Breslow thickness
<1.50 mm† 1.00 1.00
1.50–3.50 mm 1.64 (0.96–2.79) 0.07 2.46 (1.34–4.53) 0.004
>3.50 mm 3.82 (2.19–6.66) <0.001 4.32 (2.31–8.09) <0.001
Ulceration: present vs. absent 1.97 (1.40–2.77) <0.001 2.17 (1.55–3.05) <0.001
Site of melanoma
Arm or leg† 1.00 1.00
Head or neck 0.81 (0.44–1.48) 0.49 1.60 (0.96–2.66) 0.07
Trunk 1.26 (0.89–1.77) 0.19 1.05 (0.74–1.49) 0.80
No. of positive sentinel nodes
1† 1.00 1.00
2 1.08 (0.71–1.62) 0.73 1.27 (0.87–1.84) 0.21
≥3 1.17 (0.61–2.24) 0.64 2.01 (0.82–4.95) 0.13
Nonsentinel nodes: positive vs. negative 1.78 (1.19–2.67) 0.005 NA

* Patients with positive findings on RT-PCR were excluded from this analysis. NA denotes not applicable.
† This group served as the reference group.

and professional organizations for the treatment siderable risk of selection bias.9-11 The findings
of melanoma.4-7 of one prospective study were similar to those in
However, in patients with sentinel-node metas- our trial, but its size (483 patients underwent
tases, the value of completion lymph-node dis- randomization) and most recent follow-up left
section remains controversial. Since most such enough statistical uncertainty to preclude defini-
patients have all nodal metastases removed by tive conclusions.12 MSLT-II, in which 1939 pa-
means of the sentinel-node biopsy procedure, tients underwent randomization with a median
they cannot derive additional therapeutic value follow-up of 43 months, provided sufficient data
from completion lymph-node dissection. Even to resolve the central question: no significant
microscopic nonsentinel-node metastases por- survival benefit was imparted by immediate com-
tend a markedly worse prognosis, similar to that pletion lymph-node dissection among patients
of patients with bulky, clinically diagnosed metas- with sentinel-node metastases. However, com-
tases,13,14 than the prognosis in patients with pletion lymph-node dissection did provide other
metastases that are limited to the sentinel lymph potential value for patients with melanoma,
nodes. Patients with nonsentinel-node metasta- including improved staging and an increased
sis may be unlikely to benefit from early dissec- rate of regional disease control.
tion. Finally, completion lymph-node dissection Most patients in the trial population had a
is associated with higher morbidity than sentinel- low-volume nodal tumor burden. Indeed, some
node biopsy alone, so an appraisal of the value patients had only molecular indications of mela-
of the procedure is important. noma in the sentinel node, determined by means
Previous data regarding this clinical question of RT-PCR. Those patients had outcomes that
have been inconclusive. Retrospective series have were not as poor as those in retrospective studies
produced varied results and are subject to a con- using the same assay.15,17 However, any variance

n engl j med 376;23 nejm.org June 8, 2017 2219

 The n e w e ng l a n d j o u r na l of m e dic i n e

from pretrial event-rate estimates is unlikely to evaluations or in patients who receive treatment
have affected the overall result, since the RT-PCR– at institutions that are not able to perform
positive group constituted only 12% of the ran- nodal ultrasonography.
domized study population. Furthermore, the num- The advantages of immediate completion
ber of patients with pathologically detected lymph-node dissection are tempered by the com-
metastases actually exceeded the number in the plications of the procedure. As of this writing,
statistical plan. Patients with a larger sentinel- lymphedema has been observed in 24% of the
node tumor burden are more likely than patients patients in the dissection group and 6% of the
with a smaller burden to have nonsentinel-node patients in the observation group. As expected,
metastases, and the small number of patients there were significantly more complications
with a larger sentinel-node tumor burden in this among patients who underwent completion
trial limits statistical confidence for those pa- lymph-node dissection than among those who
tients specifically. It may be possible to use an did not, although the adverse events associated
estimation of the risk of nonsentinel-node metas- with the surgical procedure were often transient.
tases based on sentinel-node tumor burden and Although a complete assessment and compari-
primary tumor characteristics to help identify son of lymphedema with other complications
patients who may benefit from completion lymph- would require additional follow-up, we think
node dissection.18-20 However, a subgroup evalu- that the decreased overall number of dissections
ation of patients with a greater disease burden among patients in the observation group will
(maximal tumor diameter >1 mm) did not indi- translate into decreased complications.
cate that a benefit from completion lymph-node The lack of a survival advantage associated
dissection was more likely in high-risk groups with immediate completion lymph-node dissec-
than in low-risk groups. tion in this trial contrasts with the results of the
The current trial confirms that the patho- MSLT-I. In that trial, patients with nodal disease
logic status of nonsentinel nodes has indepen- and intermediate-thickness melanomas had bet-
dent prognostic value, whereas the number of ter outcomes with immediate surgery than with
involved sentinel nodes was not significantly delayed surgery. The lack of a survival benefit
related to melanoma-specific survival. Although with completion lymph-node dissection in patients
this finding is somewhat counterintuitive, it in MSLT-II suggests that any increase in survival
echoes retrospective data from multiple institu- with early surgery occurred among patients with
tions.13,14 This confirmation in a prospective disease that was limited to the sentinel node.
trial of the large effect of nonsentinel node Patients with nonsentinel-node metastases may
status on prognosis reaffirms its staging value. still undergo salvage treatment with completion
A lack of this information may impede the most lymph-node dissection, but the timing of that
appropriate risk stratification and selection of intervention does not appear to be critical.
adjuvant therapy for patients who do not under- Early completion lymph-node dissection did
go completion lymph-node dissection. not increase survival in the MSLT-II population.
Immediate completion lymph-node dissection It is possible that this was due to dilution of a
reduced the rate of regional nodal recurrence by therapeutic effect, since approximately three
nearly 70%, leading to a small but significant quarters of the population did not have mela-
decrease in the overall risk of recurrence. Since noma in nonsentinel nodes. A comparison of
no significant difference between the groups results in patients with nonsentinel-node metas-
was noted in the primary end point, differences tases in this trial, similar to the latent subgroup
with respect to the secondary end points must analysis in MSLT-I, might address this possibil-
be interpreted with caution. A nonsignificant ity, but it would be difficult to accomplish.21
difference in distant metastasis–free survival First, in this population with a low disease
was noted at late time points, but as of this writ- burden and with the most recent follow-up, ad-
ing, events at those time points have been few, ditional nodal recurrences are expected. Second,
and additional follow-up is necessary. Our trial even at the most recent follow-up, an imbalance
was unable to determine the safety of avoiding in the observed proportion of patients with non-
completion lymph-node dissection in patients sentinel node–positive disease was noted, with
who are unable to undergo frequent follow-up an excess in the observation group. This may be

2220 n engl j med 376;23 nejm.org June 8, 2017

 Completion Dissection or Observation in Melanoma

due to small nonsentinel-node metastases that board from GlaxoSmithKline, Bristol-Myers Squibb, and Provec-
tus Biopharmaceuticals; Dr. Dummer, receiving consulting fees
were not detected on standard pathological ex-
from Novartis, Bristol-Myers Squibb, Roche, Takeda, Pierre
amination. Intensive evaluation of nonsentinel Fabre, Merck Sharp & Dohme, and Amgen; Dr. Wright, receiving
nodes with the use of immunohistochemical grant support from Roche; Dr. Hsueh, receiving lecture fees
tests indicates that the frequency of these occult from Amgen and Castle Biosciences; Dr. MacKenzie-Ross, re-
ceiving fees for serving on an advisory board from Amgen; Dr.
metastases in completion lymph-node dissection Johnson, receiving grant support from Incyte and fees for serving
specimens is very similar to that of excess nodal on an advisory board from Bristol-Myers Squibb and Genoptix;
recurrences (8 to 10%).18 Dr. Terheyden, receiving honoraria and travel support from
Bristol-Myers Squibb and Roche and honoraria from Merck and
Overall, some value may be derived from im- Novartis; Dr. Gershenwald, receiving fees for serving on an ad-
mediate completion lymph-node dissection with visory board from Merck and Castle Biosciences; Dr. McMasters,
regard to staging and an increased rate of re- receiving fees for serving on the board of directors from Provectus
Biopharmaceuticals and fees for serving on an advisory board
gional disease control. However, this value comes and uncompensated equity interest from Elucida Oncology; Dr.
at the cost of increased complications. Barth, holding a pending patent on systems and methods for
The content of this report is solely the responsibility of the guiding tissue resection (patent no. US 14/919,411); and Dr.
authors and does not necessarily represent the official view of the Hoon, holding an issued patent on detection of micrometastasis
National Cancer Institute or the National Institutes of Health. of melanoma and breast cancer in paraffin-embedded, tumor-
Supported by grants (CA189163 and CA29605, to Dr. Faries) draining lymph nodes by means of multimarker quantitative
from the National Cancer Institute and by funding from the reverse-transcriptase–polymerase-chain-reaction assay (patent
Borstein Family Foundation, the Amyx Foundation, the Dr. no. US 7910295). No other potential conflict of interest relevant to
Miriam and Sheldon G. Adelson Medical Research Foundation, this article was reported. Disclosure forms provided by the au-
and the John Wayne Cancer Institute Auxiliary. thors are available with the full text of this article at NEJM.org.
Dr. Faries reports receiving fees for serving on an advisory We thank Donald L. Morton, M.D., (deceased) who founded the
board from Myriad Genetic Laboratories, Amgen, and Immune MSLT Group and whose contributions not only to this trial but
Design; Dr. Thompson, receiving fees for serving on an advisory also to the care of patients with cancer cannot be overestimated.

Appendix
The authors’ full names and academic degrees are as follows: Mark B. Faries, M.D., John F. Thompson, M.D., Alistair J. Cochran,
M.D., Robert H. Andtbacka, M.D., Nicola Mozzillo, M.D., Jonathan S. Zager, M.D., Tiina Jahkola, M.D., Ph.D., Tawnya L. Bowles, M.D.,
Alessandro Testori, M.D., Peter D. Beitsch, M.D., Harald J. Hoekstra, M.D., Ph.D., Marc Moncrieff, M.D., Christian Ingvar, M.D.,
Ph.D., Michel W.J.M. Wouters, M.D., Ph.D., Michael S. Sabel, M.D., Edward A. Levine, M.D., Doreen Agnese, M.D., Michael Hen-
derson, M.D., Reinhard Dummer, M.D., Carlo R. Rossi, M.D., Rogerio I. Neves, M.D., Steven D. Trocha, M.D., Frances Wright, M.D.,
David R. Byrd, M.D., Maurice Matter, M.D., Eddy Hsueh, M.D., Alastair MacKenzie-Ross, M.D., Douglas B. Johnson, M.D., Patrick
Terheyden, M.D., Adam C. Berger, M.D., Tara L. Huston, M.D., Jeffrey D. Wayne, M.D., B. Mark Smithers, M.B., B.S., Heather B. Neu-
man, M.D., Schlomo Schneebaum, M.D., Jeffrey E. Gershenwald, M.D., Charlotte E. Ariyan, M.D., Ph.D., Darius C. Desai, M.D., Lisa
Jacobs, M.D., Kelly M. McMasters, M.D., Ph.D., Anja Gesierich, M.D., Peter Hersey, M.D., Ph.D., Steven D. Bines, M.D., John M. Kane,
M.D., Richard J. Barth, M.D., Gregory McKinnon, M.D., Jeffrey M. Farma, M.D., Erwin Schultz, M.D., Sergi Vidal-Sicart, M.D., Ph.D.,
Richard A. Hoefer, D.O., James M. Lewis, M.D., Randall Scheri, M.D., Mark C. Kelley, M.D., Omgo E. Nieweg, M.D., Ph.D., R. Dirk
Noyes, M.D., Dave S.B. Hoon, Ph.D., He-Jing Wang, M.D., David A. Elashoff, Ph.D., and Robert M. Elashoff, Ph.D.
From the John Wayne Cancer Institute at Saint John’s Health Center, Santa Monica (M.B.F., D.S.B.H.), and the Departments of Pa-
thology (A.J.C.), Biomathematics (H.-J.W., D.A.E., R.M.E.), and Medicine (D.A.E.), University of California, Los Angeles — both in
California; Melanoma Institute Australia and the University of Sydney, Sydney (J.F.T., O.E.N.), Peter MacCallum Cancer Centre, Mel-
bourne, VIC (M.H.), Princess Alexandra Hospital, Brisbane, QLD (B.M.S.), and Newcastle Melanoma Unit, Waratah, NSW (P.H.) — all
in Australia; Huntsman Cancer Institute, Salt Lake City (R.H.A., R.D.N.), and Intermountain Healthcare Cancer Services–Intermountain
Medical Center, Murray (T.L.B.) — both in Utah; Istituto Nazionale dei Tumori Napoli, Naples (N.M.), Istituto Europeo di Oncologia,
Milan (A.T.), and Istituto Oncologico Veneto–University of Padua, Padua (C.R.R.) — all in Italy; H. Lee Moffitt Cancer Center, Tampa,
FL (J.S.Z.); Helsinki University Hospital, Helsinki (T.J.); Dallas Surgical Group, Dallas (P.D.B.); Universitair Medisch Centrum Gronin-
gen, Groningen (H.J.H.), and Netherlands Cancer Institute, Amsterdam (M.W.J.M.W.) — both in the Netherlands; Norfolk and Nor-
wich University Hospital, Norwich (M. Moncrieff), and Guy’s and St. Thomas’ NHS Foundation Trust, London (A.M.-R.) — both in the
United Kingdom; Swedish Melanoma Study Group–University Hospital Lund, Lund, Sweden (C.I.); University of Michigan, Ann Arbor
(M.S.S.); Wake Forest University, Winston-Salem (E.A.L.), and Duke University, Durham (R.S.) — both in North Carolina; Ohio State
University, Columbus (D.A.); University of Zurich, Zurich (R.D.), and Centre Hospitalier Universitaire Vaudois, Lausanne (M. Matter)
— both in Switzerland; Penn State Hershey Cancer Institute, Hershey (R.I.N.), Thomas Jefferson University (A.C.B.) and Fox Chase
Cancer Center (J.M.F.), Philadelphia, and St. Luke’s University Health Network, Bethlehem (D.C.D.) — all in Pennsylvania; Greenville
Health System Cancer Center, Greenville, SC (S.D.T.); Sunnybrook Research Institute, Toronto (F.W.), and Tom Baker Cancer Centre,
Calgary, AB (G.M.) — both in Canada; University of Washington, Seattle (D.R.B.); Saint Louis University, St. Louis (E.H.); Vanderbilt
University (D.B.J., M.C.K.), Nashville, and University of Tennessee, Knoxville (J.M.L.) — both in Tennessee; University Hospital
Schleswig–Holstein–Campus Lübeck, Lübeck (P.T.), University Hospital of Würzburg, Würzburg (A.G.), and City Hospital of Nürnberg,
Nuremberg (E.S.) — all in Germany; SUNY at Stony Brook Hospital Medical Center, Stony Brook (T.L.H.), Memorial Sloan Kettering
Cancer Center, New York (C.E.A.), and Roswell Park Cancer Institute, Buffalo (J.M.K.) — all in New York; Northwestern University
Feinberg School of Medicine (J.D.W.) and Rush University Medical Center (S.D.B.), Chicago; University of Wisconsin, Madison
(H.B.N.); Tel Aviv Sourasky Medical Center, Tel Aviv, Israel (S.S.); M.D. Anderson Medical Center, Houston (J.E.G.); Johns Hopkins
University School of Medicine, Baltimore (L.J.); University of Louisville, Louisville, KY (K.M.M.); Dartmouth–Hitchcock Medical Center,
Lebanon, NH (R.J.B.); Hospital Clinic Barcelona, Barcelona (S.V.-S.); and Sentara CarePlex Hospital, Hampton, VA (R.A.H.).

n engl j med 376;23 nejm.org June 8, 2017 2221

 Completion Dissection or Observation in Melanoma

References
1. Morton DL, Cochran AJ, Thompson lymphadenectomy in melanoma: results 15. Takeuchi H, Morton DL, Kuo C, et al.
JF, et al. Sentinel node biopsy for early- of the Multicenter Selective Lymphade- Prognostic significance of molecular up-
stage melanoma: accuracy and morbidity nectomy Trial (I). Ann Surg Oncol 2010; staging of paraffin-embedded sentinel
in MSLT-I, an international multicenter 17:3324-9. lymph nodes in melanoma patients. J Clin
trial. Ann Surg 2005;242:302-11. 9. Lee DY, Lau BJ, Huynh KT, et al. Im- Oncol 2004;22:2671-80.
2. Morton DL, Thompson JF, Cochran pact of completion lymph node dissection 16. Snow HM. Melanotic cancerous dis-
AJ, et al. Sentinel-node biopsy or nodal on patients with positive sentinel lymph ease. Lancet 1892;140:869-922.
observation in melanoma. N Engl J Med node biopsy in melanoma. J Am Coll Surg 17. McMasters KM, Egger ME, Edwards
2006;355:1307-17. 2016;223:9-18. MJ, et al. Final results of the Sunbelt Mela-
3. Morton DL, Thompson JF, Cochran 10. Bamboat ZM, Konstantinidis IT, Kuk noma Trial: a multi-institutional prospec-
AJ, et al. Final trial report of sentinel-node D, Ariyan CE, Brady MS, Coit DG. Observa- tive randomized phase III study evaluating
biopsy versus nodal observation in mela- tion after a positive sentinel lymph node the role of adjuvant high-dose interferon
noma. N Engl J Med 2014;370:599-609. biopsy in patients with melanoma. Ann alfa-2b and completion lymph node dis-
4. Wong SL, Balch CM, Hurley P, et al. Surg Oncol 2014;21:3117-23. section for patients staged by sentinel
Sentinel lymph node biopsy for melanoma: 11. Wong SL, Morton DL, Thompson JF, lymph node biopsy. J Clin Oncol 2016;34:
American Society of Clinical Oncology et al. Melanoma patients with positive 1079-86.
and Society of Surgical Oncology joint sentinel nodes who did not undergo com- 18. Cochran AJ, Wen DR, Huang RR,
clinical practice guideline. J Clin Oncol pletion lymphadenectomy: a multi-insti- Wang HJ, Elashoff R, Morton DL. Predic-
2012;30:2912-8. tutional study. Ann Surg Oncol 2006;13: tion of metastatic melanoma in nonsenti-
5. Australian Cancer Network Melanoma 809-16. nel nodes and clinical outcome based on
Guidelines Revision Working Party. Clini- 12. Leiter U, Stadler R, Mauch C, et al. the primary melanoma and the sentinel
cal practice guidelines for the manage- Complete lymph node dissection versus node. Mod Pathol 2004;17:747-55.
ment of melanoma in Australia and New no dissection in patients with sentinel 19. Starz H, Welzel J, Bertsch HP, Kretsch-
Zealand. Wellington, New Zealand: Can- lymph node biopsy positive melanoma mer L. Tumor penetrative depth consid-
cer Council Australia/Australian Cancer (DeCOG-SLT): a multicentre, randomised, ers both the size of sentinel lymph node
Network, October 2008. phase 3 trial. Lancet Oncol 2016;17:757-67. metastases and their location in relation
6. Garbe C, Peris K, Hauschild A, et al. 13. Reintgen M, Murray L, Akman K, et al. to the nodal capsule. J Clin Oncol 2011;
Diagnosis and treatment of melanoma: Evidence for a better nodal staging system 29:4843-4.
European consensus-based interdisciplin- for melanoma: the clinical relevance of 20. Chakera AH, Hesse B, Burak Z, et al.
ary guideline — update 2012. Eur J Can- metastatic disease confined to the senti- EANM-EORTC general recommendations
cer 2012;48:2375-90. nel lymph nodes. Ann Surg Oncol 2013; for sentinel node diagnostics in melano-
7. Coit DG, Thompson JA, Algazi A, et al. 20:668-74. ma. Eur J Nucl Med Mol Imaging 2009;36:
NCCN guidelines insights: melanoma, 14. Leung AM, Morton DL, Ozao-Choy J, 1713-42.
version 3.2016. J Natl Compr Canc Netw et al. Staging of regional lymph nodes in 21. Altstein LL, Li G, Elashoff RM. A
2016;14:945-58. melanoma: a case for including nonsenti- method to estimate treatment efficacy
8. Faries MB, Thompson JF, Cochran A, nel lymph node positivity in the American among latent subgroups of a randomized
et al. The impact on morbidity and length Joint Committee on Cancer staging sys- clinical trial. Stat Med 2011;30:709-17.
of stay of early versus delayed complete tem. JAMA Surg 2013;148:879-84. Copyright © 2017 Massachusetts Medical Society.

TRACK THIS ARTICLE’S IMPACT AND REACH

Visit the article page at NEJM.org and click on the Metrics tab for a dashboard
that logs views, citations, media references, and commentary, with easy linking.
Learn more at www.nejm.org/page/article-metrics-faq.

2222 n engl j med 376;23 nejm.org June 8, 2017

Reproduced with permission of copyright owner.
Further reproduction prohibited without
permission.