1551

Predilection of Brain Metastasis in Gray and White Matter Junction and Vascular Border Zones
Te-Long Hwang, M.D.lS3 Timothy P. Close, M.D? John M. Grego, Ph.D! William L. Brannon, M . D . ~ ~ Francisco Gonzales, M . D . ~ , ~
BACKGROUND. The purpose of this study was to assess the importance of the vascular border zone and the gray and white matter junction in the distribution of brain metastases. METHODS. We reviewed the medical records, computed tomography (CT) or magnetic resonance imaging (MRI) of 105 patients with secondary brain tumors. The metastatic lesions noted on CT scans or MRI were matched with a predetermined standard sheet containing axial images with shading on the border zones. To be included in the border zones, the center or more than 50% of the lesion had to be situated within these zones. RESULTS. Among 100 evaluable patients, there were 302 metastatic brain lesions. Of the 302 lesions, 210 lesions were 2 cm or smaller in greatest dimension and located in cerebral and cerebellar hemispheres. The major vascular border zones were the site of predilection for 103 lesions (62%) although the border zones constitute only 29% of the area. Gray and white matter junction was the preferred site for 135 lesions (64%). CONCLUSIONS. The results demonstrated that brain metastasis occurs in the vascular border zone regions and the gray and white matter junction more frequently than previously recognized, and also supported the notion that metastatic emboli tend to lodge in an area of sudden reduction of vascular caliber (gray/white matter junction) and in the area most distal vascular field (border zone). Cancer 1996; 721551-5. 0 1996 American Cancer Society. KEYWORDS brain, metastasis, border zone, gray and white matter junction.

Department of Neuropsychiatry, University of

Saluth Carolina School of Medicine, Columbia, South Carolina.

* Department of Internal Medicine, University of

Scluth Carolina School of Medicine, Columbia, South Carolina. Department of Neurology, Richland Memorial Hospital, Columbia, South Carolina. Department of Radiology, Richland Memorial Hospital, Columbia, South Carolina. Oepartment of Internal Medicine, Richland Memorial Hospital, Columbia, South Carolina. Department of Statistics, University of South Carolina, Columbia, South Carolina.

Supported by a research award from the South Carolina Cancer Research and Treatment Center.

The authors thank Karen Andrews for providing thle cancer registry, Carol G. Crain for typing the manuscript, Susan E. Hilfer for drawing the cerebrovascular border zones, and Wanette Janes for preparing statistical graphs.
Address for reprints: Te-Long Hwang, M.D., Department of Neuropsychiatry, USC School of Medicine, Suite 1048, 3555 Harden St. Ext., Columbia, SC 29203. Received September 5, 1995; revision received December 6,1995; accepted December 6,1995. 0 1996 American Cancer Society

t has been considered that brain metastasis occurs in random distribution (trapping and arrest) proportional to the blood flow to specific brain areas. Approximately 80% of brain metastases are found in the cerebral hemispheres, 17% in the cerebellum, and 3% in the brain stem. Nonrandom patterns of brain metastases have also been observed. Both pelvic (prostate and uterus) and gastrointestinal carcinoma are disproportionally represented in the ~erebellum.~ Among specific regions of the cerebral hemisphere, frontal (21%), parietal (19%), and temporoparietal-occipital (19%) are more often involved in metastases than temporal and occipital 10bes.~ predilection for certain vascular areas of the A brain has also been described. Traditional teaching has emphasized the gray and white matter junction as a frequent location for metastatic tum o r ~but its frequency has not been reported in the English literature. ,~ On the contrary, some have suggested a predilection for the gray as opposed to the white matter because of its superior capillary d e n ~ i t y A~ . previous report showed that the watershed areas were overrepresented as metastatic sites (37%).3 This study reassesses the importance of the gray and white matter junction and the vascular border zone for brain metastasis.

1552

CANCER April 15,1996 / Volume 77 / Number 8

MATERIALS AND METHODS

Between January 1988 and December 1993, 164 patients with brain metastases were enrolled at the Cancer Registry of Richland Memorial Hospital, a 600-bed university teaching hospital in Columbia, South Carolina. Medical records, computerized tomograms (CT scans), or magnetic resonance images (MRI) of the brain were available for review for 124 cases. Nineteen cases were excluded because: (1) only extraparenchymal metastasis were present (leptomeningeal carcinomatosis and intracranial epidural); (2) the brain images failed to demonstrate a brain lesion; (3) only post craniotomy images were available; and (4) the tumor type was not clearly defined. Brain images (CT or MRI) were reviewed without knowledge of the patients clinical information. CT scans of the brain were performed with contiguous 8 mm thick axial images on a Siemens Somaton Plus, Somaton or DR3, 7 mm thick axial images on a General Electric High X Speed Advantage CT scanner or a Philips L CT scanner with 5 mm thick axial images in the posterior fossa and 10 mm thick axial images in the supratentorial regions. MRI images were obtained with a Phillips Gyroscan S-15 1.5 tesla magnet. CT and MRI studies performed at other institutions were also reviewed. Lesions were considered to be metastatic if they represented distinct areas of abnormally increased or decreased CT attenuation, generally with contrast enhancement, with or without vasogenic edema, and with mass effect except for the lesions less than 1 cm in greatest dimension. Lesions were considered metastatic if they demonstrated an abnormal signal on MRI, usually with gadolinium (Gd) enhancement, vasogenic edema, and mass effect. When possible, serial examinations were obtained to assess change. Routine brain biopsy was not performed. Lesions noted on CT scans or MRI were matched with a predetermined standard sheet containing axial images with shaded areas for the border zones (Fig. 1).The border zone map was developed from previous reports of the major cerebral vascular territories.- Border zones were arbitrarily defined within a 1 cm width corresponding with the areas of border zone infarctions.-6 Subcortical and cerebellar border zones were determined to be 0.5 cm in greatest dimension. To be included, the center or more than 50% of the lesion had to be situated within these zones. The junction between gray and white matter can usually be determined since they both show different tissue attenuation on CT scan and different intensities on MRI. MRI is particularly good in delineating the gray and white matter junction. If the junction was not clearly shown or if the anatomical landmarks were distorted by the tumor or edema, the metastatic lesions were classified as uncertain in location. Metastatic lesions larger than 2 cm in greatest dimension were excluded from the analysis since the border zone

SUBCORTICAL BORDER ZONE POSTERIOR BORDER ZONE CEREBELLAR BORDER ZONE

FIGURE 1. Major cerebrovascular border zones.

and gray and white matter junction could not be determined with certainty. The four major vascular border zones were defined as follows (Fig. 1).

Anterior Border Zone

The anterior border zone is situated between the supply territories of the anterior cerebral artery (ACA) and the middle cerebral artery (MCA). It includes a portion of the centrum semiovale and corona radiata.

Posterior Border Zone

The posterior border zone is situated between the supply territories of the ACA and the posterior cerebral artery (PCA), or the zone between the MCA and the PCA. This zone lies between the parietal lobe and the occipital lobe, including the triple vessel border zone of the ACA, MCA, and the PCA, or between the medial temporal lobe and the rest of the temporal lobe.

Subcortical Border Zone (Internal Border Zone)

This area is located in the deep white matter between the following: (1) territory of deep perforators, including the Heubners artery, and anterior striate branch of the ACA, the lenticulostriate arteries of MCA, and the anterior choroidal artery of the internal carotid artery; (2) vascular territory of superficial perforators, e g , white matter medullary branches of the MCA. Anatomically, the structures involved in the internal

Preferred Sites of Brain MetastaWHwang et al.

1553

border zone include the white matter of the corona radiata and the extreme lateral portions of the basal ganglia.I6 This area is more of a junctional zone between two terminal arterial networks than a watershed area.",I3

Cerebellar Border Zone

The cerebellar border zone is the area of blood supply overlapping the posterior inferior cerebellar artery (PICA), the superior cerebellar artery, and between the PICA and the anterior inferior cerebellar artery (AICA). Vasculature in the brain stem, basal ganglia, thalamus, pituitary, and pineal glands were considered too cornplicated to assign a border zone to by neuroimages. Border zone localization includes minor variations that exist in the medical literature reporting arterial territories. It also includes minor variations on axial planes between CT scan and MRI images. The data set was further analyzed with loglinear models for overdispersed count data. The count (or frequency) of metastatic tumor in each region (graylwhite, gray, white, and uncertain) and border zone (border zone, nortborder zone) was recorded for each patient. Therefore, 8 (4 x 2) observations were recorded for each patien t. Since many patients had only one lesion, numerous counts of zero were noted. The natural log of the mean number of lesions was modeled as a linear function of age, race, sex, tumor origin, tumor type, border zone, and gray and white matter region. Tumor origin and type were collapsed into a discrete variable with five levels (squamow cell lung cancer, non small cell lung cancer, adenocarcinoma cell lung cancer, small cell lung cancer, and non lung cancer); each with a reasonable number of cases, though the number of small cell lung cancers was small.

FIGURE 3. Metastatic foci at gray and white matter junction (210 lesions).

RESULTS
Among 71 male and 34 female patients with brain metastase.s, 70 were white (67%) and 35 were black (33%).Patients' ages ranged from 21 to 85 years (median: 62). Peak age ,was between 60 and 69 years. There was no pediatric case of brain metastasis in this series. Both CT scan and MRI were available for 15 patients, CT scan alone was available for 55 patients and MRI alone for 35. Lung carcinoma was the predominant primary cancer accounting for 75% of the case breast carcinoma accounted for 5.7%, and unknown primary carcinoma accounted for 9.5%. Single metastasis was observed in 56 cases (53%) and multiple metastasis was observed in 49 cases (47%). Five patients with metastatic tumors solely in the thalamus, basal ganglia, brainstem, and pituitary and pineal glands were excluded from evaluation of border zone and gray and white matter distribution. Among the 100 remaining patients, there were 302 metastatic brain lesions. Eighty of the lesions were larger than 2 cm in great-

est dimension and another 12 lesions were located in the brainstem, basal ganglia, thalamus, pituitary and pineal glands. These lesions were excluded from the assessment. Among 210 remaining metastatic brain lesions, the border zones were the site of predilection for 130 lesions (62%), although they were estimated to represent only 29% of the surface areas.3 The anterior border zone was found to be most commonly involved. The border zone distribution of 210 metastatic lesions is shown in Figure 2. The gray and white matter junction was the preferred site for 135 of 210 lesions (64%).Gray matter and white matter metastasis accounted for 11% and 16%, respectively. The remaining 18 lesions (9%)were uncertain (Fig. 3). The estimated mean number of metastases in gray and white matter junction was 3.48 times greater than the estimated mean number of tumors in gray matter ( P < 0.001) and 5.37 times greater than the estimated mean number of tumors in white matter ( P < 0.001). Finally, the estimated mean number of metastases in the border zones was 86% greater than the estimated mean number of tumors in non border zones ( P < 0.011). This analysis did not take into account the size of the border zones.

DISCUSSION
Since this is a retrospective study, some methodologic deficiencies deserve mention. We reviewed only 124 of 164 patients because some of the medical records or images were unavailable. Among 124 patients, only 100 met the requirement for final analysis. Some of the patients

1554

CANCER April 15, 1996 / Volume 77 / Number 8

which supply the cerebral cortex as they enter the white matter.4 This gray and white matter junction region is probably the site of tumor emboli impaction. Border zone or watershed areas are formed by the terminal capillary beds of cerebral arteries. There are many end-to-end anastomoses between the capillaries of different arteries, and also some pial or precapillary arteriolar anastomoses. Arterial pressure is lower in the border zones owing to the arborization and reduction of arteriolar diameter. These distal fields receive the lowest cerebral blood flow and are the first areas to suffer ischemia and infarction during systemic hypotension,20intracranial or extracranial arterial occlusion,L6 stenosis. The capillary netor work and its low blood flow are probably the causes of precipitation of the tumor emboli. The human cerebrovascular arrangement is not uniform; variations in the size and lengths of the vessel, the peripheral resistance of the arteries, and the potential for collateral flow from a distant source are pr~minent..~ Variations in the border zones is therefore e~pected.~,~ The border zones in the regions of the thalamus, basal ganglia, brainstem, pituitary gland, and pineal ganglia are too complicated and variable to be evaluated. For instance, thalamus is supplied by tuberothalamic, thalamoperforating, thalamogeniculate, and posterior choroidal a r t e r i e ~ , ~and ~ ~ . basal ganglia are supplied by recurrent artery of Heubner, medial and lateral lenticulostriate, and anterior choroidal arteries.26It is therefore difficult to define the border zones in these areas because of multiple vascular territories in relatively small regions and common variations in the vasculature.26This study was based on neuroimaging (CT scan and MRI) analysis, and assessed metastatic tumors in the major border zones of the brain only. The pituitary gland and basal ganglia have been conConversely, sidered to be common sites for metastasi~.~.~ the pineal gland has been considered a rare metastatic site. The metastasis to the basal ganglia, thalamus, and pituitary gland were under represented in our series. In our 105 patients, 4 had brain metastases in globus pallidus (1 case), thalamus (2 cases), and pituitary gland (1 case). These cases were excluded from the analysis. Among 302 evaluable metastatic lesions, there were only 4 in thalamus (1.3%), 1 in the pineal gland (0.3%); 1 in the pituitary gland (0.3%),and none in the basal ganglia.

TABLE 1
The Site and Tissue Type of Primary Cancers for 105 Patients with Brain Metastases
~ ~~

Site

Tissue type

No. of patients
79 18 27 8 26 6 3
1

Percent
(75%)

Lung Adenocarcinoma Squarnous cell carcinoma Small cell carcinoma Nan-small cell carcinoma Carcinoma Melanoma Renal cell carcinoma Carcinoma Hepatoma Choriocarcinoma Adenocarcinoma Mesothelioma Adenocarcinoma Squamous cell carcinoma Carcinoma, poorly differentiated Total

Breast Skin Kidney Ovary Liver Placenta Cervix Pleura Unknown

(5.7%) (3%)

1
1

2 1

I
10 5
L

(9.5%)

3
105

in the Cancer Registry were referred from other hospitals solely for radiation therapy. All of the above allowed possible sample bias. In our example, we expected the counts to show more variation under the standard model for count data (Poisson) for two reasons: (1) tumor lesions occur in clusters in individuals; and (2) lesion counts for a given individual (eight separate counts for each subject) are positively correlated. There are sophisticated models for overdispersed Poisson data, but a simple model captures all of the overdispersion in a single parameter. Lung carcinoma was the predominant primary cancer (75%),which reflected the cancer type distribution at this institution. Single (53%) and multiple (47%) metastases were evenly divided, which was in agreement with previous reports.I3 This study demonstrated that brain metastasis occurs in border zone regions of the cerebral vascular supply (62%) and in the gray and white matter junction (64%) more frequently than previously recognized. This supports the notion that tumor emboli tend to pass along the arterial tree as far distally as their size permits (border zones) or to lodge in the region with sudden reduction of vascular caliber (gray/white matter junction). Experimental evidence suggests that clusters of cells are necessary for the formation of tumor emboli and the subsequent development of metastases.I8 The size of tumor emboli was estimated to be about 100 um to 200 um in greatest dimension. These may effect arteries or arterioles from 50 um to 150 um in greater dimension.s There is a rather sudden narrowing the diameter of the arterioles

REFERENCES
Tikhtman AJ, Patchell RA. Brain metastasis. In: Morantz FL4, Walsh JW, editors. Brain tumors. New York Marcel Dekker, Inc., 1995;553-6. 2. Cairncross JG, Posner JB. The management of brain metastases. In: Walker MD, editor. Oncology of nervous system. Boston: Martinus Nijhoff Publishers, 1983:341-77. Krol G, Thaler HT, Posner JB. Distribution of 3. Delattre JY, brain metastases. Arch Neurol 1988;45:741-4.
1.

Preferred Sites of Brain Metastasis/Hwang et al.

4. 5.

1555

6.

7.

8.

9.

10. 11.

12.

13.

14. 15.

16.

Russell DS, Rubinstein LJ. Pathology of tumors of the nervous system. 5th ed. Baltimore: Williams & Wilkins, 1989. Henson RA, Urich H. Cancer and the nervous system. The neurological manifestation of systemic malignant disease. Boston: Rlackwell Scientific Publications, 1982. Damasio H. A computerized tomographic guide to the identification of cerebral vascular territories. Arch Neurol 1983;40:138-42. Berman SA, Hayman LA, Hinck VC. Correlation of CT cerebral vascular territories with function: 1. Anterior cerebral artery. AJR 1980; 135253-7. Hayman LA, Berman SA, Hinck VC. Correlation of CT cerebral vascular territories with function: 11. Posterior cerebral artery. A/ 1981;2:219-25. Bogousslansky 1, Regli F. Centrum ovale infarcts: Sub-cortical infarction in the superficial territory of the middle cerebral artery. Neurology 1992;42:1992-8. Torvik A. The pathogenesis of watershed infarcts in the brain. Srroke 1984; 15:221-3. Bogousslavsky J, Regli F. Border zone infarctions distal to internal carotid artery occlusion: prognostic implications. Ann Neurol 1986;20:346-50. Weiller C, Ringelstein EB, Reiche W, Buell U. Clinical and heinodynamic aspects of low-flow infarcts. Srroke 1991;22:1117-23. Bogousslavsky J. Subcortical infarcts. In: Fisher M. Bogousslavsky J, editors. Current Review of Cerebrovascular Disease. 1st ed. Philadelphia: Current Medicine, Inc., 1993:31-40. Bogousslavsky J, Regli F. Unilateral watershed cerebral infarcts. Neurolow/ 1986;36:373-7. Waterston JA, Brown MM, Butler P, Swash M. Small deep cerebral infarcts associated with occlusive internal carotid artery disease. Arch Neurol 1990;47:953-7. Angeloni U, Bozzao L, Fantozzi L, Bastianello S, Kushner M, Fieschi C. Internal border zone infarction following acute

middle cerebral artery occlusion. Neurology 1990;40:11968.

17. Breslow NE. Extra-Poisson variation in log-linear models. Appl Stat 1984;33:38-44. 18. Watanabe S. The metastasizability of tumor cells. Cancer 1954;7215-23. 19. Eecken HV, Adams RD. The anatomy and functional significance of the meningeal arterial anastomoses of the human brain. J Neuropnthol Exp Neurol 1953;12:132-57. 20. Penning B. Pathophysiology of stroke. Neuroimag Clin N A m 1992;2:389-408. 21. Ringlestein EB, Zoumer H, Anglou D. The pathogenesis of stroke from internal carotid occlusion. Diagnostic and therapeutic implications. Stroke 1983;14967-75. 22. Moody DM, Bell MA, Challa VR. Features of the cerebral vascular pattern that predict vulnerability to perfusion or oxygenation deficiency: an anatomic study. AJNR 1990; 11:431-9. 23. Van der Zwan A, Hillen B, Tulleken CAF, Dujovny M, Dragovic L. Variability of the territories of the major cerebral arteries. J Neurosurg 1992;77:927-40. 24. Lang EW, Daffertshofer M, Daffertshofer A, Wirth SB, Chesnut RM, Hennerici M. Variability vascular territory in stroke pitfalls and failure of stroke pattern interpretation. Stroke 1995;26:942-5. 25. Moulin T, Bogousslavsky J. Vertebrobasilar stroke. In: Fisher M, BogousslavskyJ, editors. Current review of cerebrovascular disease. 2nd ed. Philadelphia: Current Medicine, 1996: 41-58. 26. Pullicino PM. The course and territories of cerebral small arteries. In: Pullicino PM, Caplan LR, Hommel M, editors. Advances in Neurology. New York Raven Press, 1993;62:1139. 27. Paillas JE, Pellet W. Brain Metastases. In: Vinken PJ, Bruyn GW, editors. Handbook of Clinical Neurology. Vol. 18. New York American Elsevier Publishing Co., Inc., 1975201-32.