European Journal of Cancer 84 (2017) 44e54

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.ejcancer.com

Original Research

Survival of melanoma patients treated with targeted

therapy and immunotherapy after systematic upfront
control of brain metastases by radiosurgery

C. Gaudy-Marqueste a,1, A.S. Dussouil a,1, R. Carron b, L. Troin a,

N. Malissen a, A. Loundou c, S. Monestier a, S. Mallet a, M.A. Richard a,
J.M. Régis b, J.J. Grob a,*

a
Dermatology and Skin Cancers Department, UMR911 CRO2, Aix-Marseille University, APHM, Marseille, France
b
Department of Stereotaxic and Functional Neurosurgery, Gamma-knife Unit, Inserm U751, Aix-Marseille University,
APHM, Marseille, France
c
Public Health Department, Aix-Marseille University, APHM, Marseille, France

Received 6 April 2017; received in revised form 16 June 2017; accepted 11 July 2017

KEYWORDS Abstract Background: Targeted therapy (TT) and immunotherapies (ITs) have dramatically
Melanoma; improved survival in metastatic melanoma (MM). However, their efficacy on brain metastasis
Brain metastasis; (BM) remains limited and poorly documented.
Gamma-Knife Patients and methods: Retrospective cohort of consecutive MM patients (pts) with BMs, all
radiosurgery; systematically upfront treated by Gamma-Knife (GK) at first BM and retreated in case of
Check-point new BMs, from 2010 to 2015 at the time when ipilimumab BRAF
MEK inhibitors and
inhibitors; anti-PD1 were introduced in practice. Survival after 1st GK (OSGK1) according to prognostic
Targeted therapies factors and treatment.
Results: Among 179 consecutive pts treated by GK, 109 received IT and/or TT after the 1st
GK. Median OSGK1 was 10.95 months and 1- and 2-year survival rates were 49.5% and
27.4%, respectively, versus a median overall survival (OS) of 2.29 months (p < .001) in those
who did not receive IT or TT. In pts who initially had a single BM, median OS and 1- and 2-
year survival rates were 14.46 months, 66.7% and 43.4%, respectively; in pts with 2e3 BMs:
8.85 months, 46.4% and 31%, respectively; in pts with >3 BMs: 7.25 months, 37.2% and
11.9%, respectively. Multivariate analysis for OSGK1 confirmed that IT and TT were signifi-
cantly and highly protective. Best OSGK1 was observed in BRAFewild-type pts receiving
anti-PD1 or in BRAF-mutated pts receiving BRAF-inhibitors and anti-PD1 (12.26 and

* Corresponding author: Pr Jean-Jacques GROB, Dermatology and Skin Cancers Department, 264 rue Saint Pierre, 13385 Marseille Cedex 5,
France.
E-mail address: jj.grob@ap-hm.fr (J.J. Grob).
1
CGM and AS contributed equally to the work.

http://dx.doi.org/10.1016/j.ejca.2017.07.017
0959-8049/ª 2017 Elsevier Ltd. All rights reserved.
 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54 45

14.82 months, respectively).

Conclusion: In real-life MM pts with BMs, a strategy aiming at controlling BM with GK
together with TT and/or TT seems to achieve unprecedented survival rates.
ª 2017 Elsevier Ltd. All rights reserved.

1. Introduction control and avoid the occurrence of life-threatening

Targeted and immune treatments have dramatically
improved the course of metastatic melanoma (MM);
however, patients (pts) with brain metastases (BMs) still
carry a poor prognosis [1]. Despite similar response rate
in BMs as in extra-cerebral sites, the impact of BRAF-
inhibitors (BRAF-inh) on progression-free survival
and overall survival (OS) seems much lower than in pts
without BMs [2,3]. Similarly, although BMs can respond
to immunotherapy (IT), they are a difficult target for
ipilimumab [4], and few data are available for anti-PD1
[5,6,7]. Stereotactic radiosurgery (SRS) with Gamma-
Knife (GK) is used to control BMs in different can-
cers, providing both a survival and functional benefit [8].
SRS was initially performed only once in pts with a
maximal number of three BMs. Many centres have now
adopted an ‘on-demand’ strategy (GKOD), which can
address as many BMs as required and can be repeated
during follow-up when new BMs develop (Fig. 1)
[8e10].
Combining an upfront and repeated direct control of
BMs by SRS and a systemic treatment with targeted
therapies (TT) or IT, is a way to keep BMs under
situations or impairment of neurological function.
Meanwhile, IT or TT control extra-cerebral disease and
hopefully prevent new BMs. Furthermore, there may be
a real synergy between radiation and IT [11e13], and no
deleterious effect when combining GK with BRAF-inh
[14].
In the present paper, we analyse the survival in a
large retrospective cohort of MM pts, whose BMs were
systematically treated by GKOD as soon as they were
detected at the time of the introduction of TT and
checkpoint blockers in our practice.
2. Methods
2.1. Population
All pts with BMs treated in our department (La Timone,
Marseille, France) by a standardised strategy of GKOD
from the period of March 2010 to November 2015 were
included, independently of the medical treatment they
received thereafter. First GK was systematically per-
formed as soon as 1st BM was diagnosed and repeated
as many times and for as many BMs as required

Fig. 1. Strategy of Gamma-knife (GK) on demand. BM, brain metastasis.

46 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54

according to the disease course (Fig. 1). All BMs at a was retained for comparison. HRs for OS1GK were
given time were treated in a single session. Whenever adjusted for the prognostic variables available in both
ongoing, IT or TT was not stopped for GK. cohorts: age, sex, number of BMs, RPA and extra-
cerebral disease.
2.2. Gamma-Knife strategy and procedure
2.6. Statistics
All pts were treated with the Gamma-knife Perfexion
(Elekta, Stockholm, Sweden) in the University Hospital Statistical analysis was performed using PASW Statistics,
La Timone, Marseille, France. Magnetic resonance im- version 17.02 (IBM SPSS Inc., Chicago, IL, USA).
aging (MRI)ecompatible Leskell stereotactic G frame Continuous variables are expressed as means
standard
(Elekta Instrument) was applied under local anaesthesia. deviation (SD) or as median with range (min and
Neuroimaging consisted of a stereotactic 3D T1- max) and categorical variables as count and percentages.
weighted post-contrast acquisition (double dose of Means values were compared by Student’s t-test and
gadolinium)dmagnetisation-prepared RAGE gradient percentages by -square test (or Fisher’s exact test, as
ECHO or MP-RAGE, Siemens, 1.5T, Magnetom, ste- appropriate). Survival was estimated by the
reotactic magnetisation transfer axial cuts and a ste- KaplaneMeier method and compared by the log-rank
reotactic brain computed tomography (CT) scan.
Selection of dosimetric parameters (maximum dose, Table 1a
marginal isodose and number of isocentres) was made Baseline (at 1st GK) characteristics in the BRAF-mutated population
according to size, shape, location and relationship of the (n Z 105).
BM to critical structures. TT or IT after No TT or IT p
1st GK (n Z 76) after 1st GK
2.3. Follow-up data (n Z 29)
Age (year)
Brain MRI-scans were performed every 3 months after Mean 54.3 54.5 0.199
1st GK. Body CT-scans were performed every 2e3 <60 48 (63.2) 19 (65.5) 0$822
60 28 (36.8) 10 (34.5)
months, depending on the medical treatment delivered.
Sex
Male 42 (55.3) 18 (62.1) 0.529
2.4. Univariate and multivariate analysis for survival Female 34 (44.7) 11 (37.9)
Karnofsky
Survival was defined from the 1st GK (OS1GK) which >70% 74 (97.4) 28 (96.6) 0.822
70% 2 (2.6) 1 (3.4)
always took place within 4 weeks of the 1st diagnosis of
Number of BMs treated
BM. The following variables, collected at the time of 1st 1 23 (30.3) 11 (37.9) 0.570
GK, were included in univariate and multivariate ana- 2e3 23 (30.3) 5 (17.2)
lyses for OS1GK: age, sex, Karnofsky (>70 versus  70), 4e5 11 (14.5) 4 (13.8)
BRAF status, lactate dehydrogenase (LDH) level >5 19 (29) 9 (31)
RPA
(versus > upper limit of the normal), extension of
1 9 (11.8) 2 (6.9) 0.748
extra-cerebral metastatic disease (no other metastasis/ 2 65 (85.5) 26 (89.7)
only skin or nodes/lungs/other), number of BMs, need 3 2 (2.6) 1 (3.4)
for steroids, neurological signs (yes/no), recursive par- Neurological signs
titioning analysis (RPA) score [15] and treatments (ipi- Yes 18 (23.7) 7 (24.1) 0.961
No 58 (76.3) 22 (75.9)
limumab, anti-PD1 and BRAF-inh [
MEK-inh])
Use of corticosteroids
received before 1st GK on the one hand, and after 1st Yes 20 (26.3) 9 (31) 0.629
GK, on the other hand. Two separate Cox analyses were No 56 (73.7) 20 (69)
conducted in BRAFewild-type (WT) pts and in BRAF- LDH
mutated pts comparing four different groups: pts who ULN 60 (78.9) 14 (48.3) 0.002
>ULN (250 UI) 16 (21.1) 15 (51.7)
received only TT, IT but never TT, TT and thereafter
Extra-cerebral mets
IT and no TT or IT. No 9 (11.8) 3 (10.3) 0.445
Skin or nodes only 8 (10.5) 3 (10.3)
2.5. Comparison with a 2005 historical cohort treated by Lung 24 (31.6) 5 (17.2)
GKOD alone Other visceral mets 35 (46.1) 18 (62.1)
TT or IT before 1st GK
Yes 28 (36.8) 10 (34.5) 0.822
A 2005 cohort from our centre, treated by the same No 48 (63.2) 19 (65.5)
strategy of GKOD [9] at a time when no systemic drug
TT, targeted therapy (BRAF-inh
MEK-inh); IT, immunotherapy
impacted on survival, was used as an historical control. (ipilimumab or anti-PD1); GK, Gamma-Knife; BMs, brain metastases;
As pts presenting with >4 BMs were not treated by GK mets, metastasis; RPA, recursive partitioning analysis; ULN, upper
in 2005, only an adjusted subset from the 2016 cohort limit of the normal.
 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54 47

test. Univariate and multivariate Cox proportional haz- extra-cerebral metastases. No pt received whole-brain
ard regression models were used to estimate the hazard radiotherapy (WBRT) at any time. Among the 179 pts,
ratio (HR). The results are reported as two-sided p values BRAF status was mutated in 105 (60.7%), WT in 68
with 95% confidence intervals. All the tests were two- (39.3%) and unknown in 6 (3.3%) pts. Among the
sided. The statistical significance was defined as p < .05. 179 pts, 51 (28$5%) had received at least one TT or IT
before BM diagnosis.
3. Results
3.2. Treatments received after initial treatment by GK
3.1. Cohort
All 179 pts underwent a GK session at 1st BM diag-
nosis. Among them, 109 (60.3%) received either IT
During the period of March 2010 to November 2015,
(56 pts) or TT (65 pts) any time thereafter. Others did
179 consecutive pts with a median age of 59.3 years were
not, for a number of different reasons as follows: pre-
treated with GKOD (Table 1a and b). At the time of 1st
vious failure of IT or TT (n Z 17), exclusion criteria for
GK, 62 (34.6%) pts had a solitary BM, and 66 (36.9%)
clinical trials (n Z 8), no availability of these drugs at
pts had more than three BMs. Only 16 were free of
the time of 1st GK (n Z 29), early post-GK death
(n Z 4), absence of extra-cerebral disease (n Z 6) or
Table 1b decision of palliative care (n Z 6).
Baseline (at 1st GK) characteristics in the BRAFewild-type popula- Among the 105 BRAF-mutated pts, 76 (72.4%)
tion (n Z 68).
received TT or IT after 1st GK. Among the 68 BRAF-
TT or IT No TT or IT p WT pts, 32 (47%) received IT after 1st GK. Based on
after 1st GK after 1st GK post-GK treatments (Fig. 2), 70 pts (29 BRAF-mutated,
(n Z 32) (n Z 36) 36 BRAF-WT and 5 unknown) did not received IT or
Age (year) TT, 65 BRAF-mutated pts received BRAF
MEK (of
Mean 63.6 66.4 0.491 which 31 also received ipilimumab or anti-PD1
<60 11 (34.4) 10 (27.8) 0$557
60 21 (65.6) 26 (72.2)
thereafter) and 32 BRAF-WT pts received ipilimumab
Sex or anti-PD1 after 1st GK.
Male 16 (50) 20 (55.6) 0$647 The number of GK sessions per pt ranged from 1 to
Female 16 (50) 16 (44.4) 7, with a median number of three BMs treated by ses-
Karnofsky sion (min 1 and max 20). Neurological symptoms after
>70% 30 (93.8) 33 (91.7) 0$743
70% 2 (6.3) 3 (8.3)
GK appeared in the range of what can be expected
Number of BMs treated depending on the number and size of BMs. We presently
1 12 (37.5) 16 (44.4) 0$217 conduct a long-term follow-up of radiological images to
2e3 7 (21.9) 10 (27.8) document the rate of adverse radiation effects.
4e5 3 (9.4) 6 (16.7)
>5 10 (31.3) 4 (11.1)
RPA
3.3. Comparability of the groups
1 0 1 (2.8) 0$597
2 30 (93.8) 32 (88.9) At 1st GK, characteristics of pts who did or did not
3 2 (6.3) 3 (8.3) receive new treatment (NT) thereafter were well
Neurological signs balanced, except for an excess of neurological signs in
Yes 3 (9.4) 12 (33.3) 0$017
No 29 (90.6) 24 (66.7)
BRAF-WT pts who did not receive NT after 1st GK
Use of corticosteroids compared with BRAF-WT pts who received NT, and a
Yes 7 (21.9) 14 (38.9) 0$130 higher proportion of elevated LDH in pts who did not
No 25 (78.1) 22 (61.1) receive NT after 1st GK, whatever the mutation status,
LDH compared with pts who received NT (Table 1a and b). In
ULN 24 (75) 16 (44.4) 0.011
>ULN (250 UI) 8 (25) 20 (55.6)
BRAF-mutated pts, baseline characteristics were not
Extra-cerebral mets different for most demographics and prognostic vari-
No 3 (9.4) 1 (2.8) 0$104 ables, according to whether they were treated after GK
Skin or nodes only 4 (12.5) 5 (13.9) either with TT alone, IT alone or IT and TT (data not
Lung 12 (37.5) 6 (16.7) shown). Those who were treated by IT alone after 1st
Other visceral mets 13 (40.6) 24 (66.7)
TT or IT before 1st GK
GK had more frequently received a NT before 1st GK.
Yes 7 (21.9) 6 (16.7) 0$586
No 25 (78.1) 30 (83.3) 3.4. Survival after 1st GK
TT: targeted therapy (BRAF-inh
MEK-inh); IT: immunotherapy
(ipilimumab or anti-PD1); GK, Gamma-Knife; BMs: brain metastasis; With a median follow-up of 9.8 months
mets: metastasis; RPA, recursive partitioning analysis; ULN: upper (1.41e79.21 months), median OSGK1 was 6.79 months
limit of the normal. in the whole cohort (179 pts) treated by GK,
48 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54

Fig. 2. Flow chart of new treatments (NT) received after 1st GK*. Mut, mutated; IPI, ipilimumab; NT, new treatment; WT, wild-type;
UNK, unknown; IT, immunotherapy; BRAFI, BRAF-inhibitor.

10.95 months in the 109 pts who additionally received an
IT or a TT and 2.29 months in those who did not.
In the 109 pts who received IT or TT after GK,
median OS (1-year and 2-year survival rates) was
14.46 months (66.7% and 43.4%) in pts with one BM at
1st diagnosis, 8.85 months (46.4% and 31%) in pts with
2e3 BMs and 7.25 months (37.2% and 11.9%) in pts
with >3 BMs, respectively. For comparison, in the
70 pts who did not receive IT or TT, median OS was
3.57 months, 3.1 months and 1.63 months, in those with
1, 2e3 or >3 BMs, respectively (Fig. 3).
In these 109 pts, median OSGK1 (1-year and 2-year
survival rates) was 7.31 months (35.3% and 8.8%) in
the BRAF-mutated pts treated by BRAF
MEK and
12.26 months (63.6% and 23.9%) in the BRAF-WT
treated by anti-PD1.
Patients who were NT naı̈ve before 1st GK did not
have significantly different outcomes compared with
patients who received NT before 1st GK (data not
shown).
Multivariate analyses for OS1GK in the whole cohort
showed that initial number of BMs (>3) and elevated
LDH were pejorative risk factors, whereas treatment with
IT or TT after 1st GK was significantly and highly pro-
tective (Table 2). More specifically, in the BRAF-mutated
pts, post-GK treatments with TT or TT and IT were

Fig. 3. Survival from 1st GK according to the treatment by TT or IT after 1st GK (n Z 179). 3a: All population. 3b: By initial number of
brain metastases. GK, Gamma-Knife; IT, immunotherapy; OS1GK, survival from 1st GK; TT, targeted therapy.
 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54 49

Table 2
Impact of baseline patient and disease characteristics (at 1st GK) on the survival from 1st GK: Univariate and multivariate analyses.
Variables Univariate analysis Multivariate analysis
HR (death) p HR (death) p
Age (years)
<60 1 0.97 1 0.136
60 1.10 (0.73e1.38) 0.75 (0.51e1.09)
Sex
Male 1 0.511 1 0.328
Female 0.89 (0.65e1.24) 0.84 (0.58e1.19)
Karnofsky
>70 1 <0$001 1 0.007
70 5.26 (2.62e10.58) 3.22 (1.38e7.50)
BRAF statusa
Mutation V600 1 0$030 1 0.111
Wild-type 1.44 (1.04e2.01) 1.38 (0.93e2.05)
Number of BMs treated
1 1 1
2e3 1.27 (0.83e1.95) 0.262 0.93 (0.58e1.50) 0.776
>3 2.22 (1.52e3.25) <0$001 1$71 (1$12e2$63) 0$013
RPAb
1 1
2 1.21 (0.65e2.25) 0.544
3 6.29 (2.54e15.6) <0$001
Neurological signs
No 1 <0$001 1
Yes 1.98 (1.37e2.84) 1.63 (1.04e2.54) 0.032
Use of corticosteroids
No 1 <0.001 1
Yes 1$88 (1.33e2.64) 1.13 (0.75e1.72) 0.548
LDH
Normal 1 <0$001 1 <0$001
>ULN 4.60 (3.21e6.57) 2$70 (1$77ee4$11)
Extra-cerebral mets
No 1 1
Skin or nodes only 0.67 (0.32e1.38) 0.275 0.52 (0.23e1.15) 0.107
Lung 1.30 (0.72e2.35) 0.393 1.36 (0.71e2.61) 0.349
Other visceral mets 2.18 (1.25e3.80) 0.006 1.78 (0.92e3.44) 0.086
Treatment with TT or IT before 1st GK
No 1 0.525 1
Yes 1.12 (0.79e1.60) 099 (0.65e1.49) 0.947
Treatment with TT or IT after 1st GK
No 1 <0$001 1 <0$001
Yes 0$39 (0$28e0$53) 0$33 (0$22e0$49)
BMs, brain metastasis, HR, hazard ratio; RPA, recursive partitioning analysis; TT, targeted therapies, IT, immunotherapies, ULN, upper limit of
the normal, NS: non-significant.
Bold values indicate statistically significative data (p  0.05).
a
Exclusion of six patients with unknown BRAF status.
b
RPA was not included in the multivariate model to avoid collinearity effect with the Karnofsky.

significantly and strongly protective: HR Z 0.60 and 0.34,
respectively (Table 3 and Fig. 4). In the BRAF-WT pts,
post-GK treatments with ipilimumab or anti-PD1 were
significantly and strongly protective: HR Z 0.48 0.33,
respectively (Table 3 and Fig. 5).
Although it is impossible to state what the ultimate
cause of death was, 85 pts had neurological signs at last
follow-up before death: 52 (47%) in the NT group and
33 (50%) in the no NT group. Seventy pts had additional
GK sessions: 62 (56.8%) in the NT group and eight
(12.3%) in the no NT group. There were no differences
in the number of subsequent GK sessions according to
the type of NT administered (data not shown). Thirteen
pts underwent subsequent craniotomy: nine (8.2%) in
the NT group and four (6.1%) in the no NT group.
3.5. Comparison with the historical cohort treated by
GKOD alone
Baseline characteristics of the 2005 cohort and the
matched sub-cohort from 2016 were balanced for most
of the relevant characteristics, despite a slight difference
in age and sex distribution (data not shown). Median
OSGK1 was longer in the 2016 cohort than in the 2005
cohort, 8.40 months versus 5$02 months (p < .001), with
a HR Z 0$54 (0$41e0$72, p < .001; Fig. 6).
50 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54

Table 3
Medical treatment effect on survival after 1st GK: Univariate and multivariate analyses in BRAF-mutated patients and in BRAF-WT patients.
Median OS Univariate analysis Multivariate analysisa
(months) HR p HRa 95% CI p
BRAF-mutated patients (n Z 105)
No targeted or immune therapies (n Z 29) 2.62 1 1
BRAF
MEK-inh alone (n Z 34) 7.31 0.60 0.063 0.42 0.22 0.8 0.009
BRAF
MEK-inh and immunotherapy (n Z 31) 14.82 0.34 <0.001 0$15 0.07 0.31 <0.001
Immunotherapy alone (n Z 11) NR 0.18 0.001 0$13 0.04 0.39 <0.001
BRAF-WT patients (n Z 68)
No immunotherapy (n Z 36) 2.29 1 1
Anti-PD1 alone (n Z 11) 12.26 0.33 0.022 0.16 0.05 0.48 0.001
Ipilimumab alone (n Z 17) 8.62 0.48 0.016 0.18 0.07 0.44 <0.001
Ipilimumab þ anti-PD1 (n Z 4) 14.07 0.19 0.023 0.28 0.05 1.38 0.12
CI, confidence interval; OS, overall survival; NR, not reached; WT, wild-type.
a
Multivariate analysis including demographics variables (age and sex), karnofsky (>70 versus  70), LDH level (N versus > ULN); extra-
cerebral metastatic disease (in four classes: no other metastasis/only skin or nodes/lungs/other visceral metastasis); neurological severity
including number of BMs, need for steroids, presence of neurological signs and recursive partitioning analysis and treatments (ipilimumab, anti-
PD1, and BRAF-inh [
MEK-inh]) before 1st GK.

Fig. 4. Survival from 1st GK (OS1GK) in BRAF-mutated patients according to the treatment received after 1st GK (n Z 105).
 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54
Fig. 5. Survival from 1st GK (OS1GK) in BRAF-WT patients according to immunotherapy received after 1st GK (n Z 68).
4. Discussion
To our knowledge, this is the largest cohort assessing
survival of non-selected MM pts treated by TT or IT
after a systematic upfront treatment of their BMs by
GK. In this real-life cohort of BM pts usually excluded
from clinical trials, the survival after BMs seems quite
favourable as compared with a recent review which de-
scribes a median survival of 7.1 months in the most
favourable group and 2.3 months in the least favourable
one [1]. In pts treated by GK, who thereafter received
TT or IT, 1- and 2-year survival rates after BMs ranged
between 66.7% and 43.4% in the best group with only
one BM at first presentation and 37.2% and 11.9% in the
worse group initially presenting with more than three
BMs, respectively. The risk of death was decreased by
40e82% depending on the TT or IT received, in com-
parison with those treated by GK who did not receive IT
or TT. The best survival was apparently obtained in pts
51
who received anti-PD1 at any time, upfront or after
BRAF-inh. The benefit with BRAF-inh seems a bit
lower, but most pts in this cohort did not receive the
combination of BRAF þ MEK-inh.
To date, only two retrospective studies have assessed
the survival of pts treated with SRS and new systemic
therapies. In the first study, 96 pts were treated with SRS
within 3 months of receiving systemic therapy; signifi-
cant differences were noted in OS according to systemic
treatments with 1-year OS rates following SRS of 48%
(anti-PD1), 50% (anti-CTLA4), 65% (BRAF/MEK-
inh), 24% (BRAF-inh) and 10% (chemotherapy) [16]. In
the second study [7], 79 pts received systemic therapy
within 6 weeks of SRS, median OS was 10.8 months in
the 26 pts who did not receive any systemic therapy,
17.8 months in the 39 pts treated with BRAF
MEK
inhibitors, 7.5 months in the 28 pts who received anti-
CTLA4 and 20.4 months in the 11 pts who received an
anti-PD1. It should be mentioned that almost half of
52 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54
Fig. 6. Overall survival after 1st GK (OS1GK) in the 2006 cohort (n Z 106) versus the matched sub-cohort of 2016 (n Z 121). HR, hazard
ratio.
these pts had only a single BM and that WBRT was
used in addition to SRS in 40%, which makes difficult
comparison with our cohort including a large subset of
more severe cases and cases treated by SRS only.
However, these data are all in line with our very
favourable results.
Other small reports addressing the combination of
BRAF-inh and SRS, using BRAF-WT pts treated by
SRS alone as a control, showed a benefit in terms of
local control and survival [17,18]. In two retrospective
studies of pts, ipilimumab after SRS did not show a
clear benefit [13,19]. Only one small report of 26 pts
(including 10 treated in the setting of tumoural bed-
efractionated RT following BM surgical resection)
addressed the combination of SRS with anti-PD1 and
did not detect any toxicity issue [6].
In addition to the advantage of providing high local
control of BMs with SRS enabling to avoid or delay
clinical neurological deterioration, the survival in our
cohort seems also superior to the few results published in
pts treated by TT or IT without upfront SRS. Indeed,
dabrafenib achieved a median OS of 7.7 months in pre-
viously untreated pts [3], whereas vemurafenib achieved
an OS of 5.3 months in pts with symptomatic pretreated
BMs [20]. In a phase II study with ipilimumab, median
survival was 7 months in the most favourable group
(asymptomatic pts without corticosteroids) and
3.8 months in the other [21]. Even more scant data are
available for anti-PD1. In the early analysis of an
ongoing phase II study of pembrolizumab including
18 MM asymptomatic pts with BMs, a cerebral response
was achieved in 22% while median OS was not reached
after a median follow-up of 11.6 months [22].
Among pts treated by SRS, those who received NT
lived longer but had the same proportion of neurolog-
ical signs at the time of last evaluation than those who
did not receive NT which may seem paradoxical. It is
thus likely that, without NT, there is often a rapid death
from extra-cerebral disease, despite BM control by SRS.
Conversely, in those receiving NT, extra-cerebral disease
is often controlled, thus making cerebral disease the
ultimate cause of death in most of the long survivors.
Our study has the double advantage to be representa-
tive of real-life cases of BMs, including pts with multiple
BMs upfront and to use a homogeneous standardised
strategy of GKOD without any WBRT. However, this
study has also the limits of a retrospective cohort, in which
systemic treatments were neither randomly allocated nor
given in a standardised schedule, and the limits of the
rather small number of pts. Systemic treatments were
indeed given to pts from the moment the new molecules
became available for clinical practice, or as a function of
inclusion criteria of trials, while the respective timing of
each treatment was case-dependent. As compared with pts
treated by GK, who did not receive IT or TT, the char-
acteristics of the pts who additionally received TT or IT
 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54
were significantly better for two major prognostic factors
LDH and neurological signs. The direct benefit of IT and
TT is thus difficult to interpret. However, the multivariate
analysis suggests that the superiority of the combination
of GKOD þ IT or TT over GKOD alone is not only due to a
more severe disease in those who did not receive IT or TT.
When all confounding factors are taken into account, the
protective effect of BRAF-inh on the survival of pts
treated by GKOD remains significant and important
(HR Z 0.60). The protective effect of ipilimumab and/or
anti-PD1 in pts is even higher (HR Z 0.34e0.19 according
to the groups). Moreover, an adjusted comparison with a
historical cohort managed by the same GKOD strategy in
the same centre, before the availability of these effective
treatments, shows a clear survival advantage in favour of
IT or TT. Taken together, these data strongly suggest that
systemic treatments by IT and TT provide a clear addi-
tional survival benefit over the mere local and repeated
control of BMs by GKOD.
Our retrospective study is not suitable to address the
issue of the best timing for the combination of SRS and
systemic treatments, since there was no standardisation
of the timing of different treatments. However, there is
probably no ideal absolute schedule for all pts. The OS
benefit probably results less from the direct potentiation
of IT or TT by GK or vice-versa, than from reciprocal
optimisation of pt’s management adapted to each situ-
ation. In many cases, the upfront direct control of BMs
by GK is the unique way to prevent immediate death or
functional neurological deterioration and to offer the
time window during which IT or TT can be started. In
other cases, when brain is the privileged site of recur-
rence after IT or TT, GK often provides the only chance
for a pt to survive long enough before a next line of IT
or TT is introduced. Finally, even when there is proper
control of BMs by GK, there is no possible long sur-
vival, unless active medical molecules like TT or IT are
given to fix the extra-cerebral metastatic disease.
Compared to our past experience with GK, we did
not notice an increased incidence of neurological com-
plications during the follow-up period, although neither
BRAF-inh nor IT was stopped during the GK sessions.
As the problem of a potential increased toxicity of SRS
by IT and TT has been raised, a more detailed radio-
logical investigation is ongoing, with a longer follow-up,
especially in anti-PD1 treated pts. However, given the
OS benefit, we suggest that until proven otherwise this
strategy should be applied to any GK eligible BM pt.
In conclusion, this study shows that combining a
systematic management of BMs by GKOD and a treat-
ment by IT or TT allows to achieve unprecedented
survival rates so far in this population excluded from
most trials. Our data suggest that anti-PD1 seems to be
of particular interest in this context, but the ideal timing
of GK for the best synergy with IT and the assessment
of long-term radiotoxicity warrant further study.
53
Conflict of interest statement
None declared.
References
[1] Goyal S, Silk AW, Tian S, et al. Clinical management of multiple
melanoma brain metastases: a systematic review. JAMA Oncol
2015;1:668e76.
[2] Gibney GT, Gauthier G, Ayas C, et al. Treatment patterns and
outcomes in BRAF V600E-mutant melanoma patients with brain
metastases receiving vemurafenib in the real-world setting. Cancer
Med 2015;4:1205e13.
[3] Long GV, Trefzer U, Davies MA, et al. Dabrafenib in patients
with Val600Glu or Val600Lys BRAF-mutant melanoma meta-
static to the brain (BREAK-MB): a multicentre, open-label, phase
2 trial. Lancet Oncol 2012;13:1087e95.
[4] Ahmad SS, Qian W, Ellis S, et al. Ipilimumab in the real world:
the UK expanded access programme experience in previously
treated advanced melanoma patients. Melanoma Res 2015;25.
432e442V.
[5] Kluger HM. Safety and activity of pembrolizumab in melanoma
patients with untreated brain metastases. In: ASCO annual
meeting abstract number: 9009; 2015.
[6] Ahmed KA, Stallworth DG, Kim Y, et al. Clinical outcomes of
melanoma brain metastases treated with stereotactic radiation
and anti-PD-1 therapy. Ann Oncol 2016;27:434e41.
[7] Choong ES, Lo S, Drummond M, et al. Survival of patients with
melanoma brain metastasis treated with stereotactic radiosurgery
and active systemic drug therapies. Eur J Cancer 2017 Feb 22;75:
169e78.
[8] Lippitz B, Lindquist C, Paddick I, Peterson D, O’Neill K,
Beaney R. Stereotactic radiosurgery in the treatment of brain me-
tastases: the current evidence. Cancer Treat Rev 2014;40:48e59.
[9] Gaudy-Marqueste C, Regis JM, Muracciole X, et al. Gamma-
Knife radiosurgery in the management of melanoma patients with
brain metastases: a series of 106 patients without whole brain
radiotherapy. Int J Radiat Oncol Biol Phys 2006:809e16.
[10] Marchan EM, Sheehan J. Stereotactic radiosurgery of brain
metastasis from melanoma. Prog Neurol Surg 2012;25:176e89.
[11] Sharabi AB, Nirschl CJ, Kochel CM, et al. Stereotactic radiation
therapy augments antigen-apecific PD-1-mediated antitumor im-
mune responses via cross-presentation of tumor antigen. Cancer
Immunol Res 2015;3:345e55.
[12] Okwan-Duodu D, Pollack BP, Lawson D, Khan MK. Role of
radiation therapy as immune activator in the era of modern
immunotherapy or metastatic malignant melanoma. Am J Clin
Oncol 2015;38:119e25.
[13] Patel KR, Shoukat S, Oliver DE, et al. Ipilimumab and stereo-
tactic radiosurgery versus stereotactic radiosurgery alone for
newly diagnosed melanoma brain metastases. Am J Clin Oncol
2015;May 16.
[14] Gaudy-Marqueste C, Carron R, Delsanti C, et al. On demand
Gamma-Knife strategy can be safely combined with BRAF in-
hibitors for the treatment of melanoma brain metastases. Ann
Oncol 2014;25:2086e91.
[15] Gaspar L, Scott C, Rotman M, et al. Recursive partitioning
analysis (RPA) of prognostic factors in three Radiation Therapy
Oncology Group (RTOG) brain metastases trials. Int J Radiat
Oncol Biol Phys 1997;37:745e51.
[16] Ahmed KA, Abuodeh YA, Echevarria MI, et al. Clinical out-
comes of melanoma brain metastases treated with stereotactic
radiosurgery and anti-PD-1 therapy, anti-CTLA-4 therapy,
BRAF/MEK inhibitors,BRAF inhibitor, or conventional
chemotherapy. Ann Oncol 2016 Dec;27(12):2288e94.
54 C. Gaudy-Marqueste et al. / European Journal of Cancer 84 (2017) 44e54
[17] Ly D, Bagshaw HP, Anker CJ, et al. Local control after stereo-
tactic radiosurgery for brain metastases in patients with mela-
noma with and without BRAF mutation and treatment. J
Neurosurg 2015;123:395e401.
[18] Wolf A, Zia S, Verma R, et al. Impact on overall survival of the
combination of BRAF inhibitors and stereotactic radiosurgery in
patients with melanoma brain metastases. J Neurooncol 2016;127:
607e15.
[19] Kiess AP, Wolchok JD, Barker CA, et al. Stereotactic radio-
surgery for melanoma brain metastases in patients receiving ipi-
limumab: safety profile and efficacy of combined treatment. Int J
Radiat Oncol Biol Phys 2015;92:368e75.
[20] Dummer R, Goldinger SM, Turtschi CP, et al. Vemurafenib in
patients with BRAF (V600) mutation-positive melanoma with
symptomatic brain metastases: final results of an open-label pilot
study. Eur J Cancer 2014;50:611e21.
[21] Margolin K, Ernstoff MS, Hamid O, et al. Ipilimumab in patients
with melanoma and brain metastases: an open-label, phase 2 trial.
Lancet Oncol 2012;13:459e65.
[22] Goldberg SB, Gettinger SN, Mahajan A, et al. Pembrolizumab
for patients with melanoma or non-small-cell lung cancer and
untreated brain metastases: early analysis of a non-randomised,
open-label, phase 2 trial. Lancet Oncol 2016 Jul;17(7):976e83.