Distribution of Brain Metastases
Jean Yves Delattre, MD; George Krol, MD;
Howard T. Thaler, MD; Jerome B. Posner, MD
\s=b\ The number and site of brain metas-
tases were identified on the computed
tomographic scans of 288 patients. There
was one brain metastasis in 49%, two in
21%, three in 13%, four in 6%, and five or
more in 11% of scans. In patients with one
metastasis, the posterior fossa was
involved in 50% of patients when the
primary tumor was pelvic (prostate or
uterus) or gastrointestinal, but it was
involved in only 10% of patients with
other primary tumors. Hemispheral metas-
tases preferred the anatomic "watershed
areas" (29% of the brain surface con-
tained 37% of the metastases), indicating
that tumoral microemboli tend to lodge in
the capillaries of the distal parts of the
superficial arteries. The charts of 134
patients with brain metastases from a
primary tumor originating outside the
lung revealed that the incidence of lung
and spine metastases was the same,
whether the primary tumor was pelvic or
gastrointestinal or from another site.
These data suggest that the high inci-
dence of subtentorial lesions in patients
with pelvic and gastrointestinal primary
tumors cannot be explained by arterial
embolization alone, and that this peculiar
distribution is probably not explained by
seeding of the brain through Batson's
plexus.
(Arch Neurol 1988;45:741-744)
rPhe distribution of métastases in the
brain and the relationship between
the distribution and the primary
Accepted for publication Feb 29, 1988.
From the Departments of Neurology (Drs
Delattre and Posner), Radiology (Dr Krol), and
Biostatistics (Dr Thaler), Memorial Sloan-Ket-
tering Cancer Center and Cornell University
Medical College, New York.
Reprint requests to Department of Neurology,
Memorial Sloan-Kettering Cancer Center, 1275
York Ave, New York, NY 10021 (Dr Posner).
Downloaded From: http://archneur.jamanetwork.com/ by a DALHOUSIE UNIVERSITY-DAL-11762 User on 06/20/2015
tumor have long been controversial six main sections of a head CT scan and the
subjects. Some authors have found a arterial territories of the brain in similar
preferential involvement of the retro- planes8 (Pig 1). Although such mapping
rolandic13 or subtentorial areas,46 and may result in minor inaccuracies, it
others have found a random distribu¬ seemed the only method allowing a com¬
parative study in such a large number of
tion, proportional only to the weight patients.
or blood supply of the various regions The location of the métastases was
of the brain.7 Previous studies were all studied by comparing the CT scans with
performed post mortem, but since all anatomic sections in similar planes.9·10
but the smallest métastases may be However, because CT resolution is inferior
identified by computed tomography to anatomic sections, it was usually impos¬
(CT), we studied the number and loca¬ sible to distinguish precentrai and postcen¬
tion of brain métastases by CT scan, tral gyri on CT scans; these areas were
in the hope that the information denominated "rolandic." Areas corre¬
sponding to the distal fields of the main
might lead to a better understanding
of the pathophysiology of brain seed¬
Fig 1.—Standard sheet onto which lesions
ing. noted on computed tomographic scans were
mapped and schematic representation of dis¬
PATIENTS AND METHODS
tal arterial fields in similar planes. Only three
CT Study
of six sections (D, E, and F) are presented
Between January 1979 and October 1985, (from Toole8).
656 patients with neoplasms originating
outside the nervous system were re¬
corded in the computer files of Memorial
Sloan-Kettering Cancer Center (New
York) as having brain métastases. Of
these, 302 patients had a CT scan available
for study. The CT scan was performed
using either of two scanners (General Elec¬
tric 9800 or Ohio Nuclear 2020). Scan
sequences consisted of 12 axial sections
extending from the base to the vertex,
repeated after intravenous administration
of 60% diatrizoate meglumine injection
(100 mL). In 288 patients, the CT scan was
consistent with brain métastases (ie, cir¬
cumscribed parenchymatous lesion or
lesions, isodense, hyperdense, or hypo¬
dense, enhancing after contrast infusion in
a homogeneous, ring-enhancing or, more
rarely, heterogeneous fashion with mild-
to-severe surrounding edema and mass
effect on the adjacent structures and ven¬
tricular system). There were 14 patients
who were excluded from the study (durai
métastases, CT scans suggestive of lepto-
meningeal disease, and part of the CT scan
not present in the file). For each patient,
lesions noted on CT scans were mapped
onto a standard sheet containing both the
 cerebral arteries were also defined on the
standard sheet (Fig 1). The size of these
areas was defined after reviewing the dis¬
tribution of distal-field infarcts1113 and
took into account the minor variations that
exist in the medical literature reporting
arterial distribution on CT scans.811 To be
included, the center and more than 75% of
the lesion had to be situated within these
areas. The different surfaces were mea¬
sured after copying the standard sheet on
millimeter paper. The number, location,
and vascular distribution of the métasta¬
ses were related to the site of the primary
tumor. Statistical comparisons between
the observed proportion of lesions in the
distal fields and the proportion predicted
by surface area were based on the binomial
distribution. Comparisons between propor¬
tions in two groups (eg, location of lesions
as a function of primary tumor) were made
using the 2 or Fisher's exact test for 2 X 2
contingency tables.
Chart Study of Patients With a Primary
Tumor Situated Outside the Lung
When we discovered from the above-
mentioned CT study that the distribution
of brain métastases differed, depending in
part on the location of the primary tumor,
we undertook to review the charts of
Table 1.—Primary Cancer Site in
288 Patients With Brain Métastases
Site _No. (%)
Lung 144 (50)
Breast 43 (15)
Melanoma 30 (10.5)
Pelvis-abdomen 27 (9.5)
Unknown 32(11)
Others 12 (4)
Total 288 (100)
patients with a primary tumor situated
outside the lung. Our hypothesis was that
patients whose tumors reached the brain
via Batson's plexus would have a higher
incidence of vertebral and/or skull metas-
tases14 than of lung métastases. By con¬
trast, arterial seeding should lead to a
higher incidence of lung métastases, since
the metastatic cells must traverse the
lungs to reach the arterial circulation.
Between January 1979 and December 1984,
274 patients suffered a brain metastasis
from a primary tumor arising outside the
lung. The medical records and CT reports,
available for 134 of these patients, were
reviewed in order to correlate the number
(single or multiple) and location (infraten¬
torial vs supratentorial) of brain métasta¬
ses with the simultaneous presence of lung
and/or skull or vertebral métastases. The
diagnosis of lung or bone metastasis was
made roentgenographically or by isotope
scan and was sometimes confirmed by
autopsy or surgery. All patients had one or
several chest roentgenograms at the time
of the diagnosis of brain métastases that
could be compared with previous and fol¬
lowing roentgenograms. There were 97
patients (of 134) who had bone studies
(isotope scan and/or roentgenograms of
the spine and/or myelogram) in the six
months preceding or following the develop¬
ment of brain métastases. There were 37
patients (of 134) without clinical com¬
plaints who had no bone investigations.
CT and Chart Studies
The CT and chart studies did not neces¬
sarily include the same patients. Some
patients had CT scans but had no charts
available, while other patients had charts
including CT reports, but the CT scan itself
could not be reviewed. When different
patients were involved, wé added the
Fig 2.—Distribution of métastases by computed tomographic
in 256 patients who had fewer than five métastases.
results of these two studies in order to
correlate the location of single métastases
(supratentorial or infratentorial) with the
primary tumor site. A total of 169 patients
with a single metastasis were studied.
RESULTS
CT Study
The primary cancers and number of
métastases are indicated in Tables 1
and 2. Melanoma, lung cancer, and
cancers of unknown primary site more
often caused multiple lesions, whereas
patients with breast cancer and, par¬
ticularly, patients with pelvic and
abdominal primary tumors more
often had single lesions (Table 3). The
location of the 443 métastases in the
256 patients with fewer than five
métastases is indicated in Fig 2. Both
the left side and the right side of the
brain were equally represented (right,
48%; left, 46%, and center, 6%). The
frontal, parietal, and temporoparie-
to-occipital regions were more often
involved than the temporal and occip¬
ital lobes. Study of the vascular distri¬
bution revealed that the distal fields
(anatomic watershed areas) of the
main cerebral arteries were signifi¬
cantly overrepresented. These areas
contained 37% of the métastases,
although they were estimated to rep¬
resent only 29% of the surface areas
(P < .01).
Among patients with a single
metastasis from a primary tumor sit¬
uated in the pelvis (prostate or uter¬
us) and abdomen (gastrointestinal
scan. There were 443 métastases

Table 2.—Number of Métastases in

288 Patients With Brain Métastases

No. of No. of
Métastases Patients (%)
1 141 (49)
2 60(21)
3 38 (13)
4 17(6)
5+ 32(11)
Total 288 (100)

Table 3.—Relationship Between

Primary Cancer Site and
Number of Métastases in
288 Patients With Brain Métastases

Single, Multiple,
Site % %
Lung 46 54
Melanoma 41 59
Unknown 32 68
Breast 56 44
Pelvis-abdomen 69 31

Downloaded From: http://archneur.jamanetwork.com/ by a DALHOUSIE UNIVERSITY-DAL-11762 User on 06/20/2015

 tract), the incidence of infratentorial
métastases was 50% (5/10), compared
with 10% (13/131) in patients with
other primary tumors (P .003).
=
Chart Study of Patients With a Primary
Cancer Situated Outside the Lung
Of these patients, 66% (88/134) had
lung lesions consistent with métasta¬
ses. The values were not significantly
different for patients with single or
multiple métastases (65% [42/65] and
67% [46/69], respectively), for pa¬
tients with a single infratentorial
lesion compared with a single supra¬
tentorial lesion (42% [5/12] and 70%
[37/53], respectively), or for patients
with a single metastasis from a pelvic
or gastrointestinal primary tumor
(70% [7/10]).
Spine or skull lesions were identi¬
fied in 27% (36/134). The values were
not significantly different for patients
with single or multiple métastases
(25% [16/65] and 29% [20/69], respec¬
tively), for patients with a solitary
infratentorial or supratentorial lesion
(25% [3/12] and 25% [13/53], respec¬
tively), or for patients with a gastro¬
intestinal, prostate, or uterine prima¬
ry tumor (20% [2/10]).
CT and Chart Studies
The significant overrepresentation
of the infratentorial areas in patients
with a single metastasis from a pelvic
(prostate or uterus) or gastrointesti¬
nal primary tumor, compared with
the patients with a single metastasis
from other primary tumors, was also
found when we added the results of
the CT and chart studies. Overall,
53% (8/15) of patients with a single
metastasis from pelvic and gastroin¬
testinal primary tumors had infraten-
Fig 3.—Distribution of single metastasis (computed tomographic and chart study). Posterior fossa is
significantly overrepresented in pelvic (prostate or uterus) and gastrointestinal primary tumors (left)
compared with patients with other primary tumors (P < .001).
Downloaded From: http://archneur.jamanetwork.com/ by a DALHOUSIE UNIVERSITY-DAL-11762 User on 06/20/2015
torial lesions (cerebellum, 40% [6/15]
and brain stem, 13% [2/15]), com¬
pared with 10% (15/154) of patients
with a single metastasis from another
primary tumor (cerebellum, 8% [13/
154] and brain stem, 1% [2/154]) (Fig
3) (P < .001). Such a predominance of
posterior fossa lesions was not
observed in patients with renal cancer
(7% [1/15] only had a solitary infra¬
tentorial lesion).
COMMENT
These data yield three findings. The
first is that 70% of patients have only
one or two brain métastases. The sec¬
ond is that there is a preferential
distribution in the superficial distal
arterial fields, ie, the anatomic water¬
shed areas. The third is that in a small
group of patients with abdominal and
pelvic primary tumors, the distribu¬
tion of brain métastases is different,
with a predominant involvement of
the posterior fossa.
Several méthodologie issues require
discussion.
1. We could review the CT scans
and/or the charts of only 370 of 656
patients, allowing for possible bias in
the sample. However, the frequency of
the various primary tumors that we
observed is in agreement with previ¬
ous pathologic studies from this insti¬
tution, suggesting no bias occurred
during the collection of the data.
2. We did not attempt to review the
charts of the patients with lung can¬
cer, since it is generally accepted that
lung cancer spreads to the brain via
arterial microemboli.
3. The incidence of lung, spine, and
skull métastases is probably underes¬
timated in view of the poor sensitivity
of standard techniques. However, the
issue here was to detect a difference
between subgroups harboring various
primary tumors, and we believe that,
had a significant difference existed
between these subgroups, we would
have detected it.
Given that our data truly reflect the
situation in brain métastases, what
conclusions can be drawn? The first is
that in life, as well as at autopsy,
distribution between single and mul¬
tiple brain lesions is approximately
equal.15 Furthermore, the similar val¬
ue obtained by CT scan (single, 49%,
and multiple, 51%) compared with
that obtained at autopsy demonstrate
the diagnostic sensitivity of CT scans
in detecting brain métastases.
Among patients with more than one
lesion, the therapeutic approach may
differ when a patient has two superfi¬
cial lesions, as compared with several
lesions. In this study, 70% of patients
had one or two lesions, and only 11%
of patients had more than four
lesions. Thus, diffuse metastatic inva¬
sion of the brain is not as frequent as
generally believed, even in patients
with melanoma or lung cancer who
are more prone to have multiple brain
métastases develop.1·16 Considering
that the number of long-term survi¬
vors (one year or more) after surgical
resection of a single lesion is not
negligible,17 and that local recurrence
is more frequent than the develop¬
ment of a new lesion,18 our results
support the trend to develop focal
treatment for brain métastases in
order to reduce the incidence of
delayed complications of whole brain
radiation therapy and polychemother-
apy."
Brain métastases were more fre¬
quent in frontal and parietal lobes
 than in occipital and temporal lobes,
probably related to the respective
mass of these structures. We also
found an overrepresentation of the
temporo-occipital and parieto-occipi¬
tal areas, as previously reported,1·2
which cannot be explained by the
mass of these structures. Our results
suggest that this distribution is due to
a preferential location of métastases
in the posterior border zone (between
the middle and posterior cerebral
arteries). We also found an overrepre¬
sentation of the métastases in the
anterior border zone (between the
anterior and middle cerebral arte¬
ries). This finding strongly supports
the view that most brain métastases
result from arterial tumoral micro¬
emboli, in agreement with the rule
that emboli tend to pass along the
arterial tree20 as far distally as their
size permits.
In a subgroup of patients with a
solitary metastasis from a primary
tumor situated in the pelvis or gastro¬
intestinal tract, we found a significant
overrepresentation of infratentorial
métastases compared with those from
other primary tumors. This observa¬
tion cannot be explained by arterial
embolization alone and suggested a
possible role for Batson's venous plex¬
us. During transient episodes of
1. Paillas JE, Pellet W: Brain metastases, in
Vinken PJ, Bruyn GW (eds): Handbook of Clini-
cal Neurology. New York, Elsevier Science Pub-
lishing Co Inc, 1975, vol 18, pp 201-232.
2. Kindt GW: The pattern of location of cere-
bral metastatic tumors. J Neurosurg 1964;21:54\x=req-\
57.
3. Weisberg L, Nice C, Katz M: Cerebral Com-
puted Tomography: A Text Atlas. Philadelphia,
WB Saunders Co., 1984.
4. Chason JL, Walker FB, Landers JM: Meta-
static carcinoma in the central nervous system
and dorsal root ganglia: A prospective autopsy
study. Cancer 1963;16:781-787.
5. Meyer PC, Reah TG: Secondary neoplasms
of the central nervous system and meninges. Br J
Cancer 1953;7:438-448.
6. Lesse S, Netsky MG: Metastases of neo-
plasms to the central nervous system and menin-
ges. Arch Neurol Psychiatry 1954;72:133-152.
7. Ask-Upmark E: Metastatic tumours of
brain and their localisation. Acta Med Scand
1956;154:1-9.
8. Toole JF: Cerebrovascular Disorders, ed 3.
New York, Raven Press, 1984.
9. Schnitzlein HN, Wright Hartley E, Murtagh
FR, et al: Computed Tomography of the Head and
Spine: A Photographic Color Atlas of CT, Gross,
and Microscopic Anatomy. Baltimore, Urban &
Schwarzenberg, 1983.
Downloaded From: http://archneur.jamanetwork.com/ by a DALHOUSIE UNIVERSITY-DAL-11762 User on 06/20/2015
increased abdominal pressure with
compression of the vena cava, tumor
could seed the spine directly, without
transiting through the lungs.21 The
cerebral durai sinuses are a direct
extension of the spinal epidural plex¬
us. The pathway is through the basi¬
lar plexus of veins to, in part, the
inferior petrous sinuses, which are an
important outlet for the cerebellum
and brain stem.22·23 Métastases via this
retrograde pathway could provide an
explanation for the preferential
involvement of the posterior fossa in
patients with abdominal and pelvic
primary tumors. Our results do not
support this hypothesis and provide
evidence against a role for Batson's
plexus: one should expect an increased
incidence of spine or skull lesions in
patients with abdominal or pelvic pri¬
mary tumors compared with other
primary tumors if seeding occurred
through Batson's plexus, but no such
difference was observed. Further¬
more, patients with isolated infraten¬
torial lesions, whether the primary
tumor was pelvic and abdominal or
from another site, did not differ from
patients with supratentorial lesions in
terms of lung and spine métastases.
Why, then, do some pelvic and
abdominal tumors prefer the posteri¬
or fossa? The "fertile-soil" hypothesis
References
10. Damasio H: A computed tomographic
guide to the identification of cerebral vascular
territories. Arch Neurol 1983;40:138-142.
11. Barnett HJM, Stein BM, Mohr JP, et al:
Stroke: Pathophysiology, Diagnosis, and Manage-
ment. New York, Churchill Livingstone Inc,
1986.
12. Wodarz R: Watershed infarctions and com-
puted tomography: A topographical study in
cases with stenosis or occlusion of the carotid
artery. Neuroradiology 1980;19:245-248.
13. Bogousslavsky J, Regli F: Unilateral
watershed cerebral infarcts. Neurology 1986;
36:373-377.
14. Batson OV: The role of the vertebral veins
in metastatic processes. Ann Intern Med 1942;
16:38-45.
15. Weiss L, Gilbert HA, Posner JB: Brain
Metastasis. Boston, GK Hall & Co, 1980, pp
2-29.
16. Takakura K, Sano K, Hojo S, et al: Meta-
static Tumors of the Central Nervous System.
New York, Igaku-Shoin Medical Publishers Inc,
1982.
17. White KT, Flemming JR, Laws ER, et al:
Single metastasis to the brain: Surgical treat-
ment in 122 consecutive patients. Mayo Clin Proc
1981;56:424-428.
18. Patchell RA, Cirrincione C, Thaler HT, et
al: Single brain metastases: Surgery plus radia-
is based on clinical and experimental
evidence demonstrating that circulat¬
ing tumoral cells arrest in a wide
variety of organs, but that métastases
occur only in certain tissues.24 Fur¬
thermore, it has been shown experi¬
mentally that subgroups of métasta¬
ses could preferentially colonize dif¬
ferent areas of the meninges and
brain, suggesting different affinities
of these subgroups of tumor cells for
various parts of the brain and menin¬
ges.25 Additional evidence supporting
the hypothesis is the fact that
although prostate and uterine cancer
preferentially metastasize to the cere¬
bellum, another retroperitoneal ab¬
dominal tumor (renal carcinoma) does
not. Gastrointestinal tumors (intra-
peritoneal) metastasize preferentially
to the cerebellum. If the problem were
one of vascular access only, one would
expect renal tumors to preferentially
metastasize to the cerebellum as
well.
Although the mechanism of the
preferential involvement of the poste¬
rior fossa in this subgroup of patients
is unknown, our results suggest that
particular attention should be direct¬
ed at the posterior fossa in patients
with pelvic and abdominal primary
tumors who are suspected of having
brain métastases.
tion or radiation alone. Neurology 1986;36:447\x=req-\
453.
19. Chak LY, Zatz LM, Wasserstein P, et al:
Neurologic dysfunction in patients treated for
small cell carcinoma of the lung: A clinical and
radiological study. Int J Radiat Oncol Biol Phys
1986;12:385-389.
20. Torvik A: The pathogenesis of watershed
infarcts in the brain. Stroke 1984;15:221-223.
21. Coman DR, Delong RP: The role of the
vertebral venous system in the metastasis of
cancer to the spinal column: Experiments with
tumor cell suspension in rats and rabbits. Cancer
1951;4:610-618.
22. Carpenter MB, Sutin J: Human Neuroana-
tomy, ed 8. Baltimore, Williams & Wilkins, 1983;
p 624.
23. Capra NF, Anderson KV: Anatomy of the
central venous system, in Kapp JP, Schmidek H
(eds): The Central Venous System and Its Disor-
ders. New York, Grune & Stratton, 1984, pp
1-36.
24. Cairncross JG, Posner JB: The manage-
ment of brain metastases, in Walker MD (ed):
Oncology of the Nervous System. Hingham, Mass,
Martinus Nijhoff Publishers, 1983, pp 341-378.
25. Brunson KW, Beattie G, Nicolson GL:
Selections and altered properties of brain colo-
nising metastatic melanoma. Nature 1978;
272:543-544.