Acta Oncologica

ISSN: 0284-186X (Print) 1651-226X (Online) Journal homepage: www.tandfonline.com/journals/ionc20

Risk factors for brain metastases in patients with

metastatic colorectal cancer

Troels Dreier Christensen, Jesper Andreas Palshof, Finn Ole Larsen, Estrid
Høgdall, Tim Svenstrup Poulsen, Per Pfeiffer, Benny Vittrup Jensen, Mette
Karen Yilmaz, Ib Jarle Christensen & Dorte Nielsen

To cite this article: Troels Dreier Christensen, Jesper Andreas Palshof, Finn Ole Larsen,
Estrid Høgdall, Tim Svenstrup Poulsen, Per Pfeiffer, Benny Vittrup Jensen, Mette Karen
Yilmaz, Ib Jarle Christensen & Dorte Nielsen (2017) Risk factors for brain metastases
in patients with metastatic colorectal cancer, Acta Oncologica, 56:5, 639-645, DOI:
10.1080/0284186X.2017.1290272

To link to this article: https://doi.org/10.1080/0284186X.2017.1290272

Published online: 22 Feb 2017.

Submit your article to this journal

Article views: 2588

View related articles

View Crossmark data

Citing articles: 8 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=ionc20
ACTA ONCOLOGICA, 2017
VOL. 56, NO. 5, 639–645
http://dx.doi.org/10.1080/0284186X.2017.1290272

ORIGINAL ARTICLE

Risk factors for brain metastases in patients with metastatic colorectal cancer
Troels Dreier Christensena, Jesper Andreas Palshofa, Finn Ole Larsena, Estrid Høgdallb, Tim Svenstrup Poulsenb,
Per Pfeifferc, Benny Vittrup Jensena, Mette Karen Yilmazd, Ib Jarle Christensenb and Dorte Nielsena
a
Department of Oncology, Herlev and Gentofte Hospital, University of Copenhagen, Herlev, Denmark; bDepartment of Pathology, Herlev and
Gentofte Hospital, University of Copenhagen, Herlev, Denmark; cDepartment of Oncology, Odense University Hospital, Odense, Denmark;
d
Department of Oncology, Aalborg University Hospital, Aalborg, Denmark

ABSTRACT ARTICLE HISTORY

Background: Brain metastases (BM) from colorectal cancer (CRC) are rare, but the incidence is Received 22 October 2016
suspected to rise as treatment of metastatic (m) CRC improves. The aim of this study was to identify Accepted 30 January 2017
possible biological and clinical characteristics at initial presentation of mCRC that could predict later
risk of developing BM. Furthermore, we wished to estimate the incidence of BM in long-term surviving
patients.
Material and methods: We conducted a retrospective study on a Danish multicenter cohort of
patients with mCRC who received cetuximab and irinotecan (CetIri) as third-line treatment. All patients
had previously progression on 5-FU, irinotecan and oxaliplatin containing regimens and received CetIri
treatment independent of RAS mutations status. We subsequently performed KRAS, NRAS, BRAF,
PIK3CA, PTEN, ERBB2 and EGFR sequencing of DNA extracted from primary tumor tissue.
Results: Totally, 480 patients were included in our study. BM were diagnosed in 42 [8.8%; 95% confi-
dence interval (CI) 6.4–11.6%] patients. Patients with BM had a significantly longer survival from mCRC
diagnosis than non-BM patients (median ¼ 32 versus 28 months, p ¼ 0.001). On univariate cox regres-
sion analysis, the risk of developing BM was significantly increased in patients with rectal cancer
(HR ¼ 3.9; 95% CI ¼ 1.2–13.3), metachronous metastatic disease (HR ¼ 2.3; 95% CI ¼ 1.2–4.4) and lung
metastases (HR ¼ 4.2; 95% CI ¼ 2.2–7.9). On multivariate cox regression analysis only lung metastases
were significantly associated BM (HR ¼ 3.5; 95% CI ¼ 1.8–6.8). None of the investigated mutations were
associated with BM.
Conclusion: The incidence of BM was 8.8% in patients with mCRC who received third-line therapy.
The most important risk factor for developing BM was lung metastases. Furthermore, rectal cancer,
metachronous metastatic disease and long survival were linked to BM development.

Introduction One way to achieve earlier detection of BM is to know risk

factors for developing BM. Patients with CRC usually develop
Brain metastases (BM) are commonly diagnosed in patients
BM late in the course of metastatic disease, and the patients
with lung cancer (20% of patients), malignant melanoma
usually have metastases to other organs before BM are diag-
(7%) and breast cancer (5%) [1], but BM developed from
nosed [2,3,11]. Several risk factors have been proposed and
colorectal cancer (CRC) are relatively rare with a reported
especially lung metastases and rectal cancer have been asso-
incidence of 1–2% [2,3]. However, it has been suggested that
the reported incidence of BM from CRC will increase due to ciated with development of BM [3]. Specific gene mutations
improved diagnostics and prolonged survival of patients, but in the tumor could also potentially lead to an increased risk
this is not well-documented [2]. of BM. Mutation in the PIK3CA, RAS (KRAS and NRAS) and
A diagnosis of BM is associated with increased morbidity BRAF genes have all been investigated, but mostly in smaller
and mortality. In patients with CRC, the reported median sur- studies [12–17], and only RAS mutations have been shown to
vival after BM diagnosis is 2.0–7.4 months, and only 10–15% be significantly associated with BM in two studies [12,13].
survive more than 1 year [4–7]. Patients who receive neuro- However, the few studies examining the predictive potential
surgical intervention as part of their treatment have a more of characteristics and mutations on later BM development in
favorable outcome [8,9], but this intervention is often patients with CRC are small and retrospective.
reserved to patients with a good prognosis with a limited The aim of this study was to examine biological and clin-
number of BM, good control of systemic disease and good ical characteristics at initial presentation of metastatic disease
performance status [4,8]. It has been suggested that early that could predict later development of BM, and to estimate
detection of BM might increase the number of treatment the incidence of BM in long-surviving patients with meta-
options available and improve prognosis [10]. static CRC (mCRC).

CONTACT Troels Dreier Christensen [email protected] Department of Oncology, Herlev and Gentofte Hospital, Herlev Ringvej 75,
DK-2730 Herlev, Denmark
ß 2017 Acta Oncologica Foundation
640 T. D. CHRISTENSEN ET AL.

Material and methods and/or medical history (Figure 1: consort diagram). Thus, the
database encompassed 480 patients.
Patients KRAS mutation status was not evaluated routinely during
We used a national biobank and database encompassing the period of treatment, and all patients therefore
consecutive patients with mCRC who received third-line received CetIri independently of KRAS mutation status.
treatment with irinotecan in combination with the EGFR Routine follow-up screening for BM was not conducted, and
inhibitor, cetuximab (CetIri). This cohort was chosen because brain scans were only performed if patients were suspected
we had tumor tissue from most of the patients and the to have BM.
cohort had a long survival with a suspected high incidence
of BM. All patients received CetIri between 1 January 2005
Data collection and registration
and 1 August 2008 at the oncology wards of three Danish
hospitals (Herlev and Gentofte, Odense and Aalborg The database encompassed detailed information about
Hospital). All patients were prior exposed to oxaliplatin, irino- demographics, surgical procedures, cancer treatments,
tecan and fluoropyrimidine. A total of 531 patients with response and localization of metastases until progression on
mCRC were eligible for inclusion, but 51 patients were CetIri, death or start of following line of therapy. Survival sta-
excluded because it was not possible to retrieve tissue tus was updated on 1 December 2015.

Figure 1. Consort diagram: Patients included in study and mutation analysis performed.
 ACTA ONCOLOGICA 641

In order to obtain and validate information about BM we extractor (QIA Symphony, Qiagen, Venlo, the Netherlands).
searched ‘CSC Sundhedsdatabanken’ through the web-based Sequencing was performed using 10 ng DNA and a primer
‘CSC Patientindex’. ‘CSC Sundhedsdatabanken’ is a nation- panel covering relevant gene exons and codons (Table 1).
wide Danish patient register with information about all diag- Briefly, we used Ion AmpliSeq Library Kit 2.0 for library prep-
noses, treatments and diagnostic procedures that each aration, Ion PGMTM Template OT2 200 Kit and Ion OneTouch
patient has received. Using ‘CSC Sundhedsdatabanken’, we Instrument for emulsion PCR and Ion PGMTM Sequencing 200
registered if the patients had CT or MRI scans of the brain Kit v2 with Ion 318TM Chip Kit v2 and Ion Personal Genome
performed after diagnosis of CRC. Afterwards, brain scan Machine (PGM) as the sequencing platform (Ion Torrent, Life
reports were retrieved either through web based databases Technology, Carlsbad, CA). A background noise at 5% was
(‘E-journal’, ‘WebRis/iSite’, ‘Patientindex’) or by contacting the chosen and at least 100 reads per amplicon were required
radiological wards at the hospital where the scan was for analysis. Interpretation of the results was performed by
performed. an experienced molecular biologist, and genes were coded
as mutated if mutations were identified (Table 1).
Ethics Of the remaining 407 patients, we were able to perform
extended mutation analysis on 375 patients. Thus, KRAS,
The study was approved by the Ethical Committee of NRAS and BRAF status was available from 448 patients, and
Copenhagen (H-KA-20060094). PIK3CA, EGFR, ERBB2 and PTEN from 375 patients (Figure 1).

DNA sequencing and interpretation

Statistical analysis
The biobank contained formalin fixed—paraffin embedded
To diminish the risk of describing associations that were the
(FFPE) tissue blocks from the patients’ primary tumor and/or
results of reverse causality, we only included patient charac-
metastases.
DNA from 181 of the patients’ tumors had previously teristics obtainable at diagnosis of mCRC in the analysis.
been analyzed for KRAS (codon 12 and 13) and BRAF V600E Patient characteristics included in analysis were: sex (male/
mutations. KRAS had been analyzed with TheraScreenV KRAS
R female); age at primary diagnosis, primary tumor location
mutation kit (DxS Ltd, Manchester, UK). BRAF pyrosequencing (right colon: cecum to transverse colon/left colon: left colic
(Pyromark Q24) was performed using primers as described flexure to sigmoid colon/rectum); resection of primary tumor
by Richman et al. [18] with a sensitivity of 5% to detect a (yes/no); timing of metastatic disease (synchronous: within
mutation. Sixty-eight patients were KRAS exon 2 mutated, 2 months after diagnosis of primary/metachronous); number
and five patients were BRAF V600E mutated. of metastatic sites at mCRC diagnosis (one site/two sites/
For the 407 patients whose tumor was not analyzed previ- three or more sites), and specific metastatic sites at diagnosis
ously or the tumor was KRAS or BRAF wild-type (WT), we per- of mCRC: primary tumor or local recurrence (yes/no), node
formed a new extended mutation analysis covering coding metastases (yes/no), liver metastases (yes/no), lung metasta-
regions of KRAS, NRAS, BRAF, PIK3CA, EGFR, ERBB2 and PTEN. ses (yes/no), and peritoneal metastases (yes/no).
For the purpose of sequencing, areas of high cancer cell con- We also analyzed the effect of different mutations on the
centration was identified from tissue blocks by microscopy of risk of developing BM. To make our analysis comparable with
hematoxylin and eosin stained slides and tissue was dis- previous studies [12,13], we performed a stepwise analysis of
sected using a 1 mm disposable ‘Miltex biopsy punch’. The RAS mutations by analyzing KRAS exon 2 (codon 12 and
samples were deparaffinized using xylene and subsequently codon 13) mutation alone, KRAS exon 2, 3 and 4 mutations
DNA was extracted using a magnetic particle based DNA combined, and all KRAS and NRAS mutations pooled in a RAS

Table 1. Overview of mutation analysis.

Genes Exons and codons covered in analysis Mutations defined as relevant Mutations detected
KRAS Exon 2, 3 and 4 (codon 4-37, 41-66 and Any mutations in codon 12, 13, 59, 61, G12 (N ¼ 152), G13 (N ¼ 17), A59
112-150) 117 and 146 (N ¼ 2), Q61 (N ¼ 6), K117 (N ¼ 2),
A146 (N ¼ 12)
NRAS Exon 2, 3 and 4 (codon 4-37, 53-90, Any mutations in codons 12, 13, 59, 61, G12 (N ¼ 5), Q61 (N ¼ 10)
98-124 and 138-150) 117 and 146
BRAF Exon 11 and 15 (codon 439-472 and V600E V600E (N ¼ 30)
582-610)
PIK3CA Exon 9 and 20 (codon 523-549 and P539, E542, E545, Q546, D549, Y1021, E542 (N ¼ 18), E545 (N ¼ 18), Q546
1018-1050) M1043, H1047 and G1049 (N ¼ 8), M1043 (N ¼ 1), H1047
(N ¼ 14), G1049 (N ¼ 2)
EGFR Exon 12, 18, 19, 20 and 21 (codon S492, G719, G724, W731, P733, E734, S492 (N ¼ 1), G724 (N ¼ 1), G735
474-499, 696-725, 729-761, 762-799 and G735, V742, K745, E746, L747, A750, (N ¼ 1)
824-875) S752, P753, D761, A763 and S768
ERBB2 Exon 8, 19, 20, 21 and 22 (codon 302-340, G309, S310, L755, D769, G776, V777, L785, V777 (N ¼ 1)
753-769, 770-796, 840-881 and 884-908) V842 and R896
PTEN Exon 6, 7 and 8 (165-183, 213-215, R173, R233, R234, K267, N323 and R335 R173 (N ¼ 2), R335 (N ¼ 1)
232-267, 283-299 and 313-342)
N ¼ number of tumors with the specific mutation.
642 T. D. CHRISTENSEN ET AL.

mutations group. We also analyzed the associations between peritoneal metastases. Time from diagnosis of CRC to mCRC
BM and BRAF, PIK3CA, PTEN, ERBB2 and EGFR mutations. was longer in patients who developed BM than those who
Statistical analyzes of our data were conducted using IBM did not (median ¼ 12.5 versus 0 months, p ¼ 0.001).
SPSS Statistics 22 (SPSS Inc., Chicago, IL). For all analyzes, a Univariate cox regression analysis was used to evaluate
two sided p < 0.05 was considered statistically significant. A the effect of the different characteristics and mutations on
95% confidence intervals (CI) were calculated for the inci- risk of developing BM. Three patient characteristics were
dence of mutations and BM, using the Clopper–Pearson found to increase risk of BM significantly: rectal cancer
method for binomial data. We analyzed potential associations (HR ¼ 3.912; 95% CI ¼ 1.154–13.262) compared to right colon
between patient clinicopathologic characteristics and BM cancer, metachronous metastatic disease (HR ¼ 2.296; 95%
development by performing chi-squared test or Fisher’s exact CI ¼ 1.188–4.439) and lung metastases at mCRC diagnosis
test on categorical variables and Mann–Whitney U-test on (HR ¼ 4.196; 95% CI ¼ 2.216–7.945). None of the mutations
continuous variables. (RAS, BRAF and PIK3CA) affected the risk of BM significantly.
Lastly, univariate cox proportional hazards model was When the three significant variables were selected for
used to evaluate the effect of the different characteristics on multivariable cox regression (Table 3), only presence of lung
hazard of developing BM, using months from mCRC diagno- metastases (HR ¼ 3.096, p < 0.0005) had a significant effect
sis to detection of BM as time variable. Patients who did not on risk of BM.
develop BM before their death were censored at time of
death or date of last follow-up. Afterwards, variables having
a significant effect on univariate analysis (p < 0.05), were
Discussion
selected for multivariable cox proportional hazard model. In a large cohort of long-term surviving patients with mCRC
who all received third-line therapy, we described the inci-
dence of BM, and tried to identify tumor and patient charac-
Results
teristics that could potentially predict brain involvement.
Of the 480 patients, only three were alive by December The incidence of BM is reported to be 1–2% in patients
2015. In total, 206 of 448 patients’ tumors (46.0%; 95% with primary CRC [2,11], and 3–5% in patients with meta-
CI ¼ 41.3–50.7%) were RAS mutated. Of these, KRAS exon 2 static disease [1,3,13,14,19]. In our study, we found an inci-
mutations were detected in 169 tumors (82.0%; 95% dence of 8.8% (95% CI ¼ 6.4–11.6%). The higher incidence
CI ¼ 76.1–87.0%), KRAS exon 3 or 4 mutations in 22 tumors was most likely a result of our cohort only comprising
(10.7%; 95% CI ¼ 6.8–15.7%) and NRAS mutations in 15 patients who lived long enough to be candidates for third-
tumors (7.3%; 95% CI ¼ 4.1–11.7%). BRAF V600E mutation was line treatment.
identified in 30 of 448 patients (6.7%; 95% CI ¼ 4.6–9.4%). In our study, lung metastases at mCRC diagnosis and rec-
The RAS and BRAF V600E mutations were mutually exclusive. tal cancer significantly increased the risk of developing BM
PIK3CA mutations were detected in 61 of 375 tumors (16.3%; on univariate cox regression. However, a significant associ-
95% CI ¼ 12.7–20.4%). PTEN, ERBB2 and EGFR mutations were ation between rectal cancer and lung metastases was also
only identified in three, one and three patients respectively, observed on chi-squared test (p < 0.0005), and on multivari-
and were therefore excluded from further analysis. ate cox regression only the effect of lung metastases
BM were diagnosed in 42 of the 480 patients analyzed remained significant (HR ¼ 3.096). Both rectal cancer and
(8.8%; 95% CI ¼ 6.4–11.6%). Median time to BM diagnosis lung metastases have previously been associated with BM
was 47.5 months (range ¼ 16–142 months) after primary [3,13,20]. We also observed that patients with absence of
diagnosis and 29.0 months (range: 5–142 months) after peritoneal metastases and local recurrence had significantly
mCRC diagnosis. No patients had BM at diagnosis of meta- increased incidence of BM. Furthermore, there was a non-sig-
static disease. Patients that developed BM had significantly nificant trend towards a lower incidence of BM among
longer survival from diagnosis of mCRC than patients patient with liver metastases than in patients without. Others
without BM (median ¼ 31.5 versus 28.0 months, p ¼ 0.001). have reported similar results for both peritoneal and liver
Median survival after diagnosis of BM was 2.9 months metastases [3,13,20].
(range ¼ 0–25 months). The vascular anatomy of the colon and rectum may
Patient characteristics at mCRC diagnosis are shown in explain this pattern of metastases as proposed by others
Table 2. Several variables were significantly associated with [15]. The colon drains mostly through the portal vein to the
BM on chi-squared and Fisher’s exact test. This included pri- liver and from here to the lung and thereafter the brain. The
mary tumor location (p ¼ 0.001, highest incidence of BM in rectum drains both through the portal and the cava vein.
patients with rectal cancer), and timing of metastatic disease This might explain why primary rectal tumors more often
(p ¼ 0.003, highest incidence of BM in patients with meta- seed the lung before the liver, and consequently the
chronous metastatic disease). Also, presence or absence of brain [15].
primary tumor/local recurrence (p ¼ 0.036), lung metastases A different explanation for the pattern of metastases and
(p < 0.0005) and peritoneal metastases (p ¼ 0.024) at mCRC who develops BM is genetic alterations in the tumor that
diagnosis were associated with BM, with the highest inci- might increase the ability of the tumor to seed certain
dence of BM in patients with presence of lung metastases, organs such as the brain or lung. Mutations in the RAS onco-
and absence of primary tumor/local recurrence and genes (NRAS and KRAS) have previously been associated with
 ACTA ONCOLOGICA 643

Table 2. Patient characteristics.

Characteristics Patients without BM, N ¼ 438 (91.2%) Patients with BM, N ¼ 42 (8.8%) p valuea
Sex
Male, N (%) 256 (90.5) 27 (9.5) 0.462
Female, N (%) 182 (92.4) 15 (7.6)
Age at CRC diagnosis
Median (range), years 60 (27–84) 58 (22–74) 0.090
Primary location
Right colon, N (%) 101 (97.1) 3 (2.9) 0.001
Left colon, N (%) 183 (93.4) 13 (6.6)
Rectum, N (%) 153 (85.5) 26 (14.5)
Unknown, N 1 0
Surgery for primary tumor
No, N (%) 64 (92.8) 5 (7.2) 0.633
Yes, N (%) 374 (91.0) 37 (9.0)
Time from primary diagnosis to diagnosis of metastatic disease
Median (range), months 0 (0–125) 12.5 (0–55) 0.001
Timing of metastatic disease
Synchronous, N (%) 248 (94.7) 14 (5.3) 0.003
Metachronous, N (%) 188 (87.0) 28 (13.0)
Unknown, N 2 0
Number of metastatic sites at mCRC diagnosis
One site, N (%) 181 (89.2) 22 (10.8) 0.333
Two sites, N (%) 143 (92.3) 12 (7.7)
Three or more sites, N (%) 93 (93.9) 6 (6.1)
Unknown, N 21 2
Presence of primary tumor or local metastases at mCRC diagnosis
No, N (%) 293 (89.3) 35 (10.7) 0.036
Yes, N (%) 140 (95.2) 7 (4.8)
Unknown, N 5 0
Node metastases at mCRC diagnosis
No, N (%) 302 (91.0) 30 (9.0) 0.820
Yes, N (%) 131 (91.6) 12 (8.4)
Unknown, N 5 0
Liver metastases at mCRC diagnosis
No, N (%) 125 (88.0) 17 (12.0) 0.117
Yes, N (%) 308 (92.5) 25 (7.5)
Unknown, N 5 0
Lung metastases at mCRC diagnosis
No, N (%) 303 (95.0) 16 (5.0) <0.0005
Yes, N (%) 130 (83.3) 26 (16.7)
Unknown, N 5 0
Peritoneal metastases at mCRC diagnosis
No, N (%) 387 (90.2) 42 (9.8) 0.024
Yes, N (%) 46 (100.0) 0 (0.0)
Unknown, N 5 0
KRAS exon 2
WT, N (%) 255 (91.4) 24 (8.6) 0.756
Mutated, N (%) 153 (90.5) 16 (9.5)
Unknown, N 30 2
KRAS exon 2, 3 or 4
WT, N (%) 236 (91.7) 21 (8.2) 0.514
Mutated, N (%) 172 (90.3) 19 (9.9)
Unknown, N 30 2
All RASb
WT, N (%) 222 (91.7) 20 (8.3) 0.593
Mutated, N (%) 186 (90.3) 20 (9.7)
Unknown, N 30 2
BRAF V600E
WT, N (%) 378 (90.4) 40 (9.6) 0.095
Mutated, N (%) 30 (100.0) 0 (0.0)
Unknown, N 30 2
PIK3CA
WT, N (%) 288 (91.7) 26 (8.3) 0.420
Mutated, N (%) 54 (88.5) 7 (11.5)
Unknown, N 96 9
a
p value calculated using chi-squared test or Fisher’s exact test on categorical variables and Mann–Whitney U-test on continuous variables.
b
KRAS (exon 2, 3 and 4) or NRAS (exon 2, 3 and 4).
N: number of patients; %: percentage of patients; SD: standard deviation; Mut: mutation.

aggressive tumor biology [17,21,22] and an increased risk of that patients with mutations in RAS genes (KRAS and NRAS)
developing BM [12,13]. Tie et al. [12] showed a significantly had a significantly higher incidence of BM. Mutations in
higher frequency of KRAS mutations in brain and lung meta- PIK3CA and BRAF have also been suggested as risk factors for
stases compared to liver metastases and primary CRC tumors. brain involvement, but no significant association has been
In a large series of 918 patients, Yeager et al. [13] also found shown so far [12–14]. In our study, none of these mutations
644 T. D. CHRISTENSEN ET AL.

Table 3. Multivariate cox regression of factors affecting hazard for brain metastases.
Hazard ratio (95% CI) p value
Primary location
Right colon 0.360
Left colon 1.649 (0.463–5.865) 0.440
Rectum 2.293 (0.656–8.022) 0.194
Metachronous versus synchronous metastatic disease 1.689 (0.857–3.329) 0.130
Having lung metastases 3.534 (1.810–6.899) <0.0005
CI: confidence interval.

were statistically significant associated with BM, and muta- In conclusion, the incidence of BM was 8.8% (95%
tion status did not affect HR for BM development. The dis- CI ¼ 6.4–11.6%) in our cohort of patients with mCRC who
cordance between our study and the one by Yeager et al. received third-line therapy. Beside long survival, the most
[13] with regard to RAS mutations might be due to the inter- important risk factor for BM seems to be lung metastases.
vention with cetuximab therapy to all patients in our study Rectal cancer and metachronous metastatic disease are also
compared to their study where RAS status was used to guide linked to BM development. None of the investigated muta-
anti-EGFR [13]. However, it could also be a consequence of tions (KRAS, NRAS, BRAF, PIK3CA, PTEN, ERBB2 and EGFR) were
only including patients in our study who received third-line significantly associated with BM.
therapy, thereby excluding patients that potentially could
have developed BM at an early stage.
Patients who developed BM had a significantly longer time Disclosure statement
from diagnosis of primary CRC to mCRC, and patients with The authors report no conflicts of interest. The authors alone are respon-
metachronous disease had a significantly increased risk of BM sible for the content and writing of this article.
on univariate cox regression analysis. This observation is com-
parable with the finding by Sundermeyer et al. [20] who
observed that early stage cancer at primary diagnosis were Funding
associated with increased incidence of bone and BM in a This work was supported by the Danish Cancer Society under Grant Rl
cohort of 1020 patients with mCRC. Furthermore, survival 17-A7362-14-57. Mutation analyses were funded by Merck KGaA as part
from diagnosis of mCRC to death was significantly longer in of a different study.
None of the funders had any role in designing the study, neither in
BM patients than in non-BM patients. The difference in sur-
data collection, management, analysis or interpretation, nor in writing or
vival is most likely the consequence of BM developing late in submitting the article. Merck KGaA reviewed the article for medical
the disease, as observed both in our and other studies [2,11]. accuracy only before journal submission. The authors are fully respon-
Actually, long survival might be one of the most important sible for the content of this article, and the views and opinions described
risk factors for BM, and it possibly explains some of the associ- in the publication reflect solely those of the authors.
ations observed in our study. Patients who have synchronous
metastatic disease, presence of primary tumor at mCRC diag-
References
nosis, and/or peritoneal metastases tend to have shorter sur-
vival [23,24], which decreases the likelihood of developing BM. [1] Barnholtz-Sloan JS, Sloan AE, Davis FG, et al. Incidence propor-
On the contrary, patients with lung metastases only and meta- tions of brain metastases in patients diagnosed (1973 to 2001) in
the Metropolitan Detroit Cancer Surveillance System. J Clin Oncol.
chronous metastatic disease, probably have a better prognosis
2004;22:2865–2872.
[25], and thus increased likelihood of surviving long enough [2] Go PH, Klaassen Z, Meadows MC, et al. Gastrointestinal cancer
to develop BM. Despite the long survival until development of and brain metastasis: a rare and ominous sign. Cancer.
BM, the survival after diagnosed of BM was only median 2011;117:3630–3640.
2.9 months, which is comparable to previous studies [4–7]. [3] Christensen TD, Spindler KL, Palshof JA, et al. Systematic review:
BM from CRC are rare, and systematic screening of brain metastases from colorectal cancer-Incidence and patient
characteristics. BMC Cancer. 2016;16:260.
patients for BM are not advisable [3]. However, based on the [4] Noura S, Ohue M, Shingai T, et al. Brain metastasis from colorectal
results of our study, it is clear that clinicians should be aware cancer: prognostic factors and survival. J Surg Oncol. 2012;
of symptoms of BM in long surviving mCRC patients, espe- 106:144–148.
cially, if the patients have lung metastases, and to a lesser [5] Kruser TJ, Chao ST, Elson P, et al. Multidisciplinary management
extend rectal cancer or metachronous metastatic disease. of colorectal brain metastases: a retrospective study. Cancer.
2008;113:158–165.
The study is limited by its retrospective nature, and by
[6] Smedby KE, Brandt L, Backlund ML, et al. Brain metastases admis-
only including patients who received third-line therapy. Even sions in Sweden between 1987 and 2006. Br J Cancer. 2009;
though our study is one of the largest studies to examine 101:1919–1924.
the predictive potential of characteristics and mutations on [7] Michl M, Thurmaier J, Schubert-Fritschle G, et al. Brain metastasis
later BM development in CRC patients, the sample size might in colorectal cancer patients: survival and analysis of prognostic
have been too small to yield significant results. Given the factors. Clin Colorectal Cancer. 2015;14:281–290.
[8] Wronski M, Arbit E. Resection of brain metastases from colorectal
limitations, we recommend that others validate our results in
carcinoma in 73 patients. Cancer. 1999;85:1677–1685.
new, larger and preferably prospective studies including [9] Fowler A, Cook R, Biggs M, et al. Survival of patients following
unselected mCRC patients. neurosurgical treatment of colorectal adenocarcinoma metastasis
 ACTA ONCOLOGICA 645

in the Northern Sydney-Central Coast area. J Clin Neurosci. poor prognosis but do not preclude benefit from oxaliplatin or
2008;15:998–1004. irinotecan: results from the MRC FOCUS trial. J Clin Oncol.
[10] Damiens K, Ayoub JP, Lemieux B, et al. Clinical features and 2009;27:5931–5937.
course of brain metastases in colorectal cancer: an experience [19] Hugen N, van de Velde CJ, de Wilt JH, et al. Metastatic pattern in
from a single institution. Curr Oncol. 2012;19:254–258. colorectal cancer is strongly influenced by histological subtype.
[11] Nayak L, Lee EQ, Wen PY. Epidemiology of brain metastases. Curr Ann Oncol. 2014;25:651–657.
Oncol Rep. 2012;14:48–54. [20] Sundermeyer ML, Meropol NJ, Rogatko A, et al. Changing pat-
[12] Tie J, Lipton L, Desai J, et al. KRAS mutation is associated with terns of bone and brain metastases in patients with colorectal
lung metastasis in patients with curatively resected colorectal cancer. Clin Colorectal Cancer. 2005;5:108–113.
cancer. Clin Cancer Res. 2011;17:1122–1130. [21] Hecht JR, Douillard JY, Schwartzberg L, et al. Extended RAS ana-
[13] Yaeger R, Cowell E, Chou JF, et al. RAS mutations affect pattern lysis for anti-epidermal growth factor therapy in patients
of metastatic spread and increase propensity for brain metastasis with metastatic colorectal cancer. Cancer Treat Rev. 2015;
in colorectal cancer. Cancer. 2015;121:1195–1203. 41:653–659.
[14] Tran B, Kopetz S, Tie J, et al. Impact of BRAF mutation and micro- [22] Vauthey JN, Zimmitti G, Kopetz SE, et al. RAS mutation status
satellite instability on the pattern of metastatic spread and prog- predicts survival and patterns of recurrence in patients
nosis in metastatic colorectal cancer. Cancer. 2011;117:4623–4632. undergoing hepatectomy for colorectal liver metastases. Ann
[15] Magni E, Santoro L, Ravenda PS, et al. Brain metastases from colo- Surg. 2013;258:619–626, discussion 26–7.
rectal cancer: main clinical factors conditioning outcome. Int J [23] Franko J, Shi Q, Goldman CD, et al. Treatment of colorectal peri-
Colorectal Dis. 2014;29:201–208. toneal carcinomatosis with systemic chemotherapy: a pooled ana-
[16] Tanriverdi O, Kaytan-Saglam E, Ulger S, et al. The clinical and lysis of north central cancer treatment group phase III trials
pathological features of 133 colorectal cancer patients with brain N9741 and N9841. J Clin Oncol. 2012;30:263–267.
metastasis: a multicenter retrospective analysis of the [24] Safaee Ardekani G, Jafarnejad SM, Tan L, et al. The prog-
Gastrointestinal Tumors Working Committee of the Turkish nostic value of BRAF mutation in colorectal cancer and mel-
Oncology Group (TOG). Med Oncol. 2014;31:152. anoma: a systematic review and meta-analysis. PloS One.
[17] Kemeny NE, Chou JF, Capanu M, et al. KRAS mutation influences 2012;7:e47054.
recurrence patterns in patients undergoing hepatic resection of [25] Hwang M, Jayakrishnan TT, Green DE, et al. Systematic review of
colorectal metastases. Cancer. 2014;120:3965–3971. outcomes of patients undergoing resection for colorectal liver
[18] Richman SD, Seymour MT, Chambers P, et al. KRAS and BRAF metastases in the setting of extra hepatic disease. Eur J Cancer.
mutations in advanced colorectal cancer are associated with 2014;50:1747–1757.